Refactor(beta): extract visible_document_ids_for_user helper

The owner-aware "resolve user to visible document pks" block was duplicated
verbatim between get_context_for_document and get_taxonomy_hints_for_document.
Extract it into indexing.visible_document_ids_for_user, next to its sibling
normalize_document_ids, and call it from both paths.

No behavior change: the helper returns None when user is None (unfiltered
retrieval) and the same pk list otherwise.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

pyproject.toml

@@ -42,7 +42,6 @@ dependencies = [
   "drf-spectacular~=0.28",
   "drf-spectacular-sidecar~=2026.5.1",
   "drf-writable-nested~=0.7.1",
-  "faiss-cpu>=1.10",
   "filelock~=3.29.0",
   "flower~=2.0.1",
   "gotenberg-client~=0.14.0",
@@ -57,7 +56,6 @@ dependencies = [
   "llama-index-embeddings-openai-like>=0.2.2",
   "llama-index-llms-ollama>=0.9.1",
   "llama-index-llms-openai-like>=0.7.1",
-  "llama-index-vector-stores-faiss>=0.5.2",
   "nltk~=3.9.1",
   "ocrmypdf~=17.4.2",
   "openai>=2.32",
@@ -74,6 +72,7 @@ dependencies = [
   "scikit-learn~=1.8.0",
   "sentence-transformers>=5.4.1",
   "setproctitle~=1.3.4",
+  "sqlite-vec==0.1.9",
   "tantivy~=0.26.0",
   "tika-client~=0.11.0",
   "torch~=2.11.0",

src-ui/src/app/components/admin/tasks/tasks.component.html

@@ -11,6 +11,9 @@
     <button class="btn btn-sm btn-outline-primary me-2" (click)="dismissTasks()" *pngxIfPermissions="{ action: PermissionAction.Change, type: PermissionType.PaperlessTask }" [disabled]="visibleTasks.length === 0">
       <i-bs name="check2-all" class="me-1"></i-bs>{{dismissButtonText}}
     </button>
+    <button class="btn btn-sm btn-outline-primary me-2" (click)="dismissAllTasks()" *pngxIfPermissions="{ action: PermissionAction.Change, type: PermissionType.PaperlessTask }" [disabled]="totalTasks === 0">
+      <i-bs name="check2-all" class="me-1"></i-bs><ng-container i18n>Dismiss all</ng-container>
+    </button>
     <div class="form-check form-switch mb-0 ms-2">
       <input class="form-check-input" type="checkbox" role="switch" [(ngModel)]="autoRefreshEnabled">
       <label class="form-check-label" for="autoRefreshSwitch" i18n>Auto refresh</label>
@@ -81,7 +84,7 @@
         <button class="btn btn-sm btn-outline-primary" ngbDropdownToggle>{{filterTargetName}}</button>
         <div class="dropdown-menu shadow" ngbDropdownMenu>
           @for (t of filterTargets; track t.id) {
-            <button ngbDropdownItem [class.active]="filterTargetID === t.id" (click)="filterTargetID = t.id">{{t.name}}</button>
+            <button ngbDropdownItem [class.active]="filterTargetID === t.id" (click)="setFilterTarget(t.id)">{{t.name}}</button>
           }
         </div>
       </div>

src-ui/src/app/components/admin/tasks/tasks.component.spec.ts

@@ -11,7 +11,7 @@ import { Router } from '@angular/router'
 import { RouterTestingModule } from '@angular/router/testing'
 import { NgbModal, NgbModalRef, NgbModule } from '@ng-bootstrap/ng-bootstrap'
 import { allIcons, NgxBootstrapIconsModule } from 'ngx-bootstrap-icons'
-import { throwError } from 'rxjs'
+import { of, throwError } from 'rxjs'
 import { routes } from 'src/app/app-routing.module'
 import {
   PaperlessTask,
@@ -29,7 +29,11 @@ import { ToastService } from 'src/app/services/toast.service'
 import { environment } from 'src/environments/environment'
 import { ConfirmDialogComponent } from '../../common/confirm-dialog/confirm-dialog.component'
 import { PageHeaderComponent } from '../../common/page-header/page-header.component'
-import { TasksComponent, TaskSection } from './tasks.component'
+import {
+  TaskFilterTargetID,
+  TasksComponent,
+  TaskSection,
+} from './tasks.component'
 
 const tasks: PaperlessTask[] = [
   {
@@ -154,6 +158,13 @@ const paginatedTasks: Results<PaperlessTask> = {
   results: tasks,
 }
 
+const sectionCountResponse = {
+  all: 7,
+  needs_attention: 2,
+  in_progress: 3,
+  completed: 2,
+}
+
 describe('TasksComponent', () => {
   let component: TasksComponent
   let fixture: ComponentFixture<TasksComponent>
@@ -221,6 +232,15 @@ describe('TasksComponent', () => {
           req.params.get('page') === '1'
       )
       .flush(paginatedTasks)
+
+    httpTestingController
+      .expectOne(
+        (req) =>
+          req.url === `${environment.apiBaseUrl}tasks/status_counts/` &&
+          req.params.get('acknowledged') === 'false' &&
+          !req.params.has('status')
+      )
+      .flush(sectionCountResponse)
   })
 
   it('should display task sections with counts', () => {
@@ -295,6 +315,7 @@ describe('TasksComponent', () => {
     const headerText = header.nativeElement.textContent
 
     expect(headerText).toContain('Dismiss visible')
+    expect(headerText).toContain('Dismiss all')
     expect(headerText).toContain('Auto refresh')
     expect(headerText).not.toContain('All types')
     expect(headerText).not.toContain('All sources')
@@ -327,6 +348,74 @@ describe('TasksComponent', () => {
     expect(pagination).not.toBeNull()
   })
 
+  it('should apply the selected section to the server-side task query', () => {
+    component.setSection(TaskSection.NeedsAttention)
+
+    const req = httpTestingController.expectOne(
+      (request) =>
+        request.url === `${environment.apiBaseUrl}tasks/` &&
+        request.params.get('page') === '1' &&
+        request.params.get('page_size') === '25' &&
+        request.params.get('acknowledged') === 'false' &&
+        request.params.getAll('status').includes(PaperlessTaskStatus.Failure) &&
+        request.params.getAll('status').includes(PaperlessTaskStatus.Revoked)
+    )
+
+    req.flush({ count: 2, results: [tasks[0], tasks[1]] })
+    expect(component.totalTasks).toBe(2)
+  })
+
+  it('should apply task type and trigger source filters to the server-side task query', () => {
+    component.setTaskType(PaperlessTaskType.SanityCheck)
+
+    httpTestingController
+      .expectOne(
+        (request) =>
+          request.url === `${environment.apiBaseUrl}tasks/` &&
+          request.params.get('page_size') === '25' &&
+          request.params.get('task_type') === PaperlessTaskType.SanityCheck
+      )
+      .flush({ count: 1, results: [tasks[6]] })
+
+    component.setTriggerSource(PaperlessTaskTriggerSource.System)
+
+    httpTestingController
+      .expectOne(
+        (request) =>
+          request.url === `${environment.apiBaseUrl}tasks/` &&
+          request.params.get('page_size') === '25' &&
+          request.params.get('task_type') === PaperlessTaskType.SanityCheck &&
+          request.params.get('trigger_source') ===
+            PaperlessTaskTriggerSource.System
+      )
+      .flush({ count: 1, results: [tasks[6]] })
+  })
+
+  it('should apply text filters to the server-side task query', () => {
+    component.filterText = 'invoice'
+    jest.advanceTimersByTime(150)
+
+    httpTestingController
+      .expectOne(
+        (request) =>
+          request.url === `${environment.apiBaseUrl}tasks/` &&
+          request.params.get('page_size') === '25' &&
+          request.params.get('name') === 'invoice'
+      )
+      .flush({ count: 1, results: [tasks[0]] })
+
+    component.setFilterTarget(TaskFilterTargetID.Result)
+
+    httpTestingController
+      .expectOne(
+        (request) =>
+          request.url === `${environment.apiBaseUrl}tasks/` &&
+          request.params.get('page_size') === '25' &&
+          request.params.get('result') === 'invoice'
+      )
+      .flush({ count: 0, results: [] })
+  })
+
   it('should load a different task page when pagination changes', () => {
     component.setPage(2)
 
@@ -350,6 +439,27 @@ describe('TasksComponent', () => {
     expect(component.pagedTasks).toEqual([tasks[0]])
   })
 
+  it('should not replace section counts with current-page counts', () => {
+    component.setPage(2)
+
+    httpTestingController
+      .expectOne(
+        (req) =>
+          req.url === `${environment.apiBaseUrl}tasks/` &&
+          req.params.get('acknowledged') === 'false' &&
+          req.params.get('page_size') === '25' &&
+          req.params.get('page') === '2'
+      )
+      .flush({
+        count: 30,
+        results: [tasks[0]],
+      })
+
+    expect(component.sectionCount(TaskSection.NeedsAttention)).toBe(2)
+    expect(component.sectionCount(TaskSection.InProgress)).toBe(3)
+    expect(component.sectionCount(TaskSection.Completed)).toBe(2)
+  })
+
   it('should expose stable task type options and disable empty ones', () => {
     expect(component.taskTypeOptions.map((option) => option.value)).toContain(
       PaperlessTaskType.TrainClassifier
@@ -495,6 +605,46 @@ describe('TasksComponent', () => {
     expect(dismissSpy).toHaveBeenCalledWith(new Set([467, 466]))
   })
 
+  it('should support dismiss all tasks', () => {
+    let modal: NgbModalRef
+    modalService.activeInstances.subscribe((m) => (modal = m[m.length - 1]))
+    const dismissSpy = jest
+      .spyOn(tasksService, 'dismissAllTasks')
+      .mockReturnValue(of({}))
+    const reloadPageSpy = jest
+      .spyOn(component as any, 'reloadPage')
+      .mockImplementation(() => undefined)
+
+    component.dismissAllTasks()
+
+    expect(modal).not.toBeUndefined()
+    expect(modal.componentInstance.messageBold).toBe('Dismiss all 7 tasks?')
+    modal.componentInstance.confirmClicked.emit()
+    expect(dismissSpy).toHaveBeenCalled()
+    expect(reloadPageSpy).toHaveBeenCalledWith(false)
+    expect(component.selectedTasks.size).toBe(0)
+  })
+
+  it('should show an error and re-enable modal buttons when dismissing all tasks fails', () => {
+    const error = new Error('dismiss all failed')
+    const toastSpy = jest.spyOn(toastService, 'showError')
+    const dismissSpy = jest
+      .spyOn(tasksService, 'dismissAllTasks')
+      .mockReturnValue(throwError(() => error))
+
+    let modal: NgbModalRef
+    modalService.activeInstances.subscribe((m) => (modal = m[m.length - 1]))
+
+    component.dismissAllTasks()
+    expect(modal).not.toBeUndefined()
+
+    modal.componentInstance.confirmClicked.emit()
+
+    expect(dismissSpy).toHaveBeenCalled()
+    expect(toastSpy).toHaveBeenCalledWith('Error dismissing tasks', error)
+    expect(modal.componentInstance.buttonsEnabled).toBe(true)
+  })
+
   it('should dismiss the currently visible scoped and filtered tasks', () => {
     component.setSection(TaskSection.InProgress)
     component.setTaskType(PaperlessTaskType.SanityCheck)
@@ -673,6 +823,9 @@ describe('TasksComponent', () => {
   })
 
   it('should keep clearing selection independent from resetting filters', () => {
+    component.resetFilter()
+    expect(component.filterText).toBe('')
+
     component.setTaskType(PaperlessTaskType.ConsumeFile)
     component.toggleSelected(tasks[0])
     expect(component.selectedTasks.size).toBe(1)

src-ui/src/app/components/admin/tasks/tasks.component.ts

@@ -40,7 +40,7 @@ export enum TaskSection {
   Completed = 'completed',
 }
 
-enum TaskFilterTargetID {
+export enum TaskFilterTargetID {
   Name,
   Result,
 }
@@ -167,6 +167,12 @@ export class TasksComponent
   public readonly pageSize = 25
   public page: number = 1
   public totalTasks: number = 0
+  public sectionCounts: Record<TaskSection, number> = {
+    [TaskSection.All]: 0,
+    [TaskSection.NeedsAttention]: 0,
+    [TaskSection.InProgress]: 0,
+    [TaskSection.Completed]: 0,
+  }
   public pagedTasks: PaperlessTask[] = []
   public selectedSection: TaskSection = TaskSection.All
   public selectedTaskType: PaperlessTaskType | null = null
@@ -282,6 +288,7 @@ export class TasksComponent
       .subscribe((query) => {
         this._filterText = query
         this.clearSelection()
+        this.reloadPage(true)
       })
   }
 
@@ -334,6 +341,30 @@ export class TasksComponent
     }
   }
 
+  dismissAllTasks() {
+    let modal = this.modalService.open(ConfirmDialogComponent, {
+      backdrop: 'static',
+    })
+    modal.componentInstance.title = $localize`Confirm Dismiss All`
+    modal.componentInstance.messageBold = $localize`Dismiss all ${this.totalTasks} tasks?`
+    modal.componentInstance.btnClass = 'btn-warning'
+    modal.componentInstance.btnCaption = $localize`Dismiss`
+    modal.componentInstance.confirmClicked.pipe(first()).subscribe(() => {
+      modal.componentInstance.buttonsEnabled = false
+      modal.close()
+      this.tasksService.dismissAllTasks().subscribe({
+        next: () => {
+          this.reloadPage(false)
+        },
+        error: (e) => {
+          this.toastService.showError($localize`Error dismissing tasks`, e)
+          modal.componentInstance.buttonsEnabled = true
+        },
+      })
+      this.clearSelection()
+    })
+  }
+
   expandTask(task: PaperlessTask) {
     this.expandedTask = this.expandedTask == task.id ? undefined : task.id
   }
@@ -446,9 +477,7 @@ export class TasksComponent
   }
 
   sectionCount(section: TaskSection): number {
-    return this.pagedTasks.filter((task) =>
-      this.taskBelongsToSection(task, section)
-    ).length
+    return this.sectionCounts[section]
   }
 
   sectionShowsResults(section: TaskSection): boolean {
@@ -458,16 +487,27 @@ export class TasksComponent
   setSection(section: TaskSection) {
     this.selectedSection = section
     this.clearSelection()
+    this.reloadPage(true)
   }
 
   setTaskType(taskType: PaperlessTaskType | null) {
     this.selectedTaskType = taskType
     this.clearSelection()
+    this.reloadPage(true)
   }
 
   setTriggerSource(triggerSource: PaperlessTaskTriggerSource | null) {
     this.selectedTriggerSource = triggerSource
     this.clearSelection()
+    this.reloadPage(true)
+  }
+
+  setFilterTarget(filterTargetID: TaskFilterTargetID) {
+    this.filterTargetID = filterTargetID
+    if (this._filterText.length) {
+      this.clearSelection()
+      this.reloadPage(true)
+    }
   }
 
   taskTypeOptionCount(taskType: PaperlessTaskType | null): number {
@@ -505,19 +545,32 @@ export class TasksComponent
   }
 
   public resetFilter() {
+    if (!this._filterText.length) {
+      return
+    }
+
     this._filterText = ''
+    this.clearSelection()
+    this.reloadPage(true)
   }
 
   public resetFilters() {
+    const hadFilter = this.isFiltered
     this.selectedTaskType = null
     this.selectedTriggerSource = null
-    this.resetFilter()
+    this._filterText = ''
     this.clearSelection()
+
+    if (hadFilter) {
+      this.reloadPage(true)
+    }
   }
 
   filterInputKeyup(event: KeyboardEvent) {
     if (event.key == 'Enter') {
       this._filterText = (event.target as HTMLInputElement).value
+      this.clearSelection()
+      this.reloadPage(true)
     } else if (event.key === 'Escape') {
       this.resetFilter()
     }
@@ -606,19 +659,86 @@ export class TasksComponent
     )
   }
 
+  private reloadSectionCounts() {
+    this.tasksService
+      .statusCounts(this.getParamsForSection(TaskSection.All))
+      .pipe(first(), takeUntil(this.unsubscribeNotifier))
+      .subscribe((counts) => {
+        this.sectionCounts[TaskSection.All] = counts.all
+        this.sectionCounts[TaskSection.NeedsAttention] = counts.needs_attention
+        this.sectionCounts[TaskSection.InProgress] = counts.in_progress
+        this.sectionCounts[TaskSection.Completed] = counts.completed
+      })
+  }
+
+  private getParamsForSection(
+    section: TaskSection
+  ): Record<string, string | number | boolean | readonly string[]> {
+    const params: Record<
+      string,
+      string | number | boolean | readonly string[]
+    > = {
+      acknowledged: false,
+    }
+
+    const statuses = this.statusesForSection(section)
+    if (statuses.length) {
+      params.status = statuses
+    }
+
+    if (this.selectedTaskType !== null) {
+      params.task_type = this.selectedTaskType
+    }
+
+    if (this.selectedTriggerSource !== null) {
+      params.trigger_source = this.selectedTriggerSource
+    }
+
+    if (this._filterText.length) {
+      params[
+        this.filterTargetID === TaskFilterTargetID.Name ? 'name' : 'result'
+      ] = this._filterText
+    }
+
+    return params
+  }
+
+  private statusesForSection(section: TaskSection): PaperlessTaskStatus[] {
+    switch (section) {
+      case TaskSection.NeedsAttention:
+        return [PaperlessTaskStatus.Failure, PaperlessTaskStatus.Revoked]
+      case TaskSection.InProgress:
+        return [PaperlessTaskStatus.Pending, PaperlessTaskStatus.Started]
+      case TaskSection.Completed:
+        return [PaperlessTaskStatus.Success]
+      default:
+        return []
+    }
+  }
+
   private reloadPage(resetToFirstPage: boolean = false) {
     if (resetToFirstPage) {
       this.page = 1
     }
 
+    this.reloadSectionCounts()
+
     this.loading = true
     this.tasksService
-      .list(this.page, this.pageSize, { acknowledged: false })
+      .list(
+        this.page,
+        this.pageSize,
+        this.getParamsForSection(this.selectedSection)
+      )
       .pipe(first(), takeUntil(this.unsubscribeNotifier))
       .subscribe({
         next: (result) => {
           this.pagedTasks = result.results
           this.totalTasks = result.count
+          this.sectionCounts[TaskSection.All] = result.count
+          if (this.selectedSection !== TaskSection.All) {
+            this.sectionCounts[this.selectedSection] = result.count
+          }
           this.loading = false
           if (
             this.page > 1 &&

src-ui/src/app/components/chat/chat/chat.component.html

@@ -8,7 +8,7 @@
       <div class="chat-messages font-monospace small">
         @for (message of messages; track message) {
           <div class="message d-flex flex-row small" [class.justify-content-end]="message.role === 'user'">
-            <div class="p-2 m-2" [class.bg-dark]="message.role === 'user'">
+            <div class="p-2 m-2" [class.bg-body]="message.role === 'user'">
               <span>
                 {{ message.content }}
                 @if (message.isStreaming) { <span class="blinking-cursor">|</span> }

src-ui/src/app/components/chat/chat/chat.component.spec.ts

@@ -188,4 +188,14 @@ describe('ChatComponent', () => {
     component.searchInputKeyDown(event)
     expect(component.sendMessage).toHaveBeenCalled()
   })
+
+  it('should not send message on Enter key press while composing with IME', () => {
+    jest.spyOn(component, 'sendMessage')
+    const event = new KeyboardEvent('keydown', {
+      key: 'Enter',
+      isComposing: true,
+    })
+    component.searchInputKeyDown(event)
+    expect(component.sendMessage).not.toHaveBeenCalled()
+  })
 })

src-ui/src/app/components/chat/chat/chat.component.ts

@@ -155,7 +155,10 @@ export class ChatComponent implements OnInit {
   }
 
   public searchInputKeyDown(event: KeyboardEvent) {
-    if (event.key === 'Enter') {
+    if (
+      event.key === 'Enter' &&
+      !(event.isComposing || event.keyCode === 229)
+    ) {
       event.preventDefault()
       this.sendMessage()
     }

src-ui/src/app/components/common/confirm-dialog/confirm-dialog.component.html

@@ -5,10 +5,10 @@
 </div>
 <div class="modal-body">
   @if (messageBold) {
-    <p><b>{{messageBold}}</b></p>
+    <p class="text-break"><b>{{messageBold}}</b></p>
   }
   @if (message) {
-    <p class="mb-0" [innerHTML]="message"></p>
+    <p class="mb-0 text-break" [innerHTML]="message"></p>
   }
 </div>
 <div class="modal-footer">

src-ui/src/app/components/common/confirm-dialog/merge-confirm-dialog/merge-confirm-dialog.component.html

@@ -9,8 +9,11 @@
         <label class="form-label" for="metadataDocumentID" i18n>Documents:</label>
         <ul class="list-group"
             cdkDropList
+            [cdkDropListData]="documentIDs"
             (cdkDropListDropped)="onDrop($event)">
-            @for (document of documents; track document.id) {
+            @for (documentID of documentIDs; track documentID) {
+              @let document = getDocument(documentID);
+              @if (document) {
                 <li class="list-group-item d-flex align-items-center" cdkDrag>
                     <i-bs name="grip-vertical" class="me-2"></i-bs>
                     <div class="d-flex flex-column">
@@ -27,6 +30,7 @@
                         </small>
                     </div>
                 </li>
+              }
             }
         </ul>
         </div>

src-ui/src/app/components/common/suggestions-dropdown/suggestions-dropdown.component.html

@@ -1,5 +1,5 @@
 <div class="btn-group">
-  <button type="button" class="btn btn-sm btn-outline-primary" (click)="clickSuggest()" [disabled]="loading || (suggestions && !aiEnabled)">
+  <button type="button" class="btn btn-sm btn-outline-primary" (click)="clickSuggest()" [disabled]="disabled || loading || (suggestions && !aiEnabled)">
     @if (loading) {
       <div class="spinner-border spinner-border-sm" role="status"></div>
     } @else {
@@ -13,7 +13,7 @@
 
   @if (aiEnabled) {
     <div class="btn-group" ngbDropdown #dropdown="ngbDropdown" [popperOptions]="popperOptions">
-      <button type="button" class="btn btn-sm btn-outline-primary" ngbDropdownToggle [disabled]="loading || !suggestions" aria-expanded="false" aria-controls="suggestionsDropdown" aria-label="Suggestions dropdown">
+      <button type="button" class="btn btn-sm btn-outline-primary" ngbDropdownToggle [disabled]="disabled || loading || !suggestions" aria-expanded="false" aria-controls="suggestionsDropdown" aria-label="Suggestions dropdown">
         <span class="visually-hidden" i18n>Show suggestions</span>
       </button>
 

src-ui/src/app/components/common/suggestions-dropdown/suggestions-dropdown.component.spec.ts

@@ -37,6 +37,18 @@ describe('SuggestionsDropdownComponent', () => {
     expect(component.getSuggestions.emit).toHaveBeenCalled()
   })
 
+  it('should not emit getSuggestions when disabled', () => {
+    jest.spyOn(component.getSuggestions, 'emit')
+    component.disabled = true
+    component.suggestions = null
+    fixture.detectChanges()
+
+    component.clickSuggest()
+
+    expect(component.getSuggestions.emit).not.toHaveBeenCalled()
+    expect(fixture.nativeElement.querySelector('button').disabled).toBeTruthy()
+  })
+
   it('should toggle dropdown when clickSuggest is called and suggestions are not null', () => {
     component.aiEnabled = true
     fixture.detectChanges()

src-ui/src/app/components/common/suggestions-dropdown/suggestions-dropdown.component.ts

@@ -47,6 +47,14 @@ export class SuggestionsDropdownComponent {
   addCorrespondent: EventEmitter<string> = new EventEmitter()
 
   public clickSuggest(): void {
+    if (
+      this.disabled ||
+      this.loading ||
+      (this.suggestions && !this.aiEnabled)
+    ) {
+      return
+    }
+
     if (!this.suggestions) {
       this.getSuggestions.emit(this)
     } else {

src-ui/src/app/data/paperless-task.ts

@@ -64,3 +64,10 @@ export interface PaperlessTaskSummary {
   last_success: Date | null
   last_failure: Date | null
 }
+
+export interface PaperlessTaskStatusCounts {
+  all: number
+  needs_attention: number
+  in_progress: number
+  completed: number
+}

src-ui/src/app/services/tasks.service.spec.ts

@@ -80,6 +80,27 @@ describe('TasksService', () => {
       .flush({ count: 0, results: [] })
   })
 
+  it('calls acknowledge_tasks api endpoint on dismiss all and reloads', () => {
+    tasksService.dismissAllTasks().subscribe()
+    const req = httpTestingController.expectOne(
+      `${environment.apiBaseUrl}tasks/acknowledge/`
+    )
+    expect(req.request.method).toEqual('POST')
+    expect(req.request.body).toEqual({
+      all: true,
+    })
+    req.flush([])
+    // reload is then called
+    httpTestingController
+      .expectOne(
+        (req: HttpRequest<unknown>) =>
+          req.url === `${environment.apiBaseUrl}tasks/` &&
+          req.params.get('acknowledged') === 'false' &&
+          req.params.get('page_size') === '1000'
+      )
+      .flush({ count: 0, results: [] })
+  })
+
   it('groups mixed task types by status when reloading', () => {
     expect(tasksService.total).toEqual(0)
     const mockTasks = [
@@ -221,4 +242,34 @@ describe('TasksService', () => {
       task_id: 'abc-123',
     })
   })
+
+  it('loads filtered task status counts', () => {
+    tasksService
+      .statusCounts({
+        acknowledged: false,
+        task_type: PaperlessTaskType.ConsumeFile,
+      })
+      .subscribe((res) => {
+        expect(res).toEqual({
+          all: 10,
+          needs_attention: 2,
+          in_progress: 3,
+          completed: 5,
+        })
+      })
+
+    const req = httpTestingController.expectOne(
+      (req: HttpRequest<unknown>) =>
+        req.url === `${environment.apiBaseUrl}tasks/status_counts/` &&
+        req.params.get('acknowledged') === 'false' &&
+        req.params.get('task_type') === PaperlessTaskType.ConsumeFile
+    )
+    expect(req.request.method).toEqual('GET')
+    req.flush({
+      all: 10,
+      needs_attention: 2,
+      in_progress: 3,
+      completed: 5,
+    })
+  })
 })

src-ui/src/app/services/tasks.service.ts

@@ -5,6 +5,7 @@ import { first, map, takeUntil, tap } from 'rxjs/operators'
 import {
   PaperlessTask,
   PaperlessTaskStatus,
+  PaperlessTaskStatusCounts,
   PaperlessTaskType,
 } from 'src/app/data/paperless-task'
 import { Results } from 'src/app/data/results'
@@ -88,7 +89,7 @@ export class TasksService {
   public list(
     page: number,
     pageSize: number,
-    extraParams?: Record<string, string | number | boolean>
+    extraParams?: Record<string, string | number | boolean | readonly string[]>
   ): Observable<Results<PaperlessTask>> {
     return this.http.get<Results<PaperlessTask>>(
       `${this.baseUrl}${this.endpoint}/`,
@@ -102,6 +103,17 @@ export class TasksService {
     )
   }
 
+  public statusCounts(
+    extraParams?: Record<string, string | number | boolean | readonly string[]>
+  ): Observable<PaperlessTaskStatusCounts> {
+    return this.http.get<PaperlessTaskStatusCounts>(
+      `${this.baseUrl}${this.endpoint}/status_counts/`,
+      {
+        params: extraParams,
+      }
+    )
+  }
+
   public dismissTasks(task_ids: Set<number>): Observable<any> {
     return this.http
       .post(`${this.baseUrl}tasks/acknowledge/`, {
@@ -116,6 +128,20 @@ export class TasksService {
       )
   }
 
+  public dismissAllTasks(): Observable<any> {
+    return this.http
+      .post(`${this.baseUrl}tasks/acknowledge/`, {
+        all: true,
+      })
+      .pipe(
+        first(),
+        takeUntil(this.unsubscribeNotifer),
+        tap(() => {
+          this.reload()
+        })
+      )
+  }
+
   public cancelPending(): void {
     this.unsubscribeNotifer.next(true)
   }

src/documents/bulk_edit.py

@@ -904,6 +904,19 @@ def remove_password(
                 doc.id,
                 pair.source_doc.source_path,
             )
+            try:
+                with pikepdf.open(source_path) as pdf:
+                    if not pdf.is_encrypted:
+                        logger.info(
+                            "Skipping password removal for document %s because the "
+                            "source PDF is not encrypted",
+                            pair.root_doc.id,
+                        )
+                        continue
+            except pikepdf.PasswordError:
+                # Password-protected PDFs need the supplied password below.
+                pass
+
             with pikepdf.open(source_path, password=password) as pdf:
                 filepath: Path = (
                     Path(tempfile.mkdtemp(dir=settings.SCRATCH_DIR))

src/documents/filters.py

@@ -28,6 +28,7 @@ from django.db.models.functions import Cast
 from django.utils.translation import gettext_lazy as _
 from django_filters import DateFilter
 from django_filters.rest_framework import BooleanFilter
+from django_filters.rest_framework import CharFilter
 from django_filters.rest_framework import DateTimeFilter
 from django_filters.rest_framework import Filter
 from django_filters.rest_framework import FilterSet
@@ -900,6 +901,16 @@ class ShareLinkBundleFilterSet(FilterSet):
 
 
 class PaperlessTaskFilterSet(FilterSet):
+    name = CharFilter(
+        method="filter_name",
+        label="Name",
+    )
+
+    result = CharFilter(
+        method="filter_result",
+        label="Result",
+    )
+
     task_type = MultipleChoiceFilter(
         choices=PaperlessTask.TaskType.choices,
         label="Task Type",
@@ -939,7 +950,58 @@ class PaperlessTaskFilterSet(FilterSet):
 
     class Meta:
         model = PaperlessTask
-        fields = ["task_type", "trigger_source", "status", "acknowledged", "owner"]
+        fields = [
+            "task_type",
+            "trigger_source",
+            "status",
+            "acknowledged",
+            "owner",
+            "name",
+            "result",
+        ]
+
+    def filter_name(self, queryset, name, value):
+        if not value:
+            return queryset
+
+        matching_task_types = [
+            task_type
+            for task_type, label in PaperlessTask.TaskType.choices
+            if value.lower() in str(label).lower()
+        ]
+        matching_trigger_sources = [
+            trigger_source
+            for trigger_source, label in PaperlessTask.TriggerSource.choices
+            if value.lower() in str(label).lower()
+        ]
+
+        return queryset.filter(
+            Q(input_data__filename__icontains=value)
+            | Q(task_type__in=matching_task_types)
+            | Q(trigger_source__in=matching_trigger_sources),
+        )
+
+    def filter_result(self, queryset, name, value):
+        if not value:
+            return queryset
+
+        query = Q(result_data__reason__icontains=value) | Q(
+            result_data__error_message__icontains=value,
+        )
+
+        try:
+            numeric_value = int(value)
+        except (TypeError, ValueError):
+            pass
+        else:
+            query |= Q(result_data__document_id=numeric_value) | Q(
+                result_data__duplicate_of=numeric_value,
+            )
+
+        if "duplicate" in value.lower():
+            query |= Q(result_data__duplicate_of__isnull=False)
+
+        return queryset.filter(query)
 
     def filter_is_complete(self, queryset, name, value):
         if value:

src/documents/management/commands/document_llmindex.py

@@ -2,6 +2,7 @@ from typing import Any
 
 from documents.management.commands.base import PaperlessCommand
 from documents.tasks import llmindex_index
+from paperless_ai.indexing import llm_index_compact
 
 
 class Command(PaperlessCommand):
@@ -12,9 +13,12 @@ class Command(PaperlessCommand):
 
     def add_arguments(self, parser: Any) -> None:
         super().add_arguments(parser)
-        parser.add_argument("command", choices=["rebuild", "update"])
+        parser.add_argument("command", choices=["rebuild", "update", "compact"])
 
     def handle(self, *args: Any, **options: Any) -> None:
+        if options["command"] == "compact":
+            llm_index_compact()
+            return
         llmindex_index(
             rebuild=options["command"] == "rebuild",
             iter_wrapper=lambda docs: self.track(

src/documents/serialisers.py

@@ -48,6 +48,7 @@ from rest_framework import serializers
 from rest_framework.exceptions import PermissionDenied
 from rest_framework.fields import SerializerMethodField
 from rest_framework.filters import OrderingFilter
+from rest_framework.utils import model_meta
 
 if settings.AUDIT_LOG_ENABLED:
     from auditlog.context import set_actor
@@ -121,6 +122,45 @@ class DynamicFieldsModelSerializer(serializers.ModelSerializer[Any]):
                 self.fields.pop(field_name)
 
 
+class DocumentUpdateFieldsModelSerializer(DynamicFieldsModelSerializer):
+    stale_update_excluded_fields = frozenset({"filename", "archive_filename"})
+
+    def _get_update_fields(self, validated_data) -> list[str]:
+        model_fields = {
+            field.name
+            for field in self.Meta.model._meta.concrete_fields
+            if field.name not in self.stale_update_excluded_fields
+        }
+        update_fields = [
+            field_name for field_name in validated_data if field_name in model_fields
+        ]
+        if "modified" in model_fields and "modified" not in update_fields:
+            update_fields.append("modified")
+        return update_fields
+
+    def update(self, instance, validated_data):
+        serializers.raise_errors_on_nested_writes("update", self, validated_data)
+        info = model_meta.get_field_info(instance)
+
+        m2m_fields = []
+        for attr, value in validated_data.items():
+            if attr in info.relations and info.relations[attr].to_many:
+                m2m_fields.append((attr, value))
+            else:
+                setattr(instance, attr, value)
+
+        # File names are managed by post-save file handling.  Saving only the
+        # serializer-updated fields prevents stale in-memory path values from
+        # overwriting a concurrent move.
+        instance.save(update_fields=self._get_update_fields(validated_data))
+
+        for attr, value in m2m_fields:
+            field = getattr(instance, attr)
+            field.set(value)
+
+        return instance
+
+
 class MatchingModelSerializer(serializers.ModelSerializer[Any]):
     document_count = serializers.IntegerField(read_only=True)
 
@@ -989,7 +1029,7 @@ class DocumentVersionInfoSerializer(serializers.Serializer[_DocumentVersionInfo]
 class DocumentSerializer(
     OwnedObjectSerializer,
     NestedUpdateMixin,
-    DynamicFieldsModelSerializer,
+    DocumentUpdateFieldsModelSerializer,
 ):
     correspondent = CorrespondentField(allow_null=True)
     tags = TagsField(many=True)
@@ -1128,10 +1168,9 @@ class DocumentSerializer(
         return super().validate(attrs)
 
     def update(self, instance: Document, validated_data):
-        if "created_date" in validated_data and "created" not in validated_data:
-            instance.created = validated_data.get("created_date")
-            instance.save()
         if "created_date" in validated_data:
+            if "created" not in validated_data:
+                validated_data["created"] = validated_data["created_date"]
             logger.warning(
                 "created_date is deprecated, use created instead",
             )
@@ -1201,11 +1240,13 @@ class DocumentSerializer(
                     for tag in instance.tags.all()
                     if tag not in inbox_tags_not_being_added
                 ]
+
         if settings.AUDIT_LOG_ENABLED:
             with set_actor(self.user):
                 super().update(instance, validated_data)
         else:
             super().update(instance, validated_data)
+
         # hard delete custom field instances that were soft deleted
         CustomFieldInstance.deleted_objects.filter(document=instance).delete()
         return instance
@@ -2632,18 +2673,25 @@ class RunTaskSerializer(serializers.Serializer[dict[str, str]]):
 
 class AcknowledgeTasksViewSerializer(serializers.Serializer[dict[str, Any]]):
     tasks = serializers.ListField(
-        required=True,
+        required=False,
         label="Tasks",
         write_only=True,
         child=serializers.IntegerField(),
     )
+    all = serializers.BooleanField(
+        required=False,
+        default=False,
+        label="All",
+        write_only=True,
+    )
 
     def _validate_task_id_list(self, tasks, name="tasks") -> None:
         if not isinstance(tasks, list):
             raise serializers.ValidationError(f"{name} must be a list")
         if not all(isinstance(i, int) for i in tasks):
             raise serializers.ValidationError(f"{name} must be a list of integers")
-        count = PaperlessTask.objects.filter(id__in=tasks).count()
+        queryset = self.context.get("queryset", PaperlessTask.objects.all())
+        count = queryset.filter(id__in=tasks).count()
         if not count == len(tasks):
             raise serializers.ValidationError(
                 f"Some tasks in {name} don't exist or were specified twice.",
@@ -2653,6 +2701,21 @@ class AcknowledgeTasksViewSerializer(serializers.Serializer[dict[str, Any]]):
         self._validate_task_id_list(tasks)
         return tasks
 
+    def validate(self, attrs):
+        acknowledge_all = attrs.get("all", False)
+        task_ids = attrs.get("tasks")
+
+        if acknowledge_all and task_ids is not None:
+            raise serializers.ValidationError(
+                "Set either all or tasks, not both.",
+            )
+        if not acknowledge_all and task_ids is None:
+            raise serializers.ValidationError(
+                "Either all must be true or tasks must be provided.",
+            )
+
+        return attrs
+
 
 class ShareLinkSerializer(OwnedObjectSerializer):
     class Meta:

src/documents/signals/handlers.py

@@ -1,7 +1,6 @@
 from __future__ import annotations
 
 import datetime
-import hashlib
 import logging
 import shutil
 import traceback as _tb
@@ -16,6 +15,7 @@ from celery.signals import task_postrun
 from celery.signals import task_prerun
 from celery.signals import task_revoked
 from celery.signals import worker_process_init
+from celery.signals import worker_process_shutdown
 from django.conf import settings
 from django.contrib.auth.models import Group
 from django.contrib.auth.models import User
@@ -54,6 +54,7 @@ from documents.models import WorkflowTrigger
 from documents.permissions import get_objects_for_user_owner_aware
 from documents.plugins.helpers import DocumentsStatusManager
 from documents.templating.utils import convert_format_str_to_template_format
+from documents.utils import compute_checksum
 from documents.workflows.actions import build_workflow_action_context
 from documents.workflows.actions import execute_email_action
 from documents.workflows.actions import execute_move_to_trash_action
@@ -410,8 +411,7 @@ def _path_matches_checksum(path: Path, checksum: str | None) -> bool:
     if checksum is None or not path.is_file():
         return False
 
-    with path.open("rb") as f:
-        return hashlib.md5(f.read()).hexdigest() == checksum
+    return compute_checksum(path) == checksum
 
 
 def _filename_template_uses_custom_fields(doc: Document) -> bool:
@@ -1340,6 +1340,20 @@ def close_connection_pool_on_worker_init(**kwargs) -> None:
             conn.close_pool()
 
 
+@worker_process_shutdown.connect
+def close_connection_pool_on_worker_shutdown(**kwargs) -> None:  # pragma: no cover
+    """
+    Close the DB connection pool when a Celery child process exits.
+
+    With CELERY_WORKER_MAX_TASKS_PER_CHILD=1 each child is replaced after a
+    single task. Without closing the pool on shutdown, its connections linger
+    on the server until TCP keepalive reaps them, accumulating over time.
+    """
+    for conn in connections.all(initialized_only=True):
+        if conn.alias == "default" and hasattr(conn, "pool") and conn.pool:
+            conn.close_pool()
+
+
 def add_or_update_document_in_llm_index(sender, document, **kwargs):
     """
     Add or update a document in the LLM index when it is created or updated.

src/documents/templating/filepath.py

@@ -1,6 +1,7 @@
 import logging
 import os
 import re
+import unicodedata
 from collections.abc import Iterable
 from pathlib import PurePath
 
@@ -36,10 +37,12 @@ class FilePathTemplate(Template):
         def clean_filepath(value: str) -> str:
             """
             Clean up a filepath by:
-            1. Removing newlines and carriage returns
-            2. Removing extra spaces before and after forward slashes
-            3. Preserving spaces in other parts of the path
+            1. Normalizing Unicode to NFC form to prevent byte-level mismatches
+            2. Removing newlines and carriage returns
+            3. Removing extra spaces before and after forward slashes
+            4. Preserving spaces in other parts of the path
             """
+            value = unicodedata.normalize("NFC", value)
             value = value.replace("\n", "").replace("\r", "")
             value = re.sub(r"\s*/\s*", "/", value)
 
@@ -181,17 +184,17 @@ def get_basic_metadata_context(
     """
     return {
         "title": pathvalidate.sanitize_filename(
-            document.title,
+            unicodedata.normalize("NFC", document.title),
             replacement_text="-",
         ),
         "correspondent": pathvalidate.sanitize_filename(
-            document.correspondent.name,
+            unicodedata.normalize("NFC", document.correspondent.name),
             replacement_text="-",
         )
         if document.correspondent
         else no_value_default,
         "document_type": pathvalidate.sanitize_filename(
-            document.document_type.name,
+            unicodedata.normalize("NFC", document.document_type.name),
             replacement_text="-",
         )
         if document.document_type
@@ -202,7 +205,10 @@ def get_basic_metadata_context(
         "owner_username": document.owner.username
         if document.owner
         else no_value_default,
-        "original_name": PurePath(document.original_filename).with_suffix("").name
+        "original_name": unicodedata.normalize(
+            "NFC",
+            PurePath(document.original_filename).with_suffix("").name,
+        )
         if document.original_filename
         else no_value_default,
         "doc_pk": f"{document.pk:07}",
@@ -269,12 +275,12 @@ def get_tags_context(tags: Iterable[Tag]) -> dict[str, str | list[str]]:
     return {
         "tag_list": pathvalidate.sanitize_filename(
             ",".join(
-                sorted(tag.name for tag in tags),
+                sorted(unicodedata.normalize("NFC", tag.name) for tag in tags),
             ),
             replacement_text="-",
         ),
         # Assumed to be ordered, but a template could loop through to find what they want
-        "tag_name_list": [x.name for x in tags],
+        "tag_name_list": [unicodedata.normalize("NFC", x.name) for x in tags],
     }
 
 
@@ -301,7 +307,7 @@ def get_custom_fields_context(
             CustomField.FieldDataType.LONG_TEXT,
         }:
             value = pathvalidate.sanitize_filename(
-                field_instance.value,
+                unicodedata.normalize("NFC", field_instance.value),
                 replacement_text="-",
             )
         elif (
@@ -310,10 +316,13 @@ def get_custom_fields_context(
         ):
             options = field_instance.field.extra_data["select_options"]
             value = pathvalidate.sanitize_filename(
-                next(
-                    option["label"]
-                    for option in options
-                    if option["id"] == field_instance.value
+                unicodedata.normalize(
+                    "NFC",
+                    next(
+                        option["label"]
+                        for option in options
+                        if option["id"] == field_instance.value
+                    ),
                 ),
                 replacement_text="-",
             )
@@ -321,7 +330,7 @@ def get_custom_fields_context(
             value = field_instance.value
         field_data["custom_fields"][
             pathvalidate.sanitize_filename(
-                field_instance.field.name,
+                unicodedata.normalize("NFC", field_instance.field.name),
                 replacement_text="-",
             )
         ] = {

src/documents/tests/management/test_management_document_llmindex.py

@@ -0,0 +1,36 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+from django.core.management import call_command
+
+if TYPE_CHECKING:
+    from pytest_mock import MockerFixture
+
+_COMPACT = "documents.management.commands.document_llmindex.llm_index_compact"
+_INDEX = "documents.management.commands.document_llmindex.llmindex_index"
+
+
+class TestDocumentLlmindexCommand:
+    def test_compact_calls_llm_index_compact(self, mocker: MockerFixture) -> None:
+        mock_compact = mocker.patch(_COMPACT)
+        call_command("document_llmindex", "compact")
+        mock_compact.assert_called_once_with()
+
+    def test_rebuild_calls_llmindex_index_with_rebuild_true(
+        self,
+        mocker: MockerFixture,
+    ) -> None:
+        mock_index = mocker.patch(_INDEX)
+        call_command("document_llmindex", "rebuild")
+        mock_index.assert_called_once()
+        assert mock_index.call_args.kwargs["rebuild"] is True
+
+    def test_update_calls_llmindex_index_with_rebuild_false(
+        self,
+        mocker: MockerFixture,
+    ) -> None:
+        mock_index = mocker.patch(_INDEX)
+        call_command("document_llmindex", "update")
+        mock_index.assert_called_once()
+        assert mock_index.call_args.kwargs["rebuild"] is False

src/documents/tests/test_api_app_config.py

@@ -844,7 +844,7 @@ class TestApiAppConfig(DirectoriesMixin, APITestCase):
 
         with (
             patch("documents.tasks.llmindex_index.apply_async") as mock_update,
-            patch("paperless.views.vector_store_file_exists") as mock_exists,
+            patch("paperless.views.llm_index_exists") as mock_exists,
         ):
             mock_exists.return_value = False
             self.client.patch(
@@ -869,7 +869,7 @@ class TestApiAppConfig(DirectoriesMixin, APITestCase):
 
         with (
             patch("documents.tasks.llmindex_index.apply_async") as mock_update,
-            patch("paperless.views.vector_store_file_exists") as mock_exists,
+            patch("paperless.views.llm_index_exists") as mock_exists,
         ):
             mock_exists.return_value = True
             self.client.patch(
@@ -890,7 +890,7 @@ class TestApiAppConfig(DirectoriesMixin, APITestCase):
 
         with (
             patch("documents.tasks.llmindex_index.apply_async") as mock_update,
-            patch("paperless.views.vector_store_file_exists") as mock_exists,
+            patch("paperless.views.llm_index_exists") as mock_exists,
         ):
             mock_exists.return_value = True
             self.client.patch(
@@ -928,7 +928,7 @@ class TestApiAppConfig(DirectoriesMixin, APITestCase):
 
         with (
             patch("documents.tasks.llmindex_index.apply_async") as mock_update,
-            patch("paperless.views.vector_store_file_exists") as mock_exists,
+            patch("paperless.views.llm_index_exists") as mock_exists,
         ):
             mock_exists.return_value = True
             self.client.patch(

src/documents/tests/test_api_post_document_nfc.py

@@ -0,0 +1,95 @@
+import unicodedata
+from typing import TYPE_CHECKING
+from unittest import mock
+
+import celery.result
+import pytest
+from django.core.files.uploadedfile import SimpleUploadedFile
+
+if TYPE_CHECKING:
+    from documents.data_models import ConsumableDocument
+    from documents.data_models import DocumentMetadataOverrides
+
+
+@pytest.fixture()
+def consume_file_mock():
+    with mock.patch("documents.tasks.consume_file.apply_async") as m:
+        m.return_value = celery.result.AsyncResult(id="test-task-id")
+        yield m
+
+
+@pytest.fixture()
+def directories(tmp_path, settings, _media_settings):
+    scratch = tmp_path / "scratch"
+    scratch.mkdir()
+    settings.SCRATCH_DIR = scratch
+    return scratch
+
+
+@pytest.mark.django_db
+class TestPostDocumentNFCNormalization:
+    def test_nfd_filename_normalized_to_nfc(
+        self,
+        admin_client,
+        consume_file_mock: mock.MagicMock,
+        directories,
+    ):
+        """Uploaded file with NFD filename must have its name stored as NFC."""
+        nfd = unicodedata.normalize("NFD", "Rechnung März.pdf")
+        nfc = unicodedata.normalize("NFC", "Rechnung März.pdf")
+
+        # Verify our test strings actually differ at the byte level
+        assert nfd != nfc
+
+        uploaded = SimpleUploadedFile(
+            nfd,
+            b"%PDF-1.4 test",
+            content_type="application/pdf",
+        )
+        response = admin_client.post(
+            "/api/documents/post_document/",
+            {"document": uploaded},
+        )
+
+        assert response.status_code == 200
+
+        task_kwargs = consume_file_mock.call_args.kwargs["kwargs"]
+        input_doc: ConsumableDocument = task_kwargs["input_doc"]
+        overrides: DocumentMetadataOverrides = task_kwargs["overrides"]
+
+        # The temp file on disk must have an NFC name
+        assert input_doc.original_file.name == nfc, (
+            f"Expected NFC filename {nfc!r}, got {input_doc.original_file.name!r}"
+        )
+        # The override filename stored for later use must also be NFC
+        assert overrides.filename == nfc, (
+            f"Expected NFC override filename {nfc!r}, got {overrides.filename!r}"
+        )
+        assert unicodedata.is_normalized("NFC", overrides.filename)
+
+    def test_already_nfc_filename_unchanged(
+        self,
+        admin_client,
+        consume_file_mock: mock.MagicMock,
+        directories,
+    ):
+        """Uploaded file with already-NFC filename must pass through unchanged."""
+        nfc = unicodedata.normalize("NFC", "Invoice_2024.pdf")
+
+        uploaded = SimpleUploadedFile(
+            nfc,
+            b"%PDF-1.4 test",
+            content_type="application/pdf",
+        )
+        response = admin_client.post(
+            "/api/documents/post_document/",
+            {"document": uploaded},
+        )
+
+        assert response.status_code == 200
+
+        task_kwargs = consume_file_mock.call_args.kwargs["kwargs"]
+        overrides: DocumentMetadataOverrides = task_kwargs["overrides"]
+
+        assert overrides.filename == nfc
+        assert unicodedata.is_normalized("NFC", overrides.filename)

src/documents/tests/test_api_tasks.py

@@ -18,6 +18,7 @@ from guardian.shortcuts import assign_perm
 from rest_framework import status
 from rest_framework.test import APIClient
 
+from documents.filters import PaperlessTaskFilterSet
 from documents.models import PaperlessTask
 from documents.tests.factories import DocumentFactory
 from documents.tests.factories import PaperlessTaskFactory
@@ -169,6 +170,165 @@ class TestGetTasksV10:
             PaperlessTask.Status.STARTED,
         }
 
+    def test_filter_by_task_name(self, admin_client: APIClient) -> None:
+        """?name= searches task filenames, task types, and trigger sources."""
+        filename_task = PaperlessTaskFactory(input_data={"filename": "invoice-123.pdf"})
+        type_task = PaperlessTaskFactory(task_type=PaperlessTask.TaskType.SANITY_CHECK)
+        source_task = PaperlessTaskFactory(
+            trigger_source=PaperlessTask.TriggerSource.EMAIL_CONSUME,
+        )
+        PaperlessTaskFactory(input_data={"filename": "unrelated.pdf"})
+
+        response = admin_client.get(ENDPOINT, {"name": "invoice"})
+
+        assert response.status_code == status.HTTP_200_OK
+        assert response.data["count"] == 1
+        assert response.data["results"][0]["task_id"] == filename_task.task_id
+
+        response = admin_client.get(ENDPOINT, {"name": "sanity"})
+
+        assert response.status_code == status.HTTP_200_OK
+        assert response.data["count"] == 1
+        assert response.data["results"][0]["task_id"] == type_task.task_id
+
+        response = admin_client.get(ENDPOINT, {"name": "email"})
+
+        assert response.status_code == status.HTTP_200_OK
+        assert response.data["count"] == 1
+        assert response.data["results"][0]["task_id"] == source_task.task_id
+
+    def test_filter_by_task_result(self, admin_client: APIClient) -> None:
+        """?result= searches common structured task result messages."""
+        reason_task = PaperlessTaskFactory(result_data={"reason": "Manual review"})
+        error_task = PaperlessTaskFactory(
+            result_data={"error_message": "Duplicate detected"},
+        )
+        document_task = PaperlessTaskFactory(result_data={"document_id": 321})
+        duplicate_task = PaperlessTaskFactory(result_data={"duplicate_of": 123})
+        PaperlessTaskFactory(result_data={"reason": "unrelated"})
+
+        response = admin_client.get(ENDPOINT, {"result": "manual"})
+
+        assert response.status_code == status.HTTP_200_OK
+        assert response.data["count"] == 1
+        assert response.data["results"][0]["task_id"] == reason_task.task_id
+
+        response = admin_client.get(ENDPOINT, {"result": "duplicate"})
+
+        assert response.status_code == status.HTTP_200_OK
+        returned_ids = {task["task_id"] for task in response.data["results"]}
+        assert returned_ids == {error_task.task_id, duplicate_task.task_id}
+
+        response = admin_client.get(ENDPOINT, {"result": "321"})
+
+        assert response.status_code == status.HTTP_200_OK
+        assert response.data["count"] == 1
+        assert response.data["results"][0]["task_id"] == document_task.task_id
+
+    def test_empty_task_name_and_result_filters(self) -> None:
+        """Empty name/result values leave the queryset unchanged."""
+        PaperlessTaskFactory.create_batch(2)
+        queryset = PaperlessTask.objects.all()
+        filterset = PaperlessTaskFilterSet()
+
+        assert filterset.filter_name(queryset, "name", "").count() == 2
+        assert filterset.filter_result(queryset, "result", "").count() == 2
+
+    def test_status_counts_respects_filters(self, admin_client: APIClient) -> None:
+        """status_counts/ returns section counts for the filtered task queryset."""
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.FAILURE,
+            input_data={"filename": "invoice-a.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.REVOKED,
+            input_data={"filename": "invoice-b.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.PENDING,
+            input_data={"filename": "invoice-c.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.STARTED,
+            input_data={"filename": "invoice-d.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.SUCCESS,
+            input_data={"filename": "invoice-e.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=True,
+            status=PaperlessTask.Status.SUCCESS,
+            input_data={"filename": "invoice-acknowledged.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.SUCCESS,
+            input_data={"filename": "unrelated.pdf"},
+        )
+
+        response = admin_client.get(
+            f"{ENDPOINT}status_counts/",
+            {"acknowledged": "false", "name": "invoice"},
+        )
+
+        assert response.status_code == status.HTTP_200_OK
+        assert response.data == {
+            "all": 5,
+            "needs_attention": 2,
+            "in_progress": 2,
+            "completed": 1,
+        }
+
+    def test_status_counts_ignores_section_filters(
+        self,
+        admin_client: APIClient,
+    ) -> None:
+        """status_counts/ ignores status-like filters for the sections it counts."""
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.FAILURE,
+            input_data={"filename": "invoice-a.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.PENDING,
+            input_data={"filename": "invoice-b.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.SUCCESS,
+            input_data={"filename": "invoice-c.pdf"},
+        )
+        PaperlessTaskFactory(
+            acknowledged=False,
+            status=PaperlessTask.Status.FAILURE,
+            input_data={"filename": "unrelated.pdf"},
+        )
+
+        response = admin_client.get(
+            f"{ENDPOINT}status_counts/",
+            {
+                "acknowledged": "false",
+                "name": "invoice",
+                "status": PaperlessTask.Status.FAILURE,
+                "is_complete": "false",
+            },
+        )
+
+        assert response.status_code == status.HTTP_200_OK
+        assert response.data == {
+            "all": 3,
+            "needs_attention": 1,
+            "in_progress": 1,
+            "completed": 1,
+        }
+
     def test_default_ordering_is_newest_first(self, admin_client: APIClient) -> None:
         """Tasks are returned in descending date_created order (newest first)."""
         base = timezone.now()
@@ -522,6 +682,27 @@ class TestAcknowledge:
         assert response.status_code == status.HTTP_200_OK
         assert response.data == {"result": 2}
 
+    def test_acknowledge_all_returns_count(self, admin_client: APIClient) -> None:
+        """POST acknowledge/ with all=true acknowledges all unacknowledged tasks."""
+        unacknowledged_task1 = PaperlessTaskFactory(acknowledged=False)
+        unacknowledged_task2 = PaperlessTaskFactory(acknowledged=False)
+        acknowledged_task = PaperlessTaskFactory(acknowledged=True)
+
+        response = admin_client.post(
+            ENDPOINT + "acknowledge/",
+            {"all": True},
+            format="json",
+        )
+
+        assert response.status_code == status.HTTP_200_OK
+        assert response.data == {"result": 2}
+        unacknowledged_task1.refresh_from_db()
+        unacknowledged_task2.refresh_from_db()
+        acknowledged_task.refresh_from_db()
+        assert unacknowledged_task1.acknowledged
+        assert unacknowledged_task2.acknowledged
+        assert acknowledged_task.acknowledged
+
     def test_acknowledged_tasks_excluded_from_unacked_filter(
         self,
         admin_client: APIClient,

src/documents/tests/test_bulk_edit.py

@@ -3,6 +3,7 @@ from datetime import date
 from pathlib import Path
 from unittest import mock
 
+import pikepdf
 from django.contrib.auth.models import Group
 from django.contrib.auth.models import User
 from django.test import TestCase
@@ -615,6 +616,18 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         self.img_doc.archive_filename = img_doc_archive
         self.img_doc.save()
 
+    @staticmethod
+    def mock_password_required_pdf(
+        mock_open: mock.Mock,
+        fake_pdf: mock.Mock,
+    ) -> None:
+        password_context = mock.MagicMock()
+        password_context.__enter__.return_value = fake_pdf
+        mock_open.side_effect = [
+            pikepdf.PasswordError("password required"),
+            password_context,
+        ]
+
     @mock.patch("documents.tasks.consume_file.s")
     def test_merge(self, mock_consume_file) -> None:
         """
@@ -1466,6 +1479,7 @@ class TestPDFActions(DirectoriesMixin, TestCase):
 
         fake_pdf = mock.MagicMock()
         fake_pdf.pages = [mock.Mock(), mock.Mock(), mock.Mock()]
+        fake_pdf.is_encrypted = True
 
         def save_side_effect(target_path):
             Path(target_path).write_bytes(b"new pdf content")
@@ -1480,7 +1494,13 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         )
 
         self.assertEqual(result, "OK")
-        mock_open.assert_called_once_with(doc.source_path, password="secret")
+        self.assertEqual(
+            mock_open.call_args_list,
+            [
+                mock.call(doc.source_path),
+                mock.call(doc.source_path, password="secret"),
+            ],
+        )
         fake_pdf.remove_unreferenced_resources.assert_called_once()
         mock_update_document.assert_not_called()
         mock_consume_delay.assert_called_once()
@@ -1494,6 +1514,33 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         self.assertEqual(task_kwargs["input_doc"].root_document_id, doc.id)
         self.assertIsNotNone(task_kwargs["overrides"])
 
+    @mock.patch("documents.tasks.consume_file.apply_async")
+    @mock.patch("documents.bulk_edit.tempfile.mkdtemp")
+    @mock.patch("pikepdf.open")
+    def test_remove_password_update_document_skips_unencrypted_pdf(
+        self,
+        mock_open,
+        mock_mkdtemp,
+        mock_consume_delay,
+    ) -> None:
+        doc = self.doc1
+        fake_pdf = mock.MagicMock()
+        fake_pdf.is_encrypted = False
+        mock_open.return_value.__enter__.return_value = fake_pdf
+
+        result = bulk_edit.remove_password(
+            [doc.id],
+            password="secret",
+            update_document=True,
+        )
+
+        self.assertEqual(result, "OK")
+        mock_open.assert_called_once_with(doc.source_path)
+        fake_pdf.remove_unreferenced_resources.assert_not_called()
+        fake_pdf.save.assert_not_called()
+        mock_mkdtemp.assert_not_called()
+        mock_consume_delay.assert_not_called()
+
     @mock.patch("documents.bulk_edit.update_document_content_maybe_archive_file.delay")
     @mock.patch("documents.tasks.consume_file.apply_async")
     @mock.patch("documents.bulk_edit.tempfile.mkdtemp")
@@ -1513,12 +1560,12 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         mock_mkdtemp.return_value = str(temp_dir)
 
         fake_pdf = mock.MagicMock()
+        self.mock_password_required_pdf(mock_open, fake_pdf)
 
         def save_side_effect(target_path):
             Path(target_path).write_bytes(b"new pdf content")
 
         fake_pdf.save.side_effect = save_side_effect
-        mock_open.return_value.__enter__.return_value = fake_pdf
 
         result = bulk_edit.remove_password(
             [doc.id],
@@ -1528,7 +1575,13 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         )
 
         self.assertEqual(result, "OK")
-        mock_open.assert_called_once_with(source_file, password="secret")
+        self.assertEqual(
+            mock_open.call_args_list,
+            [
+                mock.call(source_file),
+                mock.call(source_file, password="secret"),
+            ],
+        )
         mock_update_document.assert_not_called()
         mock_consume_delay.assert_called_once()
 
@@ -1547,7 +1600,7 @@ class TestPDFActions(DirectoriesMixin, TestCase):
             root_document=self.doc1,
         )
         fake_pdf = mock.MagicMock()
-        mock_open.return_value.__enter__.return_value = fake_pdf
+        self.mock_password_required_pdf(mock_open, fake_pdf)
 
         result = bulk_edit.remove_password(
             [self.doc1.id],
@@ -1557,7 +1610,13 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         )
 
         self.assertEqual(result, "OK")
-        mock_open.assert_called_once_with(self.doc1.source_path, password="secret")
+        self.assertEqual(
+            mock_open.call_args_list,
+            [
+                mock.call(self.doc1.source_path),
+                mock.call(self.doc1.source_path, password="secret"),
+            ],
+        )
         mock_consume_delay.assert_called_once()
 
     @mock.patch("documents.bulk_edit.chord")
@@ -1580,12 +1639,12 @@ class TestPDFActions(DirectoriesMixin, TestCase):
 
         fake_pdf = mock.MagicMock()
         fake_pdf.pages = [mock.Mock(), mock.Mock()]
+        self.mock_password_required_pdf(mock_open, fake_pdf)
 
         def save_side_effect(target_path: Path) -> None:
             target_path.write_bytes(b"password removed")
 
         fake_pdf.save.side_effect = save_side_effect
-        mock_open.return_value.__enter__.return_value = fake_pdf
         mock_group.return_value.delay.return_value = None
 
         user = User.objects.create(username="owner")
@@ -1600,7 +1659,13 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         )
 
         self.assertEqual(result, "OK")
-        mock_open.assert_called_once_with(doc.source_path, password="secret")
+        self.assertEqual(
+            mock_open.call_args_list,
+            [
+                mock.call(doc.source_path),
+                mock.call(doc.source_path, password="secret"),
+            ],
+        )
         mock_consume_file.assert_called_once()
         call_kwargs = mock_consume_file.call_args.kwargs
         consumable_document = call_kwargs["input_doc"]
@@ -1618,6 +1683,43 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         mock_group.return_value.delay.assert_called_once()
         mock_chord.assert_not_called()
 
+    @mock.patch("documents.bulk_edit.delete")
+    @mock.patch("documents.bulk_edit.chord")
+    @mock.patch("documents.bulk_edit.group")
+    @mock.patch("documents.tasks.consume_file.s")
+    @mock.patch("documents.bulk_edit.tempfile.mkdtemp")
+    @mock.patch("pikepdf.open")
+    def test_remove_password_skips_unencrypted_pdf_without_queueing(
+        self,
+        mock_open: mock.Mock,
+        mock_mkdtemp: mock.Mock,
+        mock_consume_file: mock.Mock,
+        mock_group: mock.Mock,
+        mock_chord: mock.Mock,
+        mock_delete: mock.Mock,
+    ) -> None:
+        doc = self.doc2
+        fake_pdf = mock.MagicMock()
+        fake_pdf.is_encrypted = False
+        mock_open.return_value.__enter__.return_value = fake_pdf
+
+        result = bulk_edit.remove_password(
+            [doc.id],
+            password="secret",
+            update_document=False,
+            delete_original=True,
+        )
+
+        self.assertEqual(result, "OK")
+        mock_open.assert_called_once_with(doc.source_path)
+        fake_pdf.remove_unreferenced_resources.assert_not_called()
+        fake_pdf.save.assert_not_called()
+        mock_mkdtemp.assert_not_called()
+        mock_consume_file.assert_not_called()
+        mock_group.assert_not_called()
+        mock_chord.assert_not_called()
+        mock_delete.si.assert_not_called()
+
     @mock.patch("documents.bulk_edit.delete")
     @mock.patch("documents.bulk_edit.chord")
     @mock.patch("documents.bulk_edit.group")
@@ -1640,12 +1742,12 @@ class TestPDFActions(DirectoriesMixin, TestCase):
 
         fake_pdf = mock.MagicMock()
         fake_pdf.pages = [mock.Mock(), mock.Mock()]
+        self.mock_password_required_pdf(mock_open, fake_pdf)
 
         def save_side_effect(target_path: Path) -> None:
             target_path.write_bytes(b"password removed")
 
         fake_pdf.save.side_effect = save_side_effect
-        mock_open.return_value.__enter__.return_value = fake_pdf
         mock_chord.return_value.delay.return_value = None
 
         result = bulk_edit.remove_password(
@@ -1657,7 +1759,13 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         )
 
         self.assertEqual(result, "OK")
-        mock_open.assert_called_once_with(doc.source_path, password="secret")
+        self.assertEqual(
+            mock_open.call_args_list,
+            [
+                mock.call(doc.source_path),
+                mock.call(doc.source_path, password="secret"),
+            ],
+        )
         mock_consume_file.assert_called_once()
         mock_group.assert_not_called()
         mock_chord.assert_called_once()

src/documents/tests/test_file_handling.py

@@ -24,6 +24,7 @@ from documents.models import CustomFieldInstance
 from documents.models import Document
 from documents.models import DocumentType
 from documents.models import StoragePath
+from documents.serialisers import DocumentSerializer
 from documents.tasks import empty_trash
 from documents.tests.factories import DocumentFactory
 from documents.tests.utils import DirectoriesMixin
@@ -221,8 +222,8 @@ class TestFileHandling(DirectoriesMixin, FileSystemAssertsMixin, TestCase):
         doc = Document.objects.create(
             title="document",
             mime_type="application/pdf",
-            checksum=hashlib.md5(original_bytes).hexdigest(),
-            archive_checksum=hashlib.md5(archive_bytes).hexdigest(),
+            checksum=hashlib.sha256(original_bytes).hexdigest(),
+            archive_checksum=hashlib.sha256(archive_bytes).hexdigest(),
             filename="old/document.pdf",
             archive_filename="old/document.pdf",
             storage_path=old_storage_path,
@@ -251,6 +252,46 @@ class TestFileHandling(DirectoriesMixin, FileSystemAssertsMixin, TestCase):
         self.assertIsNotFile(settings.ORIGINALS_DIR / "old" / "document.pdf")
         self.assertIsNotFile(settings.ARCHIVE_DIR / "old" / "document.pdf")
 
+    @override_settings(FILENAME_FORMAT="{title}")
+    def test_serializer_stale_update_does_not_clobber_filename(self) -> None:
+        old_path = settings.ORIGINALS_DIR / "original.pdf"
+        old_path.touch()
+        doc = Document.objects.create(
+            title="original",
+            mime_type="application/pdf",
+            checksum=hashlib.sha256(b"").hexdigest(),
+            filename="original.pdf",
+        )
+
+        first_instance = Document.objects.get(pk=doc.pk)
+        stale_instance = Document.objects.get(pk=doc.pk)
+
+        serializer = DocumentSerializer(
+            first_instance,
+            data={"title": "first"},
+            partial=True,
+        )
+        self.assertTrue(serializer.is_valid(), serializer.errors)
+        serializer.save()
+
+        doc.refresh_from_db()
+        self.assertEqual(doc.filename, "first.pdf")
+        self.assertIsFile(settings.ORIGINALS_DIR / "first.pdf")
+
+        serializer = DocumentSerializer(
+            stale_instance,
+            data={"title": "second"},
+            partial=True,
+        )
+        self.assertTrue(serializer.is_valid(), serializer.errors)
+        serializer.save()
+
+        doc.refresh_from_db()
+        self.assertEqual(doc.filename, "second.pdf")
+        self.assertIsFile(settings.ORIGINALS_DIR / "second.pdf")
+        self.assertIsNotFile(settings.ORIGINALS_DIR / "first.pdf")
+        self.assertIsNotFile(old_path)
+
     @override_settings(FILENAME_FORMAT="{correspondent}/{correspondent}")
     def test_document_delete(self) -> None:
         document = Document()

src/documents/tests/test_filepath_nfc.py

@@ -0,0 +1,187 @@
+"""
+Tests for NFC Unicode normalization in generate_filename / FilePathTemplate.render().
+
+NFC `ü` (UTF-8: c3 bc) and NFD `ü` (UTF-8: 75 cc 88) are visually identical but
+produce different byte sequences.  On Linux (ext4, ZFS) these are distinct filenames.
+All paths produced by the templating system must be NFC-normalized.
+"""
+
+import unicodedata
+
+import pytest
+
+from documents.file_handling import generate_filename
+from documents.models import CustomField
+from documents.models import CustomFieldInstance
+from documents.tests.factories import CorrespondentFactory
+from documents.tests.factories import DocumentFactory
+from documents.tests.factories import StoragePathFactory
+from documents.tests.factories import TagFactory
+
+
+@pytest.mark.django_db
+class TestGenerateFilenameNFCNormalization:
+    @pytest.mark.parametrize(
+        "raw,display",
+        [
+            (unicodedata.normalize("NFD", "Gemüse"), "Gemüse"),
+            (unicodedata.normalize("NFD", "Café"), "Café"),
+            (unicodedata.normalize("NFD", "naïve"), "naïve"),
+        ],
+    )
+    def test_nfd_title_normalized_to_nfc(self, settings, raw, display):
+        """NFD title must produce NFC path bytes."""
+        settings.FILENAME_FORMAT = "{{ title }}"
+        nfc = unicodedata.normalize("NFC", display)
+        assert raw != nfc  # confirm byte-level difference
+
+        doc = DocumentFactory(title=raw, mime_type="application/pdf")
+        result = generate_filename(doc)
+
+        assert str(result) == f"{nfc}.pdf"
+        assert str(result).encode() == f"{nfc}.pdf".encode()
+
+    def test_nfd_correspondent_normalized_to_nfc(self, settings):
+        """NFD correspondent name must produce NFC path component."""
+        settings.FILENAME_FORMAT = "{{ correspondent }}/{{ title }}"
+        nfd = unicodedata.normalize("NFD", "Müller")
+        nfc = unicodedata.normalize("NFC", "Müller")
+
+        correspondent = CorrespondentFactory(name=nfd)
+        doc = DocumentFactory(
+            title="invoice",
+            correspondent=correspondent,
+            mime_type="application/pdf",
+        )
+        result = generate_filename(doc)
+
+        assert str(result) == f"{nfc}/invoice.pdf"
+        assert str(result).encode() == f"{nfc}/invoice.pdf".encode()
+
+    def test_nfd_storage_path_normalized_to_nfc(self, settings):
+        """NFD literal in StoragePath.path template must produce NFC path bytes."""
+        settings.FILENAME_FORMAT = None
+        nfd = unicodedata.normalize("NFD", "Büro")
+        nfc = unicodedata.normalize("NFC", "Büro")
+
+        # StoragePath.path is used directly as the format/template string.
+        # Literal NFD characters in the template must survive rendering as NFC.
+        sp = StoragePathFactory(path=f"{nfd}/{{{{ title }}}}")
+        doc = DocumentFactory(title="doc", storage_path=sp, mime_type="application/pdf")
+        result = generate_filename(doc)
+
+        assert str(result).encode() == f"{nfc}/doc.pdf".encode()
+
+    def test_nfd_raw_document_title_normalized_to_nfc(self, settings):
+        """NFD title accessed via document.title (unsanitized context) must also be NFC."""
+        settings.FILENAME_FORMAT = "{{ document.title }}"
+        nfd = unicodedata.normalize("NFD", "Café")
+        nfc = unicodedata.normalize("NFC", "Café")
+
+        doc = DocumentFactory(title=nfd, mime_type="application/pdf")
+        result = generate_filename(doc)
+
+        assert str(result) == f"{nfc}.pdf"
+        assert str(result).encode() == f"{nfc}.pdf".encode()
+
+
+@pytest.mark.django_db
+class TestContextBuilderNFCNormalization:
+    """
+    Defense-in-depth: context builder functions must NFC-normalize string inputs
+    before passing them to sanitize_filename().  Task 1 already normalizes the
+    final rendered path via clean_filepath(), so these tests may already pass;
+    they exist as regression guards for the context-builder layer.
+    """
+
+    def test_nfd_tag_name_normalized_in_tag_list(self, settings):
+        """NFD tag name must appear as NFC bytes in the {{ tag_list }} shorthand."""
+        settings.FILENAME_FORMAT = "{{ tag_list }}/{{ title }}"
+        nfd = unicodedata.normalize("NFD", "Büro")
+        nfc = unicodedata.normalize("NFC", "Büro")
+        assert nfd != nfc  # confirm they differ at byte level
+
+        tag = TagFactory(name=nfd)
+        doc = DocumentFactory(title="doc", mime_type="application/pdf")
+        doc.tags.set([tag])
+
+        result = generate_filename(doc)
+
+        assert str(result).encode() == f"{nfc}/doc.pdf".encode()
+
+    def test_nfd_original_name_normalized_to_nfc(self, settings):
+        settings.FILENAME_FORMAT = "{{ original_name }}"
+        nfd = unicodedata.normalize("NFD", "Rechnung März")
+        nfc = unicodedata.normalize("NFC", "Rechnung März")
+
+        doc = DocumentFactory(
+            original_filename=f"{nfd}.pdf",
+            mime_type="application/pdf",
+        )
+        result = generate_filename(doc)
+
+        assert str(result).encode() == f"{nfc}.pdf".encode()
+
+    def test_nfd_custom_field_string_value_normalized(self, settings):
+        """NFD value in a STRING-type custom field must appear as NFC in the context."""
+        settings.FILENAME_FORMAT = (
+            "{{ custom_fields['Location']['value'] }}/{{ title }}"
+        )
+        nfd_value = unicodedata.normalize("NFD", "Düsseldorf")
+        nfc_value = unicodedata.normalize("NFC", "Düsseldorf")
+        assert nfd_value != nfc_value
+
+        doc = DocumentFactory(title="report", mime_type="application/pdf")
+        cf = CustomField.objects.create(
+            name="Location",
+            data_type=CustomField.FieldDataType.STRING,
+        )
+        CustomFieldInstance.objects.create(
+            document=doc,
+            field=cf,
+            value_text=nfd_value,
+        )
+
+        result = generate_filename(doc)
+
+        assert str(result).encode() == f"{nfc_value}/report.pdf".encode()
+
+    def test_nfd_custom_field_name_normalized_as_key(self, settings):
+        """NFD characters in a custom field name must appear as NFC in the context dict key."""
+        nfd_name = unicodedata.normalize("NFD", "Größe")
+        nfc_name = unicodedata.normalize("NFC", "Größe")
+        assert nfd_name != nfc_name
+
+        settings.FILENAME_FORMAT = f"{{% if custom_fields['{nfc_name}'] %}}{{{{ custom_fields['{nfc_name}']['value'] }}}}/{{{{ title }}}}{{% else %}}{{{{ title }}}}{{% endif %}}"
+
+        doc = DocumentFactory(title="letter", mime_type="application/pdf")
+        cf = CustomField.objects.create(
+            name=nfd_name,
+            data_type=CustomField.FieldDataType.STRING,
+        )
+        CustomFieldInstance.objects.create(
+            document=doc,
+            field=cf,
+            value_text="Berlin",
+        )
+
+        result = generate_filename(doc)
+
+        # If field name key is NFC-normalized, the template condition succeeds
+        # and result is "Berlin/letter.pdf"; otherwise it falls back to "letter.pdf"
+        assert str(result) == "Berlin/letter.pdf"
+
+    def test_nfd_tag_name_list_normalized_to_nfc(self, settings):
+        """NFD tag names in tag_name_list must appear as NFC bytes when iterated."""
+        settings.FILENAME_FORMAT = (
+            "{% for t in tag_name_list %}{{ t }}{% endfor %}/{{ title }}"
+        )
+        nfd = unicodedata.normalize("NFD", "Büro")
+        nfc = unicodedata.normalize("NFC", "Büro")
+        assert nfd != nfc  # confirm byte-level difference
+
+        doc = DocumentFactory(title="doc", mime_type="application/pdf")
+        doc.tags.add(TagFactory(name=nfd))
+        result = generate_filename(doc)
+
+        assert str(result).encode() == f"{nfc}/doc.pdf".encode()

src/documents/tests/test_views.py

@@ -368,6 +368,7 @@ class TestAISuggestions(DirectoriesMixin, TestCase):
             self.document,
             self.user,
             None,
+            hints=None,
         )
 
     @patch("documents.views.get_ai_document_classification")
@@ -399,6 +400,7 @@ class TestAISuggestions(DirectoriesMixin, TestCase):
             self.document,
             self.user,
             "de-de",
+            hints=None,
         )
         self.assertEqual(
             get_llm_suggestion_cache(
@@ -438,6 +440,7 @@ class TestAISuggestions(DirectoriesMixin, TestCase):
             self.document,
             self.user,
             "fr-fr",
+            hints=None,
         )
         self.assertEqual(
             get_llm_suggestion_cache(

src/documents/views.py

@@ -245,6 +245,7 @@ from paperless_ai.matching import match_correspondents_by_name
 from paperless_ai.matching import match_document_types_by_name
 from paperless_ai.matching import match_storage_paths_by_name
 from paperless_ai.matching import match_tags_by_name
+from paperless_ai.taxonomy import get_taxonomy_hints_for_document
 from paperless_mail.models import MailAccount
 from paperless_mail.models import MailRule
 from paperless_mail.oauth import PaperlessMailOAuth2Manager
@@ -1400,7 +1401,7 @@ class DocumentViewSet(
         )
         if request.user is not None and not has_perms_owner_aware(
             request.user,
-            "view_document",
+            "change_document",
             doc,
         ):
             return HttpResponseForbidden("Insufficient permissions")
@@ -1460,7 +1461,7 @@ class DocumentViewSet(
         )
         if request.user is not None and not has_perms_owner_aware(
             request.user,
-            "view_document",
+            "change_document",
             doc,
         ):
             return HttpResponseForbidden("Insufficient permissions")
@@ -1494,11 +1495,14 @@ class DocumentViewSet(
             refresh_suggestions_cache(doc.pk)
             return Response(cached_llm_suggestions.suggestions)
 
+        hints = get_taxonomy_hints_for_document(doc, request.user)
+
         try:
             llm_suggestions = get_ai_document_classification(
                 doc,
                 request.user,
                 output_language,
+                hints=hints,
             )
         except ValueError as exc:
             logger.exception(
@@ -1513,18 +1517,22 @@ class DocumentViewSet(
         matched_tags = match_tags_by_name(
             llm_suggestions.get("tags", []),
             request.user,
+            hinted_names=set(hints["tags"]) if hints else None,
         )
         matched_correspondents = match_correspondents_by_name(
             llm_suggestions.get("correspondents", []),
             request.user,
+            hinted_names=set(hints["correspondents"]) if hints else None,
         )
         matched_types = match_document_types_by_name(
             llm_suggestions.get("document_types", []),
             request.user,
+            hinted_names=set(hints["document_types"]) if hints else None,
         )
         matched_paths = match_storage_paths_by_name(
             llm_suggestions.get("storage_paths", []),
             request.user,
+            hinted_names=set(hints["storage_paths"]) if hints else None,
         )
 
         resp_data = {
@@ -3126,6 +3134,7 @@ class PostDocumentView(GenericAPIView[Any]):
         serializer.is_valid(raise_exception=True)
 
         doc_name, doc_data = serializer.validated_data.get("document")
+        doc_name = normalize("NFC", doc_name)
         correspondent_id = serializer.validated_data.get("correspondent")
         document_type_id = serializer.validated_data.get("document_type")
         storage_path_id = serializer.validated_data.get("storage_path")
@@ -4011,7 +4020,7 @@ class RemoteVersionView(GenericAPIView[Any]):
 
 
 class _TasksViewSetSchema(AutoSchema):
-    _UNPAGINATED_ACTIONS = frozenset({"summary", "active"})
+    _UNPAGINATED_ACTIONS = frozenset({"summary", "active", "status_counts"})
 
     def _get_paginator(self):
         if getattr(self.view, "action", None) in self._UNPAGINATED_ACTIONS:
@@ -4033,7 +4042,7 @@ class _TasksViewSetSchema(AutoSchema):
     ),
     acknowledge=extend_schema(
         operation_id="acknowledge_tasks",
-        description="Acknowledge a list of tasks",
+        description="Acknowledge a list of tasks, or all visible unacknowledged tasks",
         request=AcknowledgeTasksViewSerializer,
         responses={
             (200, "application/json"): inline_serializer(
@@ -4071,6 +4080,19 @@ class _TasksViewSetSchema(AutoSchema):
             ),
         ],
     ),
+    status_counts=extend_schema(
+        responses={
+            200: inline_serializer(
+                name="TaskStatusCounts",
+                fields={
+                    "all": serializers.IntegerField(),
+                    "needs_attention": serializers.IntegerField(),
+                    "in_progress": serializers.IntegerField(),
+                    "completed": serializers.IntegerField(),
+                },
+            ),
+        },
+    ),
     active=extend_schema(
         description="Currently pending and running tasks (capped at 50).",
         responses={200: TaskSerializerV10(many=True)},
@@ -4124,6 +4146,7 @@ class TasksViewSet(ReadOnlyModelViewSet[PaperlessTask]):
         PaperlessTask.TaskType.SANITY_CHECK: (sanity_check, {"raise_on_error": False}),
         PaperlessTask.TaskType.LLM_INDEX: (llmindex_index, {"rebuild": False}),
     }
+    _STATUS_COUNT_EXCLUDED_FILTERS = frozenset({"status", "is_complete"})
 
     def get_serializer_class(self):
         # v9: use backwards-compatible serializer with old field names
@@ -4164,16 +4187,38 @@ class TasksViewSet(ReadOnlyModelViewSet[PaperlessTask]):
             queryset = queryset.filter(task_id=task_id)
         return queryset
 
+    def get_status_count_queryset(self):
+        """Apply task filters except the status dimensions represented by the counts."""
+        query_params = self.request.query_params.copy()
+        for param in self._STATUS_COUNT_EXCLUDED_FILTERS:
+            query_params.pop(param, None)
+
+        filterset = self.filterset_class(
+            data=query_params,
+            queryset=self.get_queryset(),
+            request=self.request,
+        )
+        if not filterset.is_valid():
+            raise ValidationError(filterset.errors)
+        return filterset.qs
+
     @action(
         methods=["post"],
         detail=False,
         permission_classes=[IsAuthenticated, AcknowledgeTasksPermissions],
     )
     def acknowledge(self, request):
-        serializer = AcknowledgeTasksViewSerializer(data=request.data)
+        queryset = self.get_queryset()
+        serializer = AcknowledgeTasksViewSerializer(
+            data=request.data,
+            context={"queryset": queryset},
+        )
         serializer.is_valid(raise_exception=True)
-        task_ids = serializer.validated_data.get("tasks")
-        tasks = self.get_queryset().filter(id__in=task_ids)
+        if serializer.validated_data.get("all", False):
+            tasks = queryset.filter(acknowledged=False)
+        else:
+            task_ids = serializer.validated_data.get("tasks")
+            tasks = queryset.filter(id__in=task_ids)
         count = tasks.update(acknowledged=True)
         return Response({"result": count})
 
@@ -4226,6 +4271,34 @@ class TasksViewSet(ReadOnlyModelViewSet[PaperlessTask]):
         serializer = TaskSummarySerializer(data, many=True)
         return Response(serializer.data)
 
+    @action(methods=["get"], detail=False)
+    def status_counts(self, request):
+        """Aggregated task counts for task UI sections."""
+        queryset = self.get_status_count_queryset()
+        counts = queryset.aggregate(
+            all=Count("id"),
+            needs_attention=Count(
+                "id",
+                filter=Q(
+                    status__in=[
+                        PaperlessTask.Status.FAILURE,
+                        PaperlessTask.Status.REVOKED,
+                    ],
+                ),
+            ),
+            in_progress=Count(
+                "id",
+                filter=Q(
+                    status__in=[
+                        PaperlessTask.Status.PENDING,
+                        PaperlessTask.Status.STARTED,
+                    ],
+                ),
+            ),
+            completed=Count("id", filter=Q(status=PaperlessTask.Status.SUCCESS)),
+        )
+        return Response(counts)
+
     @action(methods=["get"], detail=False)
     def active(self, request):
         """Currently pending and running tasks (capped at 50)."""

src/paperless/parsers/text.py

@@ -20,6 +20,7 @@ from PIL import Image
 from PIL import ImageDraw
 from PIL import ImageFont
 
+from paperless.parsers.utils import read_file_handle_unicode_errors
 from paperless.version import __full_version_str__
 
 if TYPE_CHECKING:
@@ -183,7 +184,7 @@ class TextDocumentParser:
         documents.parsers.ParseError
             If the file cannot be read.
         """
-        self._text = self._read_text(document_path)
+        self._text = read_file_handle_unicode_errors(document_path, log=logger)
 
     # ------------------------------------------------------------------
     # Result accessors
@@ -295,30 +296,3 @@ class TextDocumentParser:
             Always ``[]`` — plain text files carry no structured metadata.
         """
         return []
-
-    # ------------------------------------------------------------------
-    # Private helpers
-    # ------------------------------------------------------------------
-
-    def _read_text(self, filepath: Path) -> str:
-        """Read file content, replacing invalid UTF-8 bytes rather than failing.
-
-        Parameters
-        ----------
-        filepath:
-            Path to the file to read.
-
-        Returns
-        -------
-        str
-            File content as a string.
-        """
-        try:
-            return filepath.read_text(encoding="utf-8")
-        except UnicodeDecodeError as exc:
-            logger.warning(
-                "Unicode error reading %s, replacing bad bytes: %s",
-                filepath,
-                exc,
-            )
-            return filepath.read_bytes().decode("utf-8", errors="replace")

src/paperless/parsers/utils.py

@@ -8,6 +8,7 @@ share implementation.
 
 from __future__ import annotations
 
+import codecs
 import logging
 import re
 import tempfile
@@ -114,7 +115,7 @@ def read_file_handle_unicode_errors(
     filepath: Path,
     log: logging.Logger | None = None,
 ) -> str:
-    """Read a file as UTF-8 text, replacing invalid bytes rather than raising.
+    """Read a file as text, detecting encoding via BOM and stripping NUL bytes.
 
     Parameters
     ----------
@@ -127,15 +128,27 @@ def read_file_handle_unicode_errors(
     Returns
     -------
     str
-        File content as a string, with any invalid UTF-8 sequences replaced
-        by the Unicode replacement character.
+        File content as a string, with NUL bytes removed so the result is
+        safe to store in PostgreSQL text fields.
     """
     _log = log or logger
+    raw = filepath.read_bytes()
+
+    if raw.startswith((codecs.BOM_UTF16_LE, codecs.BOM_UTF16_BE)):
+        encoding = "utf-16"
+    elif raw.startswith(codecs.BOM_UTF8):
+        encoding = "utf-8-sig"
+    else:
+        encoding = "utf-8"
+
     try:
-        return filepath.read_text(encoding="utf-8")
+        text = raw.decode(encoding)
     except UnicodeDecodeError as e:
         _log.warning("Unicode error during text reading, continuing: %s", e)
-        return filepath.read_bytes().decode("utf-8", errors="replace")
+        text = raw.decode("utf-8", errors="replace")
+
+    # PostgreSQL rejects NUL (0x00) bytes in text fields
+    return text.replace("\x00", "")
 
 
 def get_page_count_for_pdf(

src/paperless/settings/__init__.py

@@ -97,8 +97,14 @@ MODEL_FILE = get_path_from_env(
     DATA_DIR / "classification_model.pickle",
 )
 LLM_INDEX_DIR = DATA_DIR / "llm_index"
-LLM_INDEX_LOCK = DATA_DIR / "locks" / "llm_index.lock"
-(DATA_DIR / "locks").mkdir(parents=True, exist_ok=True)
+LLM_INDEX_LOCK = LLM_INDEX_DIR / "index.lock"
+# Cross-process read/write lock guarding the LLM index compaction/migration
+# file swap. Readers hold it shared; the swap takes it exclusively so it never
+# runs while a reader connection is open. Must be a SQLite (.db) file.
+LLM_INDEX_RWLOCK = LLM_INDEX_DIR / "llmindex.rwlock.db"
+# Seconds the compaction swap waits for active readers to drain before skipping
+# this cycle (it is a maintenance operation; the next run retries).
+LLM_INDEX_COMPACTION_LOCK_TIMEOUT = 30
 
 LOGGING_DIR = get_path_from_env("PAPERLESS_LOGGING_DIR", DATA_DIR / "log")
 
@@ -644,6 +650,7 @@ LOGGING = {
         "kombu": {"handlers": ["file_celery"], "level": "DEBUG"},
         "_granian": {"handlers": ["file_paperless"], "level": "DEBUG"},
         "granian.access": {"handlers": ["file_paperless"], "level": "DEBUG"},
+        "httpx": {"level": "WARNING"},
     },
 }
 

src/paperless/settings/custom.py

@@ -252,6 +252,9 @@ def parse_db_settings(data_dir: Path) -> dict[str, dict[str, Any]]:
                 "NAME": os.getenv("PAPERLESS_DBNAME", "paperless"),
                 "USER": os.getenv("PAPERLESS_DBUSER", "paperless"),
                 "PASSWORD": os.getenv("PAPERLESS_DBPASS", "paperless"),
+                # Validate pooled connections so a connection closed server-side
+                # is replaced rather than handed out as "the connection is closed".
+                "CONN_HEALTH_CHECKS": True,
             }
 
             base_options = {

src/paperless/tests/settings/test_custom_parsers.py

@@ -398,6 +398,7 @@ class TestParseDbSettings:
                 {
                     "default": {
                         "ENGINE": "django.db.backends.postgresql",
+                        "CONN_HEALTH_CHECKS": True,
                         "HOST": "localhost",
                         "NAME": "paperless",
                         "USER": "paperless",
@@ -426,6 +427,7 @@ class TestParseDbSettings:
                 {
                     "default": {
                         "ENGINE": "django.db.backends.postgresql",
+                        "CONN_HEALTH_CHECKS": True,
                         "HOST": "paperless-db-host",
                         "PORT": 1111,
                         "NAME": "customdb",
@@ -455,6 +457,7 @@ class TestParseDbSettings:
                 {
                     "default": {
                         "ENGINE": "django.db.backends.postgresql",
+                        "CONN_HEALTH_CHECKS": True,
                         "HOST": "pghost",
                         "NAME": "paperless",
                         "USER": "paperless",
@@ -485,6 +488,7 @@ class TestParseDbSettings:
                 {
                     "default": {
                         "ENGINE": "django.db.backends.postgresql",
+                        "CONN_HEALTH_CHECKS": True,
                         "HOST": "pghost",
                         "NAME": "paperless",
                         "USER": "paperless",

src/paperless/tests/test_parser_utils.py

@@ -2,13 +2,50 @@
 
 from __future__ import annotations
 
+import codecs
 from pathlib import Path
 
 from paperless.parsers.utils import is_tagged_pdf
+from paperless.parsers.utils import read_file_handle_unicode_errors
 
 SAMPLES = Path(__file__).parent / "samples" / "tesseract"
 
 
+class TestReadFileHandleUnicodeErrors:
+    def test_plain_utf8(self, tmp_path: Path) -> None:
+        f = tmp_path / "plain.txt"
+        f.write_bytes(b"hello world")
+        assert read_file_handle_unicode_errors(f) == "hello world"
+
+    def test_utf8_bom(self, tmp_path: Path) -> None:
+        f = tmp_path / "bom.txt"
+        f.write_bytes(codecs.BOM_UTF8 + b"hello")
+        assert read_file_handle_unicode_errors(f) == "hello"
+
+    def test_utf16_le(self, tmp_path: Path) -> None:
+        f = tmp_path / "utf16le.txt"
+        f.write_bytes(codecs.BOM_UTF16_LE + "hello".encode("utf-16-le"))
+        assert read_file_handle_unicode_errors(f) == "hello"
+
+    def test_utf16_be(self, tmp_path: Path) -> None:
+        f = tmp_path / "utf16be.txt"
+        f.write_bytes(codecs.BOM_UTF16_BE + "hello".encode("utf-16-be"))
+        assert read_file_handle_unicode_errors(f) == "hello"
+
+    def test_nul_bytes_stripped(self, tmp_path: Path) -> None:
+        f = tmp_path / "null-bytes.txt"
+        f.write_bytes(b"foo\x00bar")
+        assert read_file_handle_unicode_errors(f) == "foobar"
+
+    def test_invalid_utf8_replaced(self, tmp_path: Path) -> None:
+        f = tmp_path / "bad.txt"
+        f.write_bytes(b"ok\x80\x81bad")
+        result = read_file_handle_unicode_errors(f)
+        assert "ok" in result
+        assert "bad" in result
+        assert "\x00" not in result
+
+
 class TestIsTaggedPdf:
     def test_tagged_pdf_returns_true(self) -> None:
         assert is_tagged_pdf(SAMPLES / "simple-digital.pdf") is True

src/paperless/views.py

@@ -49,7 +49,7 @@ from paperless.serialisers import GroupSerializer
 from paperless.serialisers import PaperlessAuthTokenSerializer
 from paperless.serialisers import ProfileSerializer
 from paperless.serialisers import UserSerializer
-from paperless_ai.indexing import vector_store_file_exists
+from paperless_ai.indexing import llm_index_exists
 
 
 class PaperlessObtainAuthTokenView(ObtainAuthToken):
@@ -467,7 +467,7 @@ class ApplicationConfigurationViewSet(ModelViewSet[ApplicationConfiguration]):
             or old_llm_context_size != new_llm_context_size
         )
         rebuild_needed = new_ai_index_enabled and (
-            not vector_store_file_exists() or embedding_config_changed
+            not llm_index_exists() or embedding_config_changed
         )
 
         if rebuild_needed:

src/paperless_ai/ai_classifier.py

@@ -1,15 +1,21 @@
 import json
 import logging
+from typing import TYPE_CHECKING
 
 from django.conf import settings
 from django.contrib.auth.models import User
 
 from documents.models import Document
-from documents.permissions import get_objects_for_user_owner_aware
 from paperless.config import AIConfig
 from paperless_ai.client import AIClient
+from paperless_ai.db import db_connection_released
 from paperless_ai.indexing import query_similar_documents
 from paperless_ai.indexing import truncate_content
+from paperless_ai.indexing import visible_document_ids_for_user
+from paperless_ai.taxonomy import format_hints_for_prompt
+
+if TYPE_CHECKING:
+    from paperless_ai.taxonomy import TaxonomyHints
 
 logger = logging.getLogger("paperless_ai.rag_classifier")
 
@@ -24,14 +30,26 @@ def get_language_name(language_code: str) -> str:
 
 def build_prompt_without_rag(
     document: Document,
+    config: AIConfig,
+    hints: "TaxonomyHints | None" = None,
 ) -> str:
     filename = document.filename or ""
-    content = truncate_content(document.content[:4000] or "")
+    content = truncate_content(
+        document.content[:4000] or "",
+        chunk_size=config.llm_embedding_chunk_size,
+        context_size=config.llm_context_size,
+    )
+
+    hints_block = format_hints_for_prompt(hints) if hints else ""
+    # Splice the block (if any) immediately before the "Analyze ..." instruction.
+    # When there is no block this expands to nothing, so the prompt is identical
+    # to the pre-hints baseline.
+    hints_section = f"{hints_block}\n\n    " if hints_block else ""
 
     return f"""
     You are a document classification assistant.
 
-    Analyze the following document and extract the following information:
+    {hints_section}Analyze the following document and extract the following information:
     - A short descriptive title
     - Tags that reflect the content
     - Names of people or organizations mentioned
@@ -49,10 +67,16 @@ def build_prompt_without_rag(
 
 def build_prompt_with_rag(
     document: Document,
+    config: AIConfig,
     user: User | None = None,
+    hints: "TaxonomyHints | None" = None,
 ) -> str:
-    base_prompt = build_prompt_without_rag(document)
-    context = truncate_content(get_context_for_document(document, user))
+    base_prompt = build_prompt_without_rag(document, config, hints=hints)
+    context = truncate_content(
+        get_context_for_document(document, user),
+        chunk_size=config.llm_embedding_chunk_size,
+        context_size=config.llm_context_size,
+    )
 
     return f"""{base_prompt}
 
@@ -85,20 +109,7 @@ def get_context_for_document(
     user: User | None = None,
     max_docs: int = 5,
 ) -> str:
-    visible_documents = (
-        get_objects_for_user_owner_aware(
-            user,
-            "view_document",
-            Document,
-        )
-        if user
-        else None
-    )
-    visible_document_ids = (
-        list(visible_documents.values_list("pk", flat=True))
-        if visible_documents is not None
-        else None
-    )
+    visible_document_ids = visible_document_ids_for_user(user)
     similar_docs = query_similar_documents(
         document=doc,
         document_ids=visible_document_ids,
@@ -126,30 +137,34 @@ def get_ai_document_classification(
     document: Document,
     user: User | None = None,
     output_language: str | None = None,
+    hints: "TaxonomyHints | None" = None,
 ) -> dict:
     ai_config = AIConfig()
 
     prompt = (
-        build_prompt_with_rag(document, user)
+        build_prompt_with_rag(document, ai_config, user, hints=hints)
         if ai_config.llm_embedding_backend
-        else build_prompt_without_rag(document)
+        else build_prompt_without_rag(document, ai_config, hints=hints)
     )
 
     client = AIClient()
-    result = client.run_llm_query(prompt)
-    suggestions = parse_ai_response(result)
-    if output_language:
-        localized = client.run_llm_query(
-            build_localization_prompt(suggestions, output_language),
-        )
-        localized_suggestions = parse_ai_response(localized)
-        suggestions = {
-            **suggestions,
-            "title": localized_suggestions["title"] or suggestions["title"],
-            "tags": localized_suggestions["tags"] or suggestions["tags"],
-            "document_types": localized_suggestions["document_types"]
-            or suggestions["document_types"],
-            "storage_paths": localized_suggestions["storage_paths"]
-            or suggestions["storage_paths"],
-        }
+    # Hand the pooled DB connection back while the (slow) LLM query runs so it
+    # is not pinned for the call's duration; see paperless_ai.db and #12976.
+    with db_connection_released():
+        result = client.run_llm_query(prompt)
+        suggestions = parse_ai_response(result)
+        if output_language:
+            localized = client.run_llm_query(
+                build_localization_prompt(suggestions, output_language),
+            )
+            localized_suggestions = parse_ai_response(localized)
+            suggestions = {
+                **suggestions,
+                "title": localized_suggestions["title"] or suggestions["title"],
+                "tags": localized_suggestions["tags"] or suggestions["tags"],
+                "document_types": localized_suggestions["document_types"]
+                or suggestions["document_types"],
+                "storage_paths": localized_suggestions["storage_paths"]
+                or suggestions["storage_paths"],
+            }
     return suggestions

src/paperless_ai/chat.py

@@ -3,9 +3,13 @@ import logging
 import sys
 
 from documents.models import Document
+from paperless.config import AIConfig
 from paperless_ai.client import AIClient
+from paperless_ai.db import db_connection_released
+from paperless_ai.indexing import _document_id_filters
 from paperless_ai.indexing import get_rag_prompt_helper
 from paperless_ai.indexing import load_or_build_index
+from paperless_ai.indexing import read_store
 
 logger = logging.getLogger("paperless_ai.chat")
 
@@ -75,148 +79,78 @@ def _format_chat_metadata_trailer(references: list[dict[str, int | str]]) -> str
     )
 
 
-def _get_document_filtered_retriever(index, doc_ids: set[str], similarity_top_k: int):
-    from llama_index.core.base.base_retriever import BaseRetriever
-    from llama_index.core.schema import NodeWithScore
-    from llama_index.core.vector_stores import VectorStoreQuery
-
-    class DocumentFilteredFaissRetriever(BaseRetriever):
-        def __init__(self):
-            super().__init__()
-            self._cached_query_str = None
-            self._cached_nodes = []
-
-        def _retrieve(self, query_bundle):
-            if query_bundle.query_str == self._cached_query_str:
-                return self._cached_nodes
-
-            if query_bundle.embedding is None:
-                query_bundle.embedding = (
-                    index._embed_model.get_agg_embedding_from_queries(
-                        query_bundle.embedding_strs,
-                    )
-                )
-
-            faiss_index = index.vector_store._faiss_index
-            max_top_k = faiss_index.ntotal
-            if max_top_k == 0:
-                self._cached_query_str = query_bundle.query_str
-                self._cached_nodes = []
-                return []
-
-            query_top_k = min(max(similarity_top_k, 1), max_top_k)
-            allowed_nodes: list[NodeWithScore] = []
-            seen_node_ids: set[str] = set()
-
-            while query_top_k <= max_top_k:
-                query_result = index.vector_store.query(
-                    VectorStoreQuery(
-                        query_embedding=query_bundle.embedding,
-                        similarity_top_k=query_top_k,
-                    ),
-                )
-
-                for vector_id, score in zip(
-                    query_result.ids or [],
-                    query_result.similarities or [],
-                    strict=False,
-                ):
-                    node_id = index.index_struct.nodes_dict.get(vector_id)
-                    if node_id is None or node_id in seen_node_ids:
-                        continue
-
-                    node = index.docstore.docs.get(node_id)
-                    if node is None or node.metadata.get("document_id") not in doc_ids:
-                        continue
-
-                    seen_node_ids.add(node_id)
-                    allowed_nodes.append(NodeWithScore(node=node, score=score))
-
-                    if len(allowed_nodes) >= similarity_top_k:
-                        self._cached_query_str = query_bundle.query_str
-                        self._cached_nodes = allowed_nodes
-                        return allowed_nodes
-
-                if query_top_k == max_top_k:
-                    self._cached_query_str = query_bundle.query_str
-                    self._cached_nodes = allowed_nodes
-                    return allowed_nodes
-
-                query_top_k = min(query_top_k * 2, max_top_k)
-
-            self._cached_query_str = query_bundle.query_str
-            self._cached_nodes = allowed_nodes
-            return allowed_nodes
-
-    return DocumentFilteredFaissRetriever()
-
-
 def stream_chat_with_documents(query_str: str, documents: list[Document]):
     try:
         yield from _stream_chat_with_documents(query_str, documents)
     except Exception as e:
-        logger.exception(f"Failed to stream document chat response: {e}", exc_info=True)
+        logger.exception("Failed to stream document chat response: %s", e)
         yield CHAT_ERROR_MESSAGE
 
 
 def _stream_chat_with_documents(query_str: str, documents: list[Document]):
-    client = AIClient()
-    index = load_or_build_index()
-
-    doc_ids = [str(doc.pk) for doc in documents]
-
-    # Filter only the node(s) that match the document IDs
-    nodes = [
-        node
-        for node in index.docstore.docs.values()
-        if node.metadata.get("document_id") in doc_ids
-    ]
-
-    if len(nodes) == 0:
-        logger.warning("No nodes found for the given documents.")
+    if not documents:
         yield CHAT_NO_CONTENT_MESSAGE
         return
 
     from llama_index.core.prompts import PromptTemplate
     from llama_index.core.query_engine import RetrieverQueryEngine
     from llama_index.core.response_synthesizers import get_response_synthesizer
+    from llama_index.core.retrievers import VectorIndexRetriever
 
-    retriever = _get_document_filtered_retriever(
-        index,
-        set(doc_ids),
-        CHAT_RETRIEVER_TOP_K,
-    )
+    config = AIConfig()
+    filters = _document_id_filters(str(doc.pk) for doc in documents)
 
-    top_nodes = retriever.retrieve(query_str)
-    if len(top_nodes) == 0:
-        logger.warning("Retriever returned no nodes for the given documents.")
-        yield CHAT_NO_CONTENT_MESSAGE
-        return
+    # Hold the shared read lock for the whole operation: the query engine
+    # retrieves from the vector store again during synthesis, so the connection
+    # must stay open (and the swap must not run) until the stream finishes.
+    with read_store() as store:
+        index = load_or_build_index(config, store)
+        retriever = VectorIndexRetriever(
+            index=index,
+            similarity_top_k=CHAT_RETRIEVER_TOP_K,
+            filters=filters,
+        )
 
-    references = _get_document_references(documents, top_nodes)
+        # Slow query-embedding + vector search; no Django ORM access happens
+        # during it, so release the pooled DB connection for its duration. See
+        # #12976.
+        with db_connection_released():
+            top_nodes = retriever.retrieve(query_str)
+        if not top_nodes:
+            logger.warning("No nodes found for the given documents.")
+            yield CHAT_NO_CONTENT_MESSAGE
+            return
 
-    prompt_template = PromptTemplate(template=CHAT_PROMPT_TMPL)
-    response_synthesizer = get_response_synthesizer(
-        llm=client.llm,
-        prompt_helper=get_rag_prompt_helper(),
-        text_qa_template=prompt_template,
-        streaming=True,
-    )
+        client = AIClient()
 
-    query_engine = RetrieverQueryEngine.from_args(
-        retriever=retriever,
-        llm=client.llm,
-        response_synthesizer=response_synthesizer,
-        streaming=True,
-    )
+        references = _get_document_references(documents, top_nodes)
 
-    logger.debug("Document chat query: %s", query_str)
+        prompt_template = PromptTemplate(template=CHAT_PROMPT_TMPL)
+        response_synthesizer = get_response_synthesizer(
+            llm=client.llm,
+            prompt_helper=get_rag_prompt_helper(
+                chunk_size=config.llm_embedding_chunk_size,
+                context_size=config.llm_context_size,
+            ),
+            text_qa_template=prompt_template,
+            streaming=True,
+        )
+        query_engine = RetrieverQueryEngine.from_args(
+            retriever=retriever,
+            llm=client.llm,
+            response_synthesizer=response_synthesizer,
+            streaming=True,
+        )
 
-    response_stream = query_engine.query(query_str)
+        logger.debug("Document chat query: %s", query_str)
+        # Release the pooled DB connection for the slow streaming LLM response
+        # so it is not pinned for the whole stream; see paperless_ai.db and
+        # #12976.
+        with db_connection_released():
+            response_stream = query_engine.query(query_str)
+            for chunk in response_stream.response_gen:
+                yield chunk
+                sys.stdout.flush()
 
-    for chunk in response_stream.response_gen:
-        yield chunk
-        sys.stdout.flush()
-
-    if references:
-        yield _format_chat_metadata_trailer(references)
+            if references:
+                yield _format_chat_metadata_trailer(references)

src/paperless_ai/db.py

@@ -0,0 +1,30 @@
+from __future__ import annotations
+
+from contextlib import contextmanager
+
+from django.db import connections
+
+
+@contextmanager
+def db_connection_released():
+    """
+    Return any checked-out DB connections to the pool for the duration of the
+    wrapped block.
+
+    The AI endpoints run inside a synchronous web request (``ai_suggestions``)
+    or a streaming response (``chat``). Django keeps the request's database
+    connection checked out for the entire request/response, so a blocking LLM
+    call - which can take many seconds - pins a pooled connection the whole
+    time. With connection pooling enabled, enough concurrent AI requests check
+    out every slot and all other requests then fail with
+    ``psycopg_pool.PoolTimeout`` (see issue #12976).
+
+    No Django ORM access happens during the LLM call, so we hand the connection
+    back to the pool first; Django transparently re-checks-out a connection on
+    the next ORM use after the block.
+    """
+    connections.close_all()
+    try:
+        yield
+    finally:
+        connections.close_all()

src/paperless_ai/embedding.py

@@ -1,12 +1,9 @@
-import json
 import re
 from typing import TYPE_CHECKING
 
 from django.conf import settings
 
 if TYPE_CHECKING:
-    from pathlib import Path
-
     from llama_index.core.base.embeddings.base import BaseEmbedding
 
 from documents.models import Document
@@ -23,9 +20,7 @@ OCR_LEADER_REGEX = re.compile(r"[._\-\u00b7]{4,}")
 HORIZONTAL_WHITESPACE_REGEX = re.compile(r"[ \t\u00a0]+")
 
 
-def get_embedding_model() -> "BaseEmbedding":
-    config = AIConfig()
-
+def get_embedding_model(config: AIConfig) -> "BaseEmbedding":
     match config.llm_embedding_backend:
         case LLMEmbeddingBackend.OPENAI_LIKE:
             from llama_index.embeddings.openai_like import OpenAILikeEmbedding
@@ -95,41 +90,24 @@ def get_embedding_model() -> "BaseEmbedding":
             )
 
 
-def get_embedding_dim() -> int:
-    """
-    Loads embedding dimension from meta.json if available, otherwise infers it
-    from a dummy embedding and stores it for future use.
-    """
-    config = AIConfig()
-    default_model = {
-        LLMEmbeddingBackend.OPENAI_LIKE: "text-embedding-3-small",
-        LLMEmbeddingBackend.HUGGINGFACE: "sentence-transformers/all-MiniLM-L6-v2",
-        LLMEmbeddingBackend.OLLAMA: "embeddinggemma",
-    }.get(
-        config.llm_embedding_backend,
-        "sentence-transformers/all-MiniLM-L6-v2",
+_DEFAULT_MODEL_NAMES = {
+    LLMEmbeddingBackend.OPENAI_LIKE: "text-embedding-3-small",
+    LLMEmbeddingBackend.HUGGINGFACE: "sentence-transformers/all-MiniLM-L6-v2",
+    LLMEmbeddingBackend.OLLAMA: "embeddinggemma",
+}
+
+
+def get_configured_model_name(config: AIConfig) -> str:
+    """Return the canonical name of the currently configured embedding model."""
+    # dict.get(key, default) overload resolution fails for TextChoices keys in some
+    # type checkers; use `or` fallback to avoid the ambiguity.
+    default = (
+        _DEFAULT_MODEL_NAMES.get(
+            config.llm_embedding_backend,
+        )
+        or "sentence-transformers/all-MiniLM-L6-v2"
     )
-    model = config.llm_embedding_model or default_model
-
-    meta_path: Path = settings.LLM_INDEX_DIR / "meta.json"
-    if meta_path.exists():
-        with meta_path.open() as f:
-            meta = json.load(f)
-        if meta.get("embedding_model") != model:
-            raise RuntimeError(
-                f"Embedding model changed from {meta.get('embedding_model')} to {model}. "
-                "You must rebuild the index.",
-            )
-        return meta["dim"]
-
-    embedding_model = get_embedding_model()
-    test_embed = embedding_model.get_text_embedding("test")
-    dim = len(test_embed)
-
-    with meta_path.open("w") as f:
-        json.dump({"embedding_model": model, "dim": dim}, f)
-
-    return dim
+    return config.llm_embedding_model or default
 
 
 def _normalize_llm_index_text(text: str) -> str:
@@ -138,17 +116,11 @@ def _normalize_llm_index_text(text: str) -> str:
 
 
 def build_llm_index_text(doc: Document) -> str:
+    # Short structured fields (filename, storage path, ASN, title, tags, ...) live
+    # in node.metadata: excluded from embeddings, shown to the LLM via metadata
+    # prepend. Notes and Custom Fields stay in the body: Notes can be long free
+    # text, Custom Fields are dynamic in count and best kept in the embedding.
     lines = [
-        f"Title: {doc.title}",
-        f"Filename: {doc.filename}",
-        f"Created: {doc.created}",
-        f"Added: {doc.added}",
-        f"Modified: {doc.modified}",
-        f"Tags: {', '.join(tag.name for tag in doc.tags.all())}",
-        f"Document Type: {doc.document_type.name if doc.document_type else ''}",
-        f"Correspondent: {doc.correspondent.name if doc.correspondent else ''}",
-        f"Storage Path: {doc.storage_path.name if doc.storage_path else ''}",
-        f"Archive Serial Number: {doc.archive_serial_number or ''}",
         f"Notes: {','.join([str(c.note) for c in Note.objects.filter(document=doc)])}",
     ]
 

src/paperless_ai/indexing.py

@@ -1,27 +1,32 @@
 import logging
-import shutil
-from collections import defaultdict
 from collections.abc import Iterable
+from contextlib import contextmanager
 from datetime import timedelta
-from pathlib import Path
 from typing import TYPE_CHECKING
 
 from django.conf import settings
+from django.contrib.auth.models import User
 from django.utils import timezone
 from filelock import FileLock
+from filelock import ReadWriteLock
+from filelock import Timeout
 
 from documents.models import Document
 from documents.models import PaperlessTask
+from documents.permissions import get_objects_for_user_owner_aware
 from documents.utils import IterWrapper
 from documents.utils import identity
 from paperless.config import AIConfig
+from paperless_ai.db import db_connection_released
 from paperless_ai.embedding import build_llm_index_text
-from paperless_ai.embedding import get_embedding_dim
+from paperless_ai.embedding import get_configured_model_name
 from paperless_ai.embedding import get_embedding_model
 
 if TYPE_CHECKING:
-    from llama_index.core import VectorStoreIndex
     from llama_index.core.schema import BaseNode
+    from llama_index.core.schema import NodeWithScore
+
+    from paperless_ai.vector_store import PaperlessSqliteVecVectorStore
 
 
 logger = logging.getLogger("paperless_ai.indexing")
@@ -30,21 +35,11 @@ RAG_NUM_OUTPUT = 512
 RAG_CHUNK_OVERLAP = 200
 
 
-def _index_lock_path() -> Path:
-    """Return the path used as the file lock for FAISS index mutations.
-
-    The lock file lives in DATA_DIR/locks/ (not inside LLM_INDEX_DIR) so that a
-    rebuild — which calls shutil.rmtree(LLM_INDEX_DIR) — cannot delete the lock
-    while another worker still holds it.
-    """
-    return settings.LLM_INDEX_LOCK
-
-
 def queue_llm_index_update_if_needed(*, rebuild: bool, reason: str) -> bool:
     # NOTE: The check-then-enqueue sequence below is non-atomic (TOCTOU): two
     # concurrent workers can both observe no running task and both enqueue a
     # full rebuild. This is wasteful but not data-corrupting — update_llm_index
-    # is itself protected by _index_lock_path(), so only one rebuild runs at a
+    # is itself protected by settings.LLM_INDEX_LOCK, so only one rebuild runs at a
     # time and the second one is serialised after the first completes.
     from documents.tasks import llmindex_index
 
@@ -71,46 +66,110 @@ def queue_llm_index_update_if_needed(*, rebuild: bool, reason: str) -> bool:
     return True
 
 
-def get_or_create_storage_context(*, rebuild=False):
-    """
-    Loads or creates the StorageContext (vector store, docstore, index store).
-    If rebuild=True, deletes and recreates everything.
-    """
-    if rebuild:
-        shutil.rmtree(settings.LLM_INDEX_DIR, ignore_errors=True)
-        settings.LLM_INDEX_DIR.mkdir(parents=True, exist_ok=True)
+def get_vector_store() -> "PaperlessSqliteVecVectorStore":
+    from paperless_ai.vector_store import PaperlessSqliteVecVectorStore
 
-    if rebuild or not settings.LLM_INDEX_DIR.exists():
-        import faiss
-        from llama_index.core import StorageContext
-        from llama_index.core.storage.docstore import SimpleDocumentStore
-        from llama_index.core.storage.index_store import SimpleIndexStore
-        from llama_index.vector_stores.faiss import FaissVectorStore
-
-        settings.LLM_INDEX_DIR.mkdir(parents=True, exist_ok=True)
-        embedding_dim = get_embedding_dim()
-        faiss_index = faiss.IndexFlatL2(embedding_dim)
-        vector_store = FaissVectorStore(faiss_index=faiss_index)
-        docstore = SimpleDocumentStore()
-        index_store = SimpleIndexStore()
-    else:
-        from llama_index.core import StorageContext
-        from llama_index.core.storage.docstore import SimpleDocumentStore
-        from llama_index.core.storage.index_store import SimpleIndexStore
-        from llama_index.vector_stores.faiss import FaissVectorStore
-
-        vector_store = FaissVectorStore.from_persist_dir(settings.LLM_INDEX_DIR)
-        docstore = SimpleDocumentStore.from_persist_dir(settings.LLM_INDEX_DIR)
-        index_store = SimpleIndexStore.from_persist_dir(settings.LLM_INDEX_DIR)
-
-    return StorageContext.from_defaults(
-        docstore=docstore,
-        index_store=index_store,
-        vector_store=vector_store,
-        persist_dir=settings.LLM_INDEX_DIR,
+    settings.LLM_INDEX_DIR.mkdir(parents=True, exist_ok=True)
+    return PaperlessSqliteVecVectorStore(
+        uri=str(settings.LLM_INDEX_DIR),
     )
 
 
+# --- LLM index locking ---------------------------------------------------
+#
+# Two locks guard the index; they answer different questions and are NOT
+# interchangeable:
+#
+# * settings.LLM_INDEX_LOCK (FileLock, exclusive) -- serializes WRITERS against
+#   each other, so only one rebuild/upsert/delete/compaction runs at a time.
+#   Taken by write_store(). Readers never take it, so it never blocks reads.
+#
+# * settings.LLM_INDEX_RWLOCK (ReadWriteLock) -- coordinates readers against the
+#   compaction/migration file swap. read_store() takes it SHARED (readers run
+#   concurrently); _exclude_readers() takes it EXCLUSIVE, only for the swap, so
+#   the database file is never replaced while a reader connection is open (that
+#   would alias the old WAL onto the new file and corrupt it).
+#
+#                    | vs another writer | vs a reader
+#   -----------------+-------------------+----------------------------
+#   normal write     | LLM_INDEX_LOCK    | nothing (WAL gives MVCC)
+#   compaction/swap  | LLM_INDEX_LOCK    | LLM_INDEX_RWLOCK (exclusive)
+#   reader           | nothing (WAL)     | LLM_INDEX_RWLOCK (shared)
+#
+# They can't be merged into one ReadWriteLock: a normal write must exclude other
+# writers WITHOUT blocking readers (WAL already gives reader/writer concurrency),
+# and ReadWriteLock has no "exclusive vs writers, shared vs readers" mode. Only
+# the swap needs to exclude readers.
+def _index_rwlock() -> ReadWriteLock:
+    """Return a fresh read/write lock instance for the index swap.
+
+    ``is_singleton=False`` so reads and the swap always coordinate through
+    SQLite (the actual cross-process case) rather than hitting the in-process
+    reentrant-upgrade guard; callers must ``close()`` it (the context managers
+    below do).
+    """
+    settings.LLM_INDEX_DIR.mkdir(parents=True, exist_ok=True)
+    return ReadWriteLock(str(settings.LLM_INDEX_RWLOCK), is_singleton=False)
+
+
+@contextmanager
+def read_store():
+    """Acquire the shared read lock and yield the vector store for a read.
+
+    The shared lock is held for the whole lifetime of the connection (and
+    closed on exit) so the compaction/migration swap, which takes the exclusive
+    lock, never runs while this connection is open. Concurrent readers do not
+    block each other; only the swap does.
+    """
+    lock = _index_rwlock()
+    try:
+        with lock.read_lock(), get_vector_store() as store:
+            yield store
+    finally:
+        lock.close()
+
+
+@contextmanager
+def _exclude_readers():
+    """Acquire exclusive index access, blocking until readers have drained.
+
+    The exclusive counterpart to ``read_store()``: a compaction or migration
+    must not run while any reader connection is open. Raises
+    :class:`filelock.Timeout` if active readers do not drain within
+    ``LLM_INDEX_COMPACTION_LOCK_TIMEOUT``; callers skip the operation on timeout.
+    """
+    lock = _index_rwlock()
+    try:
+        with lock.write_lock(timeout=settings.LLM_INDEX_COMPACTION_LOCK_TIMEOUT):
+            yield
+    finally:
+        lock.close()
+
+
+@contextmanager
+def write_store(embed_model_name: str | None = None):
+    """Acquire the write lock and yield the vector store.
+
+    All mutating operations (upsert, delete, rebuild, compact) must go through
+    this context manager to serialise concurrent Celery writers.
+    Read paths use ``read_store()`` so they hold the shared read lock.
+
+    Pass ``embed_model_name`` whenever the operation may create the table so
+    the model name is recorded in the schema metadata for future mismatch checks.
+    """
+    from paperless_ai.vector_store import PaperlessSqliteVecVectorStore
+
+    settings.LLM_INDEX_DIR.mkdir(parents=True, exist_ok=True)
+    with (
+        FileLock(settings.LLM_INDEX_LOCK),
+        PaperlessSqliteVecVectorStore(
+            uri=str(settings.LLM_INDEX_DIR),
+            embed_model_name=embed_model_name,
+        ) as store,
+    ):
+        yield store
+
+
 def build_document_node(
     document: Document,
     *,
@@ -130,6 +189,9 @@ def build_document_node(
         "document_type": document.document_type.name
         if document.document_type
         else None,
+        "filename": document.filename,
+        "storage_path": document.storage_path.name if document.storage_path else None,
+        "archive_serial_number": document.archive_serial_number,
         "created": document.created.isoformat() if document.created else None,
         "added": document.added.isoformat() if document.added else None,
         "modified": document.modified.isoformat(),
@@ -142,9 +204,11 @@ def build_document_node(
     # the token count and exceed embedding models with small context windows
     # (e.g. nomic-embed-text via Ollama defaults to num_ctx=2048).
     doc = LlamaDocument(
+        id_=str(document.id),
         text=text,
         metadata=metadata,
         excluded_embed_metadata_keys=list(metadata.keys()),
+        excluded_llm_metadata_keys=["document_id"],
     )
     chunk_size = chunk_size or get_rag_chunk_size()
     parser = SimpleNodeParser(
@@ -154,76 +218,33 @@ def build_document_node(
     return parser.get_nodes_from_documents([doc])
 
 
-def load_or_build_index(nodes=None):
-    """
-    Load an existing VectorStoreIndex if present,
-    or build a new one using provided nodes if storage is empty.
+def load_or_build_index(config: AIConfig, store: "PaperlessSqliteVecVectorStore"):
+    """Return a VectorStoreIndex backed by ``store``.
+
+    ``store`` is supplied by the caller's ``read_store()`` context so the shared
+    read lock and the connection stay alive for the whole retrieval.
     """
     import llama_index.core.settings as llama_settings
     from llama_index.core import VectorStoreIndex
-    from llama_index.core import load_index_from_storage
 
-    embed_model = get_embedding_model()
+    embed_model = get_embedding_model(config)
     llama_settings.Settings.embed_model = embed_model
-    storage_context = get_or_create_storage_context()
-    try:
-        return load_index_from_storage(storage_context=storage_context)
-    except ValueError as e:
-        logger.warning("Failed to load index from storage: %s", e)
-        if not nodes:
-            queue_llm_index_update_if_needed(
-                rebuild=vector_store_file_exists(),
-                reason="LLM index missing or invalid while loading.",
-            )
-            logger.info("No nodes provided for index creation.")
-            raise
-        return VectorStoreIndex(
-            nodes=nodes,
-            storage_context=storage_context,
-            embed_model=embed_model,
-        )
+    return VectorStoreIndex.from_vector_store(
+        vector_store=store,
+        embed_model=embed_model,
+    )
 
 
-def remove_document_docstore_nodes(document: Document, index: "VectorStoreIndex"):
-    """
-    Removes existing documents from docstore for a given document from the index.
-    This is necessary because FAISS IndexFlatL2 is append-only.
-    """
-    all_node_ids = list(index.docstore.docs.keys())
-    existing_nodes = [
-        node.node_id
-        for node in index.docstore.get_nodes(all_node_ids)
-        if node.metadata.get("document_id") == str(document.id)
-    ]
-    for node_id in existing_nodes:
-        # Delete from docstore, FAISS IndexFlatL2 are append-only
-        index.docstore.delete_document(node_id)
-        # Also purge the FAISS position -> UUID mapping so subsequent similarity
-        # queries don't raise KeyError on ghost vector positions.
-        stale_keys = [
-            k for k, v in index.index_struct.nodes_dict.items() if v == node_id
-        ]
-        for key in stale_keys:
-            del index.index_struct.nodes_dict[key]
-    # Re-sync the mutated index_struct so persist() writes the updated nodes_dict.
-    index.storage_context.index_store.add_index_struct(index.index_struct)
-
-
-def vector_store_file_exists():
-    """
-    Check if the vector store file exists in the LLM index directory.
-    """
-    return Path(settings.LLM_INDEX_DIR / "default__vector_store.json").exists()
+def llm_index_exists() -> bool:
+    """True when the index table exists on disk."""
+    with read_store() as store:
+        return store.table_exists()
 
 
 def get_rag_chunk_size() -> int:
     return AIConfig().llm_embedding_chunk_size
 
 
-def get_rag_context_size() -> int:
-    return AIConfig().llm_context_size
-
-
 def get_rag_chunk_overlap(chunk_size: int | None = None) -> int:
     chunk_size = chunk_size or get_rag_chunk_size()
     return min(RAG_CHUNK_OVERLAP, chunk_size - 1)
@@ -249,123 +270,149 @@ def get_rag_prompt_helper(
     )
 
 
+def _embed_nodes(nodes: list["BaseNode"], embed_model) -> None:
+    """Embed ``nodes`` in place using ``embed_model``."""
+    from llama_index.core.schema import MetadataMode
+
+    texts = [n.get_content(metadata_mode=MetadataMode.EMBED) for n in nodes]
+    for node, emb in zip(
+        nodes,
+        embed_model.get_text_embedding_batch(texts),
+        strict=True,
+    ):
+        node.embedding = emb
+
+
+def _document_id_filters(doc_ids):
+    """Return a MetadataFilters IN filter scoped to ``doc_ids``."""
+    from llama_index.core.vector_stores.types import FilterOperator
+    from llama_index.core.vector_stores.types import MetadataFilter
+    from llama_index.core.vector_stores.types import MetadataFilters
+
+    return MetadataFilters(
+        filters=[
+            MetadataFilter(
+                key="document_id",
+                operator=FilterOperator.IN,
+                value=sorted(doc_ids),
+            ),
+        ],
+    )
+
+
 def update_llm_index(
     *,
     iter_wrapper: IterWrapper[Document] = identity,
     rebuild=False,
 ) -> str:
-    """
-    Rebuild or update the LLM index.
-    """
-    from llama_index.core import VectorStoreIndex
-
-    nodes = []
-
+    """Rebuild or incrementally update the LLM index."""
+    with write_store() as store:
+        try:
+            with _exclude_readers():
+                needs_reembed = store.check_and_run_migrations()
+        except Timeout:
+            logger.info(
+                "Skipping LLM index migration check: index readers are active; "
+                "will retry next run.",
+            )
+            needs_reembed = False
+        if needs_reembed:
+            logger.warning(
+                "LLM index migration requires re-embedding; forcing rebuild.",
+            )
+            rebuild = True
     documents = Document.objects.all()
-    if not documents.exists():
+    no_documents = not documents.exists()
+
+    # Fast exit before touching config: nothing to index and no existing index.
+    if no_documents and not rebuild and not llm_index_exists():
         logger.warning("No documents found to index.")
-        if not rebuild and not vector_store_file_exists():
-            return "No documents found to index."
+        return "No documents found to index."
 
     config = AIConfig()
+    model_name = get_configured_model_name(config)
+
+    if not rebuild and llm_index_exists():
+        with read_store() as store:
+            config_mismatch = store.config_mismatch(model_name)
+        if config_mismatch:
+            logger.warning("Embedding model changed; forcing LLM index rebuild.")
+            rebuild = True
+
+    if no_documents:
+        logger.warning("No documents found to index.")
+
     chunk_size = config.llm_embedding_chunk_size
+    embed_model = get_embedding_model(config)
 
-    with FileLock(_index_lock_path()):
-        if rebuild or not vector_store_file_exists():
-            # remove meta.json to force re-detection of embedding dim
-            (settings.LLM_INDEX_DIR / "meta.json").unlink(missing_ok=True)
-            # Rebuild index from scratch
+    with write_store(embed_model_name=model_name) as store:
+        if rebuild or not store.table_exists():
             logger.info("Rebuilding LLM index.")
-            import llama_index.core.settings as llama_settings
-
-            embed_model = get_embedding_model()
-            llama_settings.Settings.embed_model = embed_model
-            storage_context = get_or_create_storage_context(rebuild=True)
+            store.drop_table()
             for document in iter_wrapper(documents):
-                document_nodes = build_document_node(document, chunk_size=chunk_size)
-                nodes.extend(document_nodes)
-
-            index = VectorStoreIndex(
-                nodes=nodes,
-                storage_context=storage_context,
-                embed_model=embed_model,
-                show_progress=False,
-            )
+                nodes = build_document_node(document, chunk_size=chunk_size)
+                _embed_nodes(nodes, embed_model)
+                store.add(nodes)
             msg = "LLM index rebuilt successfully."
         else:
-            # Update existing index
-            index = load_or_build_index()
-            existing_nodes: defaultdict[str, list] = defaultdict(list)
-            for node in index.docstore.docs.values():
-                doc_id = node.metadata.get("document_id")
-                if doc_id is not None:
-                    existing_nodes[doc_id].append(node)
-
+            existing = store.get_modified_times()
+            changed = 0
             for document in iter_wrapper(documents):
                 doc_id = str(document.id)
-                document_modified = document.modified.isoformat()
+                if existing.get(doc_id) == document.modified.isoformat():
+                    continue
+                nodes = build_document_node(document, chunk_size=chunk_size)
+                _embed_nodes(nodes, embed_model)
+                store.upsert_document(doc_id, nodes)
+                changed += 1
+            msg = (
+                "LLM index updated successfully."
+                if changed
+                else "No changes detected in LLM index."
+            )
 
-                if doc_id in existing_nodes:
-                    doc_nodes = existing_nodes[doc_id]
-                    node_modified = doc_nodes[0].metadata.get("modified")
-
-                    if node_modified == document_modified:
-                        continue
-
-                    # Delete from docstore, FAISS IndexFlatL2 are append-only
-                    for node in doc_nodes:
-                        remove_document_docstore_nodes(document, index)
-
-                nodes.extend(build_document_node(document, chunk_size=chunk_size))
-
-            if nodes:
-                msg = "LLM index updated successfully."
-                logger.info(
-                    "Updating %d nodes in LLM index.",
-                    len(nodes),
-                )
-                index.insert_nodes(nodes)
-            else:
-                msg = "No changes detected in LLM index."
-                logger.info(msg)
-
-        index.storage_context.persist(persist_dir=settings.LLM_INDEX_DIR)
+        try:
+            with _exclude_readers():
+                store.compact()
+        except Timeout:
+            logger.info(
+                "Skipping LLM index compaction: index readers are active; "
+                "will retry next run.",
+            )
     return msg
 
 
 def llm_index_add_or_update_document(document: Document):
-    """
-    Adds or updates a document in the LLM index.
-    If the document already exists, it will be replaced.
-    """
-    new_nodes = build_document_node(document, chunk_size=get_rag_chunk_size())
-    if not new_nodes:
-        logger.warning(
-            "No indexable content for document %s; skipping LLM index update.",
-            document.pk,
-        )
-        return
+    """Add or atomically replace a document's chunks in the index."""
+    config = AIConfig()
+    new_nodes = build_document_node(
+        document,
+        chunk_size=config.llm_embedding_chunk_size,
+    )
+    if new_nodes:
+        _embed_nodes(new_nodes, get_embedding_model(config))
 
-    with FileLock(_index_lock_path()):
-        index = load_or_build_index(nodes=new_nodes)
+    with write_store(embed_model_name=get_configured_model_name(config)) as store:
+        store.upsert_document(str(document.id), new_nodes)
 
-        remove_document_docstore_nodes(document, index)
 
-        index.insert_nodes(new_nodes)
-
-        index.storage_context.persist(persist_dir=settings.LLM_INDEX_DIR)
+def llm_index_compact() -> None:
+    """Compact the index immediately, rebuilding the table to reclaim space."""
+    with write_store() as store:
+        try:
+            with _exclude_readers():
+                store.compact(force=True)
+        except Timeout:
+            logger.info(
+                "Skipping LLM index compaction: index readers are active; "
+                "will retry next run.",
+            )
 
 
 def llm_index_remove_document(document: Document):
-    """
-    Removes a document from the LLM index.
-    """
-    with FileLock(_index_lock_path()):
-        index = load_or_build_index()
-
-        remove_document_docstore_nodes(document, index)
-
-        index.storage_context.persist(persist_dir=settings.LLM_INDEX_DIR)
+    """Remove a document's chunks from the LLM index."""
+    with write_store() as store:
+        store.delete(str(document.id))
 
 
 def truncate_content(
@@ -405,82 +452,102 @@ def normalize_document_ids(document_ids: Iterable[int | str] | None) -> set[str]
     return {str(document_id) for document_id in document_ids}
 
 
-def query_similar_documents(
-    document: Document,
-    top_k: int = 5,
-    document_ids: Iterable[int | str] | None = None,
-) -> list[Document]:
+def visible_document_ids_for_user(user: User | None) -> list[int] | None:
+    """Return the pks of documents ``user`` may view, or ``None`` for no filter.
+
+    Returns ``None`` when ``user`` is ``None`` so retrieval runs unfiltered. Used
+    by both the similarity-context and taxonomy-hints paths to scope RAG
+    neighbours to documents the requesting user is allowed to see.
     """
-    Runs a similarity query and returns top-k similar Document objects.
+    if user is None:
+        return None
+    visible_documents = get_objects_for_user_owner_aware(
+        user,
+        "view_document",
+        Document,
+    )
+    return list(visible_documents.values_list("pk", flat=True))
+
+
+def retrieve_similar_nodes(
+    document: Document,
+    document_ids: Iterable[int | str] | None = None,
+    top_k: int = 5,
+) -> list["NodeWithScore"]:
+    """Run ANN retrieval and return the raw NodeWithScore results.
+
+    Returns ``[]`` when the allow-list normalizes to empty, or when no index
+    exists yet (queuing a build in that case). The ``retrieve()`` call is a slow
+    embedding request, so it runs inside ``db_connection_released()`` to avoid
+    pinning the pooled DB connection (#12976). Both ``query_similar_documents``
+    and the taxonomy-hints path go through here, so they share that behavior.
     """
     allowed_document_ids = normalize_document_ids(document_ids)
     if allowed_document_ids is not None and not allowed_document_ids:
         return []
 
-    if not vector_store_file_exists():
+    if not llm_index_exists():
         queue_llm_index_update_if_needed(
             rebuild=False,
             reason="LLM index not found for similarity query.",
         )
         return []
 
-    with FileLock(_index_lock_path()):
-        index = load_or_build_index()
+    config = AIConfig()
 
-        # constrain only the node(s) that match the document IDs, if given
-        doc_node_ids = (
-            [
-                node.node_id
-                for node in index.docstore.docs.values()
-                if node.metadata.get("document_id") in allowed_document_ids
-            ]
-            if allowed_document_ids is not None
-            else None
-        )
-        if doc_node_ids is not None and not doc_node_ids:
-            return []
+    from llama_index.core.retrievers import VectorIndexRetriever
 
-        from llama_index.core.retrievers import VectorIndexRetriever
+    filters = (
+        _document_id_filters(allowed_document_ids)
+        if allowed_document_ids is not None
+        else None
+    )
 
+    query_text = truncate_content(
+        (document.title or "") + "\n" + (document.content or ""),
+        chunk_size=config.llm_embedding_chunk_size,
+        context_size=config.llm_context_size,
+    )
+    # Hold the shared read lock for the whole retrieval so the connection is
+    # never open across a compaction swap. The retrieve() call generates a
+    # query embedding (a slow external request) and searches the vector store;
+    # no Django ORM access happens during it, so release the pooled DB
+    # connection for its duration. See #12976.
+    with read_store() as store:
+        index = load_or_build_index(config, store)
         retriever = VectorIndexRetriever(
             index=index,
             similarity_top_k=top_k,
-            doc_ids=doc_node_ids,
+            filters=filters,
         )
+        with db_connection_released():
+            return retriever.retrieve(query_text)
 
-        config = AIConfig()
-        query_text = truncate_content(
-            (document.title or "") + "\n" + (document.content or ""),
-            chunk_size=config.llm_embedding_chunk_size,
-            context_size=config.llm_context_size,
-        )
-        try:
-            results = retriever.retrieve(query_text)
-        except KeyError as e:
-            # Ghost FAISS positions remain after deletion because IndexFlatL2 is
-            # append-only. Treat them as absent and return no results.
-            logger.debug(
-                "Skipping LLM similarity query for document %s due to a stale "
-                "FAISS position with no docstore node: %s",
-                document.pk,
-                e,
-            )
-            return []
+
+def query_similar_documents(
+    document: Document,
+    top_k: int = 5,
+    document_ids: Iterable[int | str] | None = None,
+) -> list[Document]:
+    """Return up to ``top_k`` Documents most similar to ``document``."""
+    allowed_document_ids = normalize_document_ids(document_ids)
+    results = retrieve_similar_nodes(
+        document=document,
+        document_ids=allowed_document_ids,
+        top_k=top_k,
+    )
 
     retrieved_document_ids: list[int] = []
     for node in results:
         document_id = node.metadata.get("document_id")
         if document_id is None:
             continue
-        normalized_document_id = str(document_id)
-        if (
-            allowed_document_ids is not None
-            and normalized_document_id not in allowed_document_ids
-        ):
+        normalized = str(document_id)
+        if allowed_document_ids is not None and normalized not in allowed_document_ids:
             continue
         try:
-            retrieved_document_ids.append(int(normalized_document_id))
-        except ValueError:
+            retrieved_document_ids.append(int(normalized))
+        except ValueError:  # pragma: no cover
             logger.warning(
                 "Skipping LLM index result with invalid document_id %r.",
                 document_id,

src/paperless_ai/matching.py

@@ -15,40 +15,56 @@ MATCH_THRESHOLD = 0.8
 logger = logging.getLogger("paperless_ai.matching")
 
 
-def match_tags_by_name(names: list[str], user: User) -> list[Tag]:
+def match_tags_by_name(
+    names: list[str],
+    user: User,
+    hinted_names: set[str] | None = None,
+) -> list[Tag]:
     queryset = get_objects_for_user_owner_aware(
         user,
         ["view_tag"],
         Tag,
     )
-    return _match_names_to_queryset(names, queryset, "name")
+    return _match_names_to_queryset(names, queryset, "name", hinted_names)
 
 
-def match_correspondents_by_name(names: list[str], user: User) -> list[Correspondent]:
+def match_correspondents_by_name(
+    names: list[str],
+    user: User,
+    hinted_names: set[str] | None = None,
+) -> list[Correspondent]:
     queryset = get_objects_for_user_owner_aware(
         user,
         ["view_correspondent"],
         Correspondent,
     )
-    return _match_names_to_queryset(names, queryset, "name")
+    return _match_names_to_queryset(names, queryset, "name", hinted_names)
 
 
-def match_document_types_by_name(names: list[str], user: User) -> list[DocumentType]:
+def match_document_types_by_name(
+    names: list[str],
+    user: User,
+    hinted_names: set[str] | None = None,
+) -> list[DocumentType]:
     queryset = get_objects_for_user_owner_aware(
         user,
         ["view_documenttype"],
         DocumentType,
     )
-    return _match_names_to_queryset(names, queryset, "name")
+    return _match_names_to_queryset(names, queryset, "name", hinted_names)
 
 
-def match_storage_paths_by_name(names: list[str], user: User) -> list[StoragePath]:
+def match_storage_paths_by_name(
+    names: list[str],
+    user: User,
+    hinted_names: set[str] | None = None,
+) -> list[StoragePath]:
     queryset = get_objects_for_user_owner_aware(
         user,
         ["view_storagepath"],
         StoragePath,
     )
-    return _match_names_to_queryset(names, queryset, "name")
+    return _match_names_to_queryset(names, queryset, "name", hinted_names)
 
 
 def _normalize(s: str) -> str:
@@ -58,10 +74,18 @@ def _normalize(s: str) -> str:
     return s
 
 
-def _match_names_to_queryset(names: list[str], queryset, attr: str):
+def _match_names_to_queryset(
+    names: list[str],
+    queryset,
+    attr: str,
+    hinted_names: set[str] | None = None,
+):
     results = []
     objects = list(queryset)
     object_names = [_normalize(getattr(obj, attr)) for obj in objects]
+    normalized_hints = (
+        {_normalize(name) for name in hinted_names} if hinted_names else set()
+    )
 
     for name in names:
         if not name:
@@ -76,6 +100,11 @@ def _match_names_to_queryset(names: list[str], queryset, attr: str):
             results.append(matched)
             continue
 
+        # A hinted name that didn't exact-match came from existing taxonomy
+        # verbatim; do not fuzzy-map it onto a different object.
+        if target in normalized_hints:
+            continue
+
         # Fuzzy match fallback
         matches = difflib.get_close_matches(
             target,
@@ -88,8 +117,6 @@ def _match_names_to_queryset(names: list[str], queryset, attr: str):
             matched = objects.pop(index)
             object_names.pop(index)
             results.append(matched)
-        else:
-            pass
     return results
 
 

src/paperless_ai/taxonomy.py

@@ -0,0 +1,115 @@
+from typing import TYPE_CHECKING
+from typing import TypedDict
+
+from django.contrib.auth.models import User
+
+from documents.models import Document
+from paperless.config import AIConfig
+from paperless_ai.indexing import retrieve_similar_nodes
+from paperless_ai.indexing import visible_document_ids_for_user
+
+if TYPE_CHECKING:
+    from llama_index.core.schema import NodeWithScore
+
+
+class TaxonomyHints(TypedDict):
+    tags: list[str]
+    document_types: list[str]
+    correspondents: list[str]
+    storage_paths: list[str]
+
+
+def build_taxonomy_hints_from_nodes(
+    nodes: list["NodeWithScore"],
+) -> TaxonomyHints:
+    """Collect the unique, sorted taxonomy names carried on retrieved nodes.
+
+    Reads ``tags`` (a list), ``document_type``, ``correspondent``, and
+    ``storage_path`` from each node's metadata. Empty / ``None`` values and
+    missing keys are skipped. The result is naturally bounded by the retrieval
+    ``top_k``, so no cap is applied.
+    """
+    tags: set[str] = set()
+    document_types: set[str] = set()
+    correspondents: set[str] = set()
+    storage_paths: set[str] = set()
+
+    for node in nodes:
+        metadata = node.metadata or {}
+
+        for tag in metadata.get("tags") or []:
+            if tag:
+                tags.add(tag)
+
+        document_type = metadata.get("document_type")
+        if document_type:
+            document_types.add(document_type)
+
+        correspondent = metadata.get("correspondent")
+        if correspondent:
+            correspondents.add(correspondent)
+
+        storage_path = metadata.get("storage_path")
+        if storage_path:
+            storage_paths.add(storage_path)
+
+    return TaxonomyHints(
+        tags=sorted(tags),
+        document_types=sorted(document_types),
+        correspondents=sorted(correspondents),
+        storage_paths=sorted(storage_paths),
+    )
+
+
+_HINT_INSTRUCTION = (
+    "Prefer existing names from these lists verbatim. Only propose a new value "
+    "if none of the existing names fits."
+)
+
+
+def format_hints_for_prompt(hints: TaxonomyHints) -> str:
+    """Render non-empty hint categories as labelled blocks plus one instruction.
+
+    Returns "" when every category is empty, so callers can treat the result
+    the same as no hints at all.
+    """
+    # Literal-key access keeps this TypedDict-safe for mypy; the order here is
+    # the order the blocks appear in the prompt.
+    labelled_values: list[tuple[str, list[str]]] = [
+        ("Available tags", hints["tags"]),
+        ("Available document types", hints["document_types"]),
+        ("Available correspondents", hints["correspondents"]),
+        ("Available storage paths", hints["storage_paths"]),
+    ]
+    blocks: list[str] = []
+    for label, values in labelled_values:
+        if values:
+            listing = "\n".join(f"- {value}" for value in values)
+            blocks.append(f"{label}:\n{listing}")
+
+    if not blocks:
+        return ""
+
+    return "\n\n".join([*blocks, _HINT_INSTRUCTION])
+
+
+def get_taxonomy_hints_for_document(
+    document: Document,
+    user: User | None,
+) -> TaxonomyHints | None:
+    """Build taxonomy hints from a document's RAG neighbours.
+
+    Returns ``None`` when no embedding backend is configured (the gate) so the
+    caller's prompt and matching are identical to today. Otherwise returns a
+    ``TaxonomyHints`` -- possibly all-empty when no similar documents exist.
+    Applies the same owner-aware visible-document filter as
+    ``get_context_for_document``.
+    """
+    if not AIConfig().llm_embedding_backend:
+        return None
+
+    nodes = retrieve_similar_nodes(
+        document=document,
+        document_ids=visible_document_ids_for_user(user),
+    )
+    return build_taxonomy_hints_from_nodes(nodes)

src/paperless_ai/tests/conftest.py

@@ -1,10 +1,36 @@
 from pathlib import Path
 
 import pytest
+import pytest_mock
+from llama_index.core.base.embeddings.base import BaseEmbedding
 from pytest_django.fixtures import SettingsWrapper
 
 
 @pytest.fixture
-def temp_llm_index_dir(tmp_path: Path, settings: SettingsWrapper):
+def temp_llm_index_dir(tmp_path: Path, settings: SettingsWrapper) -> Path:
     settings.LLM_INDEX_DIR = tmp_path
+    settings.LLM_INDEX_LOCK = tmp_path / "index.lock"
+    settings.LLM_INDEX_RWLOCK = tmp_path / "llmindex.rwlock.db"
     return tmp_path
+
+
+class FakeEmbedding(BaseEmbedding):
+    async def _aget_query_embedding(self, query: str) -> list[float]:
+        return [0.1] * self.get_query_embedding_dim()
+
+    def _get_query_embedding(self, query: str) -> list[float]:
+        return [0.1] * self.get_query_embedding_dim()
+
+    def _get_text_embedding(self, text: str) -> list[float]:
+        return [0.1] * self.get_query_embedding_dim()
+
+    def get_query_embedding_dim(self) -> int:
+        return 384
+
+
+@pytest.fixture
+def mock_embed_model(mocker: pytest_mock.MockerFixture) -> pytest_mock.MockType:
+    fake = FakeEmbedding()
+    mocker.patch("paperless_ai.indexing.get_embedding_model", return_value=fake)
+    mocker.patch("paperless_ai.embedding.get_embedding_model", return_value=fake)
+    return fake

src/paperless_ai/tests/test_ai_classifier.py

@@ -1,11 +1,15 @@
 import json
+from types import SimpleNamespace
+from typing import cast
 from unittest.mock import MagicMock
 from unittest.mock import patch
 
 import pytest
+import pytest_mock
 from django.test import override_settings
 
 from documents.models import Document
+from paperless.config import AIConfig
 from paperless_ai.ai_classifier import build_localization_prompt
 from paperless_ai.ai_classifier import build_prompt_with_rag
 from paperless_ai.ai_classifier import build_prompt_without_rag
@@ -211,11 +215,12 @@ def test_prompt_with_without_rag(mock_document):
         "paperless_ai.ai_classifier.get_context_for_document",
         return_value="Context from similar documents",
     ):
-        prompt = build_prompt_without_rag(mock_document)
+        config = AIConfig()
+        prompt = build_prompt_without_rag(mock_document, config)
         assert "Additional context from similar documents" not in prompt
         assert "for generated" not in prompt
 
-        prompt = build_prompt_with_rag(mock_document)
+        prompt = build_prompt_with_rag(mock_document, config)
         assert "Additional context from similar documents" in prompt
 
         prompt = build_localization_prompt(
@@ -259,3 +264,111 @@ def test_get_context_for_document_no_similar_docs(mock_document):
     with patch("paperless_ai.ai_classifier.query_similar_documents", return_value=[]):
         result = get_context_for_document(mock_document)
         assert result == ""
+
+
+class TestPromptHints:
+    @pytest.fixture
+    def config(self) -> AIConfig:
+        # build_prompt_* only read these two numeric settings off config;
+        # a stand-in avoids constructing a DB-backed AIConfig.
+        return cast(
+            "AIConfig",
+            SimpleNamespace(llm_embedding_chunk_size=1000, llm_context_size=8000),
+        )
+
+    def test_without_rag_includes_hints_block(
+        self,
+        mock_document: MagicMock,
+        config: AIConfig,
+    ) -> None:
+        hints = {
+            "tags": ["Bloodwork"],
+            "document_types": ["Invoice"],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+        prompt = build_prompt_without_rag(mock_document, config, hints=hints)
+        assert "Available tags:" in prompt
+        assert "- Bloodwork" in prompt
+        assert "Prefer existing names from these lists verbatim" in prompt
+
+    def test_without_rag_none_matches_baseline(
+        self,
+        mock_document: MagicMock,
+        config: AIConfig,
+    ) -> None:
+        baseline = build_prompt_without_rag(mock_document, config)
+        with_none = build_prompt_without_rag(mock_document, config, hints=None)
+        assert with_none == baseline
+        assert "Available tags:" not in with_none
+
+    def test_with_rag_includes_context_and_hints(
+        self,
+        mock_document: MagicMock,
+        config: AIConfig,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.ai_classifier.get_context_for_document",
+            return_value="TITLE: Neighbour\nsome context",
+        )
+        hints = {
+            "tags": ["Bloodwork"],
+            "document_types": [],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+        prompt = build_prompt_with_rag(mock_document, config, user=None, hints=hints)
+        assert "Additional context from similar documents" in prompt
+        assert "Available tags:" in prompt
+
+    def test_classification_forwards_hints(
+        self,
+        mock_document: MagicMock,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.ai_classifier.AIConfig",
+            return_value=SimpleNamespace(
+                llm_embedding_backend=None,
+                llm_embedding_chunk_size=1000,
+                llm_context_size=8000,
+            ),
+        )
+        build = mocker.patch(
+            "paperless_ai.ai_classifier.build_prompt_without_rag",
+            return_value="PROMPT",
+        )
+        mock_client = MagicMock()
+        mock_client.run_llm_query.return_value = {
+            "title": "t",
+            "tags": [],
+            "correspondents": [],
+            "document_types": [],
+            "storage_paths": [],
+            "dates": [],
+        }
+        mocker.patch("paperless_ai.ai_classifier.AIClient", return_value=mock_client)
+        hints = {
+            "tags": ["Bloodwork"],
+            "document_types": [],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+
+        result = get_ai_document_classification(
+            mock_document,
+            user=None,
+            hints=hints,
+        )
+
+        _, build_kwargs = build.call_args
+        assert build_kwargs["hints"] == hints
+        assert set(result.keys()) == {
+            "title",
+            "tags",
+            "correspondents",
+            "document_types",
+            "storage_paths",
+            "dates",
+        }

src/paperless_ai/tests/test_ai_indexing.py

@@ -1,15 +1,13 @@
-import json
 from pathlib import Path
+from types import SimpleNamespace
 from unittest.mock import MagicMock
 from unittest.mock import patch
 
 import pytest
 import pytest_mock
-from django.contrib.auth.models import User
 from django.test import override_settings
 from django.utils import timezone
-from faker import Faker
-from llama_index.core.base.embeddings.base import BaseEmbedding
+from llama_index.core.schema import MetadataMode
 
 from documents.models import Document
 from documents.models import PaperlessTask
@@ -19,10 +17,12 @@ from documents.tests.factories import DocumentFactory
 from documents.tests.factories import PaperlessTaskFactory
 from paperless.models import ApplicationConfiguration
 from paperless_ai import indexing
+from paperless_ai.tests.conftest import FakeEmbedding
+from paperless_ai.vector_store import PaperlessSqliteVecVectorStore
 
 
 @pytest.fixture
-def real_document(db):
+def real_document(db: None) -> Document:
     return Document.objects.create(
         title="Test Document",
         content="This is some test content.",
@@ -30,44 +30,39 @@ def real_document(db):
     )
 
 
-@pytest.fixture
-def mock_embed_model():
-    fake = FakeEmbedding()
-    with (
-        patch("paperless_ai.indexing.get_embedding_model") as mock_index,
-        patch(
-            "paperless_ai.embedding.get_embedding_model",
-        ) as mock_embedding,
-    ):
-        mock_index.return_value = fake
-        mock_embedding.return_value = fake
-        yield mock_index
-
-
-class FakeEmbedding(BaseEmbedding):
-    # TODO: maybe a better way to do this?
-    def _aget_query_embedding(self, query: str) -> list[float]:
-        return [0.1] * self.get_query_embedding_dim()
-
-    def _get_query_embedding(self, query: str) -> list[float]:
-        return [0.1] * self.get_query_embedding_dim()
-
-    def _get_text_embedding(self, text: str) -> list[float]:
-        return [0.1] * self.get_query_embedding_dim()
-
-    def get_query_embedding_dim(self) -> int:
-        return 384  # Match your real FAISS config
-
-
 @pytest.mark.django_db
-def test_build_document_node(real_document) -> None:
+def test_build_document_node(real_document: Document) -> None:
     nodes = indexing.build_document_node(real_document)
     assert len(nodes) > 0
     assert nodes[0].metadata["document_id"] == str(real_document.id)
+    assert nodes[0].metadata["filename"] == real_document.filename
+    assert nodes[0].metadata["storage_path"] == (
+        real_document.storage_path.name if real_document.storage_path else None
+    )
+    assert (
+        nodes[0].metadata["archive_serial_number"]
+        == real_document.archive_serial_number
+    )
+    assert "filename" in nodes[0].excluded_embed_metadata_keys
+    assert "filename" not in nodes[0].excluded_llm_metadata_keys
 
 
 @pytest.mark.django_db
-def test_build_document_node_excludes_metadata_from_embedding(real_document) -> None:
+def test_build_document_node_sets_ref_doc_id(real_document: Document) -> None:
+    """Every node produced by build_document_node must carry the paperless document id
+    as its ref_doc_id so that the vector store's delete(str(doc.id)) works correctly."""
+    nodes = indexing.build_document_node(real_document)
+    assert len(nodes) > 0, "Expected at least one node"
+    for node in nodes:
+        assert node.ref_doc_id == str(real_document.id), (
+            f"Expected ref_doc_id={real_document.id!r}, got {node.ref_doc_id!r}"
+        )
+
+
+@pytest.mark.django_db
+def test_build_document_node_excludes_metadata_from_embedding(
+    real_document: Document,
+) -> None:
     """Metadata keys must not be prepended to the embedding text.
 
     build_llm_index_text already encodes all metadata in the body text, so
@@ -75,8 +70,6 @@ def test_build_document_node_excludes_metadata_from_embedding(real_document) ->
     double the token count and exceed embedding models with small context
     windows (e.g. nomic-embed-text via Ollama defaults to num_ctx=2048).
     """
-    from llama_index.core.schema import MetadataMode
-
     nodes = indexing.build_document_node(real_document)
     for node in nodes:
         embed_text = node.get_content(metadata_mode=MetadataMode.EMBED)
@@ -87,7 +80,36 @@ def test_build_document_node_excludes_metadata_from_embedding(real_document) ->
 
 
 @pytest.mark.django_db
-def test_build_document_node_uses_rag_chunk_settings(real_document) -> None:
+def test_build_document_node_structured_fields_in_metadata(
+    real_document: Document,
+) -> None:
+    """Structured fields must be in node.metadata so the LLM receives them via metadata prepend."""
+    nodes = indexing.build_document_node(real_document)
+    assert len(nodes) > 0
+    for node in nodes:
+        assert "title" in node.metadata
+        assert "tags" in node.metadata
+        assert "correspondent" in node.metadata
+        assert "document_type" in node.metadata
+        assert "created" in node.metadata
+        assert "added" in node.metadata
+        assert "modified" in node.metadata
+
+
+@pytest.mark.django_db
+def test_build_document_node_excludes_document_id_from_llm_context(
+    real_document: Document,
+) -> None:
+    """document_id is an internal key and must not appear in LLM context text."""
+    nodes = indexing.build_document_node(real_document)
+    assert len(nodes) > 0
+    for node in nodes:
+        assert "document_id" in node.excluded_llm_metadata_keys
+        assert "document_id" not in node.get_content(metadata_mode=MetadataMode.LLM)
+
+
+@pytest.mark.django_db
+def test_build_document_node_uses_rag_chunk_settings(real_document: Document) -> None:
     app_config, _ = ApplicationConfiguration.objects.get_or_create()
     app_config.llm_embedding_chunk_size = 512
     app_config.save()
@@ -118,9 +140,9 @@ def test_get_rag_prompt_helper_uses_context_setting() -> None:
 
 @pytest.mark.django_db
 def test_update_llm_index(
-    temp_llm_index_dir,
-    real_document,
-    mock_embed_model,
+    temp_llm_index_dir: Path,
+    real_document: Document,
+    mock_embed_model: FakeEmbedding,
 ) -> None:
     mock_config = MagicMock()
     mock_config.llm_embedding_chunk_size = 512
@@ -138,44 +160,49 @@ def test_update_llm_index(
 
         ai_config.assert_called_once()
         build_document_node.assert_called_once_with(real_document, chunk_size=512)
-        assert any(temp_llm_index_dir.glob("*.json"))
 
 
 @pytest.mark.django_db
-def test_update_llm_index_removes_meta(
-    temp_llm_index_dir,
-    real_document,
-    mock_embed_model,
+def test_update_llm_index_rebuilds_on_model_name_change(
+    temp_llm_index_dir: Path,
+    real_document: Document,
+    mock_embed_model: FakeEmbedding,
 ) -> None:
-    # Pre-create a meta.json with incorrect data
-    (temp_llm_index_dir / "meta.json").write_text(
-        json.dumps({"embedding_model": "old", "dim": 1}),
-    )
-
+    # Build initial index with model "model-a".
     with patch("documents.models.Document.objects.all") as mock_all:
         mock_queryset = MagicMock()
         mock_queryset.exists.return_value = True
         mock_queryset.__iter__.return_value = iter([real_document])
         mock_all.return_value = mock_queryset
-        indexing.update_llm_index(rebuild=True)
+        with patch(
+            "paperless_ai.indexing.get_configured_model_name",
+            return_value="model-a",
+        ):
+            indexing.update_llm_index(rebuild=True)
 
-    meta = json.loads((temp_llm_index_dir / "meta.json").read_text())
-    from paperless.config import AIConfig
+    # Simulate config change to "model-b"; the incremental run must force a rebuild.
+    with patch("documents.models.Document.objects.all") as mock_all:
+        mock_queryset = MagicMock()
+        mock_queryset.exists.return_value = True
+        mock_queryset.__iter__.return_value = iter([real_document])
+        mock_all.return_value = mock_queryset
+        with patch(
+            "paperless_ai.indexing.get_configured_model_name",
+            return_value="model-b",
+        ):
+            indexing.update_llm_index(rebuild=False)
 
-    config = AIConfig()
-    expected_model = config.llm_embedding_model or (
-        "text-embedding-3-small"
-        if config.llm_embedding_backend == "openai-like"
-        else "sentence-transformers/all-MiniLM-L6-v2"
-    )
-    assert meta == {"embedding_model": expected_model, "dim": 384}
+    with indexing.get_vector_store() as store:
+        # Schema metadata only updates when the table is dropped and recreated, never
+        # on incremental writes -- so "model-b" here proves a full rebuild happened.
+        assert store.stored_model_name() == "model-b"
 
 
 @pytest.mark.django_db
 def test_update_llm_index_partial_update(
-    temp_llm_index_dir,
-    real_document,
-    mock_embed_model,
+    temp_llm_index_dir: Path,
+    real_document: Document,
+    mock_embed_model: FakeEmbedding,
 ) -> None:
     doc2 = Document.objects.create(
         title="Test Document 2",
@@ -210,131 +237,34 @@ def test_update_llm_index_partial_update(
         mock_queryset.__iter__.return_value = iter([updated_document, doc2, doc3])
         mock_all.return_value = mock_queryset
 
-        # assert logs "Updating LLM index with %d new nodes and removing %d old nodes."
-        with patch("paperless_ai.indexing.logger") as mock_logger:
-            indexing.update_llm_index(rebuild=False)
-            mock_logger.info.assert_called_once_with(
-                "Updating %d nodes in LLM index.",
-                2,
-            )
         indexing.update_llm_index(rebuild=False)
 
-    assert any(temp_llm_index_dir.glob("*.json"))
-
-
-def test_get_or_create_storage_context_raises_exception(
-    temp_llm_index_dir,
-    mock_embed_model,
-) -> None:
-    with pytest.raises(Exception):
-        indexing.get_or_create_storage_context(rebuild=False)
-
-
-@override_settings(
-    LLM_EMBEDDING_BACKEND="huggingface",
-)
-def test_load_or_build_index_builds_when_nodes_given(
-    temp_llm_index_dir,
-    real_document,
-    mock_embed_model,
-) -> None:
-    with (
-        patch(
-            "llama_index.core.load_index_from_storage",
-            side_effect=ValueError("Index not found"),
-        ),
-        patch(
-            "llama_index.core.VectorStoreIndex",
-            return_value=MagicMock(),
-        ) as mock_index_cls,
-        patch(
-            "paperless_ai.indexing.get_or_create_storage_context",
-            return_value=MagicMock(),
-        ) as mock_storage,
-    ):
-        mock_storage.return_value.persist_dir = temp_llm_index_dir
-        indexing.load_or_build_index(
-            nodes=[indexing.build_document_node(real_document)],
+    with indexing.get_vector_store() as store:
+        assert store.table_exists(), (
+            "Expected the vector store table to exist after incremental update"
         )
-        mock_index_cls.assert_called_once()
-
-
-def test_load_or_build_index_raises_exception_when_no_nodes(
-    temp_llm_index_dir,
-    mock_embed_model,
-) -> None:
-    with (
-        patch(
-            "llama_index.core.load_index_from_storage",
-            side_effect=ValueError("Index not found"),
-        ),
-        patch(
-            "paperless_ai.indexing.get_or_create_storage_context",
-            return_value=MagicMock(),
-        ),
-    ):
-        with pytest.raises(Exception):
-            indexing.load_or_build_index()
-
-
-@pytest.mark.django_db
-def test_load_or_build_index_succeeds_when_nodes_given(
-    temp_llm_index_dir,
-    mock_embed_model,
-) -> None:
-    with (
-        patch(
-            "llama_index.core.load_index_from_storage",
-            side_effect=ValueError("Index not found"),
-        ),
-        patch(
-            "llama_index.core.VectorStoreIndex",
-            return_value=MagicMock(),
-        ) as mock_index_cls,
-        patch(
-            "paperless_ai.indexing.get_or_create_storage_context",
-            return_value=MagicMock(),
-        ) as mock_storage,
-    ):
-        mock_storage.return_value.persist_dir = temp_llm_index_dir
-        indexing.load_or_build_index(
-            nodes=[MagicMock()],
-        )
-        mock_index_cls.assert_called_once()
 
 
 @pytest.mark.django_db
 def test_add_or_update_document_updates_existing_entry(
-    temp_llm_index_dir,
-    real_document,
-    mock_embed_model,
+    temp_llm_index_dir: Path,
+    real_document: Document,
+    mock_embed_model: FakeEmbedding,
 ) -> None:
     indexing.update_llm_index(rebuild=True)
     indexing.llm_index_add_or_update_document(real_document)
 
-    assert any(temp_llm_index_dir.glob("*.json"))
-
-
-@pytest.mark.django_db
-def test_remove_document_deletes_node_from_docstore(
-    temp_llm_index_dir,
-    real_document,
-    mock_embed_model,
-) -> None:
-    indexing.update_llm_index(rebuild=True)
-    index = indexing.load_or_build_index()
-    assert len(index.docstore.docs) == 1
-
-    indexing.llm_index_remove_document(real_document)
-    index = indexing.load_or_build_index()
-    assert len(index.docstore.docs) == 0
+    with indexing.get_vector_store() as store:
+        assert store.table_exists(), (
+            "Expected the vector store table to exist after add-or-update"
+        )
 
 
 @pytest.mark.django_db
 def test_query_after_remove_does_not_raise_key_error(
-    temp_llm_index_dir,
-    real_document,
-    mock_embed_model,
+    temp_llm_index_dir: Path,
+    real_document: Document,
+    mock_embed_model: FakeEmbedding,
 ) -> None:
     indexing.update_llm_index(rebuild=True)
 
@@ -352,8 +282,8 @@ def test_query_after_remove_does_not_raise_key_error(
 
 @pytest.mark.django_db
 def test_update_llm_index_no_documents(
-    temp_llm_index_dir,
-    mock_embed_model,
+    temp_llm_index_dir: Path,
+    mock_embed_model: FakeEmbedding,
 ) -> None:
     with patch("documents.models.Document.objects.all") as mock_all:
         mock_queryset = MagicMock()
@@ -369,6 +299,22 @@ def test_update_llm_index_no_documents(
             )
 
 
+@pytest.mark.django_db
+def test_update_no_documents_no_index_returns_early(
+    temp_llm_index_dir: Path,
+    mocker: pytest_mock.MockerFixture,
+) -> None:
+    """update with no documents and no existing index must return early."""
+    mock_qs = MagicMock()
+    mock_qs.exists.return_value = False
+    mock_qs.__iter__ = MagicMock(return_value=iter([]))
+    mocker.patch("paperless_ai.indexing.Document.objects.all", return_value=mock_qs)
+
+    result = indexing.update_llm_index(rebuild=False)
+
+    assert result == "No documents found to index."
+
+
 @pytest.mark.django_db
 def test_queue_llm_index_update_if_needed_enqueues_when_idle_or_skips_recent() -> None:
     # No existing tasks
@@ -406,20 +352,17 @@ def test_queue_llm_index_update_if_needed_enqueues_when_idle_or_skips_recent() -
     LLM_BACKEND="ollama",
 )
 def test_query_similar_documents(
-    temp_llm_index_dir,
-    real_document,
+    temp_llm_index_dir: Path,
+    real_document: Document,
 ) -> None:
     with (
-        patch("paperless_ai.indexing.get_or_create_storage_context") as mock_storage,
         patch("paperless_ai.indexing.load_or_build_index") as mock_load_or_build_index,
         patch(
-            "paperless_ai.indexing.vector_store_file_exists",
+            "paperless_ai.indexing.llm_index_exists",
         ) as mock_vector_store_exists,
         patch("llama_index.core.retrievers.VectorIndexRetriever") as mock_retriever_cls,
         patch("paperless_ai.indexing.Document.objects.filter") as mock_filter,
     ):
-        mock_storage.return_value = MagicMock()
-        mock_storage.return_value.persist_dir = temp_llm_index_dir
         mock_vector_store_exists.return_value = True
 
         mock_index = MagicMock()
@@ -453,12 +396,12 @@ def test_query_similar_documents(
 
 @pytest.mark.django_db
 def test_query_similar_documents_triggers_update_when_index_missing(
-    temp_llm_index_dir,
-    real_document,
+    temp_llm_index_dir: Path,
+    real_document: Document,
 ) -> None:
     with (
         patch(
-            "paperless_ai.indexing.vector_store_file_exists",
+            "paperless_ai.indexing.llm_index_exists",
             return_value=False,
         ),
         patch(
@@ -479,120 +422,13 @@ def test_query_similar_documents_triggers_update_when_index_missing(
     assert result == []
 
 
-@pytest.mark.django_db
-def test_query_similar_documents_normalizes_and_post_filters_allowed_ids(
-    real_document,
-) -> None:
-    real_document.owner = User.objects.create_user(username="rag-owner")
-    real_document.save()
-    private_owner = User.objects.create_user(username="rag-private-owner")
-    private_document = Document.objects.create(
-        title="Private similar document",
-        content="Similar private content that must not reach RAG.",
-        owner=private_owner,
-        added=timezone.now(),
-    )
-
-    with (
-        patch(
-            "paperless_ai.indexing.vector_store_file_exists",
-            return_value=True,
-        ),
-        patch("paperless_ai.indexing.load_or_build_index") as mock_load_or_build_index,
-        patch("llama_index.core.retrievers.VectorIndexRetriever") as mock_retriever_cls,
-    ):
-        allowed_node = MagicMock()
-        allowed_node.node_id = "allowed-node"
-        allowed_node.metadata = {"document_id": str(real_document.pk)}
-        private_node = MagicMock()
-        private_node.node_id = "private-node"
-        private_node.metadata = {"document_id": str(private_document.pk)}
-
-        mock_index = MagicMock()
-        mock_index.docstore.docs.values.return_value = [allowed_node, private_node]
-        mock_load_or_build_index.return_value = mock_index
-
-        mock_retriever = MagicMock()
-        mock_retriever.retrieve.return_value = [private_node, allowed_node]
-        mock_retriever_cls.return_value = mock_retriever
-
-        result = indexing.query_similar_documents(
-            real_document,
-            top_k=2,
-            document_ids=[real_document.pk],
-        )
-
-    mock_retriever_cls.assert_called_once_with(
-        index=mock_index,
-        similarity_top_k=2,
-        doc_ids=["allowed-node"],
-    )
-    assert result == [real_document]
-    assert private_document not in result
-
-
-class TestUpdateLlmIndexStaleNodes:
-    """Tests that update_llm_index removes ALL nodes for a multi-chunk document."""
-
-    @pytest.mark.django_db
-    def test_incremental_update_removes_all_old_nodes_for_multi_chunk_document(
-        self,
-        temp_llm_index_dir,
-        mock_embed_model: MagicMock,
-    ) -> None:
-        """Ghost nodes from all chunks of a modified document must be removed.
-
-        When a document is split into multiple chunks (chunk_size=1024), the
-        incremental update path must delete every old node, not just the last
-        one captured by a dict comprehension keyed on document_id.
-        """
-        # Content long enough to produce at least two chunks at chunk_size=1024.
-        # Generate many paragraphs so the token count comfortably exceeds 1024.
-        fake = Faker()
-        long_content = "\n\n".join(fake.paragraph(nb_sentences=20) for _ in range(20))
-        doc = DocumentFactory(content=long_content)
-
-        # Build the initial index (rebuild=True) so it has multiple nodes
-        indexing.update_llm_index(rebuild=True)
-
-        # Verify the initial index has more than one node for this document
-        initial_index = indexing.load_or_build_index()
-        initial_node_ids = [
-            nid
-            for nid, node in initial_index.docstore.docs.items()
-            if node.metadata.get("document_id") == str(doc.id)
-        ]
-        assert len(initial_node_ids) > 1, (
-            f"Expected multiple chunks but got {len(initial_node_ids)}; "
-            "increase long_content length"
-        )
-
-        # Simulate a modification so the incremental path treats it as changed.
-        # Use queryset.update() to bypass auto_now and actually change the DB value.
-        new_modified = timezone.now()
-        Document.objects.filter(pk=doc.pk).update(modified=new_modified)
-
-        # Run incremental update (rebuild=False) with the modified document
-        indexing.update_llm_index(rebuild=False)
-
-        # Reload the persisted index and check that no OLD node ids remain
-        updated_index = indexing.load_or_build_index()
-        remaining_old_node_ids = [
-            nid for nid in initial_node_ids if nid in updated_index.docstore.docs
-        ]
-        assert remaining_old_node_ids == [], (
-            f"Ghost nodes still present after incremental update: "
-            f"{remaining_old_node_ids}"
-        )
-
-
 @pytest.mark.django_db
 def test_query_similar_documents_empty_allow_list_fails_closed(
-    real_document,
+    real_document: Document,
 ) -> None:
     with (
         patch(
-            "paperless_ai.indexing.vector_store_file_exists",
+            "paperless_ai.indexing.llm_index_exists",
             return_value=True,
         ) as mock_vector_store_exists,
         patch("paperless_ai.indexing.load_or_build_index") as mock_load_or_build_index,
@@ -610,27 +446,25 @@ def test_query_similar_documents_empty_allow_list_fails_closed(
 
 
 class TestUpdateLlmIndexEmptyDocumentSet:
-    """update_llm_index must persist an empty index when all documents are deleted.
+    """update_llm_index must clear the vector store table when all documents are deleted.
 
-    Without this, the stale on-disk FAISS vectors are never cleared and
-    subsequent similarity searches return phantom hits for document IDs that
-    no longer exist in the DB.
+    Without this, the stale vectors are never cleared and subsequent similarity
+    searches return phantom hits for document IDs that no longer exist in the DB.
     """
 
     @pytest.mark.django_db
     def test_rebuild_clears_stale_index_when_no_documents_exist(
         self,
         temp_llm_index_dir: Path,
-        mock_embed_model: MagicMock,
+        mock_embed_model: FakeEmbedding,
     ) -> None:
-        """After deleting all documents, rebuild=True must persist an empty index.
+        """After deleting all documents, rebuild=True must produce a table with zero rows.
 
         Steps:
         1. Build an index with one document so the on-disk state is non-empty.
         2. Delete all documents from the DB.
         3. Call update_llm_index(rebuild=True).
-        4. Reload the index from disk.
-        5. Assert the reloaded index has zero nodes (no phantom vectors).
+        4. Open the LanceDB table directly and assert zero rows.
         """
         # Step 1: create a document and build a non-empty index
         Document.objects.create(
@@ -640,27 +474,26 @@ class TestUpdateLlmIndexEmptyDocumentSet:
         )
         indexing.update_llm_index(rebuild=True)
 
-        initial_index = indexing.load_or_build_index()
-        assert len(initial_index.docstore.docs) > 0, (
-            "Precondition failed: expected at least one node before deletion"
-        )
+        with indexing.get_vector_store() as store:
+            assert store.table_exists(), (
+                "Precondition failed: expected the vector store table to exist "
+                "before deletion"
+            )
 
         # Step 2: delete all documents
         Document.objects.all().delete()
         assert not Document.objects.exists()
 
-        # Step 3: rebuild with no documents
+        # Step 3: rebuild with no documents — drop_table is called so the table
+        # is removed (no rows to re-insert, so it stays absent).
         indexing.update_llm_index(rebuild=True)
 
-        # Step 4: reload the persisted index from disk
-        reloaded_index = indexing.load_or_build_index()
-
-        # Step 5: phantom vectors must be gone
-        assert len(reloaded_index.docstore.docs) == 0, (
-            f"Expected 0 nodes after clearing all documents, "
-            f"but found {len(reloaded_index.docstore.docs)}: "
-            f"{list(reloaded_index.docstore.docs.keys())}"
-        )
+        # Step 4: the table must be absent (no rows) — phantom vectors gone
+        with indexing.get_vector_store() as store2:
+            assert not store2.table_exists(), (
+                "Expected the vector store table to be absent after rebuilding "
+                "with no documents"
+            )
 
 
 class TestDocumentUpdatedSignalTriggersLlmReindex:
@@ -709,10 +542,14 @@ class TestLlmIndexAddOrUpdateDocumentEmptyContent:
     def test_returns_without_error_when_build_document_node_returns_empty(
         self,
         temp_llm_index_dir: Path,
+        mock_embed_model: MagicMock,
         mocker: pytest_mock.MockerFixture,
     ) -> None:
-        """When build_document_node returns [], the function must return without error
-        and must not call load_or_build_index at all."""
+        """When build_document_node returns [], the function must return without error.
+
+        The store's upsert_document treats an empty node list as a removal (no-op
+        delete), so load_or_build_index must not be called.
+        """
         mocker.patch(
             "paperless_ai.indexing.build_document_node",
             return_value=[],
@@ -720,6 +557,7 @@ class TestLlmIndexAddOrUpdateDocumentEmptyContent:
         mock_load = mocker.patch("paperless_ai.indexing.load_or_build_index")
 
         doc = MagicMock(spec=Document)
+        doc.id = 42
         # Must not raise
         indexing.llm_index_add_or_update_document(doc)
 
@@ -727,172 +565,220 @@ class TestLlmIndexAddOrUpdateDocumentEmptyContent:
 
 
 @pytest.mark.django_db
-class TestLlmIndexLocking:
-    """The FAISS index mutation functions must acquire the index lock before touching the index.
+def test_llm_index_compact_uses_force(
+    temp_llm_index_dir: Path,
+    mocker: pytest_mock.MockerFixture,
+) -> None:
+    """compact must use force=True to rebuild the table and reclaim space immediately."""
+    mock_store = mocker.MagicMock()
+    mocker.patch(
+        "paperless_ai.indexing.write_store",
+        return_value=mocker.MagicMock(
+            __enter__=mocker.MagicMock(return_value=mock_store),
+            __exit__=mocker.MagicMock(return_value=False),
+        ),
+    )
 
-    Without locking, two concurrent Celery workers can each load the same
-    on-disk index, make independent modifications, and the last writer silently
-    overwrites the first's changes.
+    indexing.llm_index_compact()
+
+    mock_store.compact.assert_called_once_with(force=True)
+
+
+@pytest.mark.django_db
+class TestLlmIndexLocking:
+    """Index mutation functions must go through write_store(), which holds the lock.
+
+    Without locking, two concurrent Celery workers can open the same store,
+    make independent modifications, and trigger CommitConflictError.
     """
 
-    def test_add_or_update_document_acquires_lock(
+    def test_add_or_update_document_uses_write_store(
         self,
         temp_llm_index_dir: Path,
+        mock_embed_model: FakeEmbedding,
         mocker: pytest_mock.MockerFixture,
     ) -> None:
-        """llm_index_add_or_update_document must enter the file lock before touching the index."""
-        call_order: list[str] = []
-
-        mock_lock_instance = MagicMock()
-        mock_lock_instance.__enter__ = MagicMock(
-            side_effect=lambda *_: call_order.append("lock_acquired"),
-        )
-        mock_lock_instance.__exit__ = MagicMock(return_value=False)
-
-        mock_file_lock_cls = mocker.patch(
-            "paperless_ai.indexing.FileLock",
-            return_value=mock_lock_instance,
-        )
-
-        mock_load = mocker.patch(
-            "paperless_ai.indexing.load_or_build_index",
-            side_effect=lambda *_a, **_kw: (
-                call_order.append("index_loaded") or MagicMock()
+        mock_store = MagicMock()
+        mocker.patch(
+            "paperless_ai.indexing.write_store",
+            return_value=mocker.MagicMock(
+                __enter__=mocker.MagicMock(return_value=mock_store),
+                __exit__=mocker.MagicMock(return_value=False),
             ),
         )
+        mock_node = MagicMock()
+        mock_node.get_content.return_value = "fake node text"
         mocker.patch(
             "paperless_ai.indexing.build_document_node",
-            return_value=[MagicMock()],
+            return_value=[mock_node],
         )
-        mocker.patch("paperless_ai.indexing.remove_document_docstore_nodes")
 
         doc = MagicMock(spec=Document)
+        doc.id = 1
         indexing.llm_index_add_or_update_document(doc)
 
-        mock_file_lock_cls.assert_called_once()
-        mock_lock_instance.__enter__.assert_called_once()
-        mock_load.assert_called_once()
-        assert call_order.index("lock_acquired") < call_order.index("index_loaded"), (
-            "Lock must be acquired before the index is loaded"
-        )
+        mock_store.upsert_document.assert_called_once()
 
-    def test_remove_document_acquires_lock(
+    def test_remove_document_uses_write_store(
         self,
         temp_llm_index_dir: Path,
         mocker: pytest_mock.MockerFixture,
     ) -> None:
-        """llm_index_remove_document must enter the file lock before loading the index."""
-        call_order: list[str] = []
-
-        mock_lock_instance = MagicMock()
-        mock_lock_instance.__enter__ = MagicMock(
-            side_effect=lambda *_: call_order.append("lock_acquired"),
-        )
-        mock_lock_instance.__exit__ = MagicMock(return_value=False)
-
-        mock_file_lock_cls = mocker.patch(
-            "paperless_ai.indexing.FileLock",
-            return_value=mock_lock_instance,
-        )
-
-        mock_load = mocker.patch(
-            "paperless_ai.indexing.load_or_build_index",
-            side_effect=lambda *_a, **_kw: (
-                call_order.append("index_loaded") or MagicMock()
+        mock_store = MagicMock()
+        mocker.patch(
+            "paperless_ai.indexing.write_store",
+            return_value=mocker.MagicMock(
+                __enter__=mocker.MagicMock(return_value=mock_store),
+                __exit__=mocker.MagicMock(return_value=False),
             ),
         )
-        mocker.patch("paperless_ai.indexing.remove_document_docstore_nodes")
 
         doc = MagicMock(spec=Document)
+        doc.id = 1
         indexing.llm_index_remove_document(doc)
 
-        mock_file_lock_cls.assert_called_once()
-        mock_lock_instance.__enter__.assert_called_once()
-        mock_load.assert_called_once()
-        assert call_order.index("lock_acquired") < call_order.index("index_loaded"), (
-            "Lock must be acquired before the index is loaded"
-        )
+        mock_store.delete.assert_called_once_with("1")
 
-    def test_update_llm_index_rebuild_acquires_lock(
+    def test_update_llm_index_rebuild_uses_write_store(
         self,
         temp_llm_index_dir: Path,
-        mock_embed_model: MagicMock,
+        mock_embed_model: FakeEmbedding,
         mocker: pytest_mock.MockerFixture,
     ) -> None:
-        """update_llm_index must enter the file lock during the rebuild/persist cycle."""
-        mock_lock_instance = MagicMock()
-        mock_lock_instance.__enter__ = MagicMock(return_value=None)
-        mock_lock_instance.__exit__ = MagicMock(return_value=False)
-
-        mock_file_lock_cls = mocker.patch(
-            "paperless_ai.indexing.FileLock",
-            return_value=mock_lock_instance,
+        mock_store = MagicMock()
+        mocker.patch(
+            "paperless_ai.indexing.write_store",
+            return_value=mocker.MagicMock(
+                __enter__=mocker.MagicMock(return_value=mock_store),
+                __exit__=mocker.MagicMock(return_value=False),
+            ),
         )
-
-        # exists=True so the code reaches the lock; iterate over an empty
-        # queryset so VectorStoreIndex is called with no nodes (still exercises
-        # the lock path without needing heavy FAISS fixture data)
         mock_qs = MagicMock()
         mock_qs.exists.return_value = True
         mock_qs.__iter__ = MagicMock(return_value=iter([]))
         mocker.patch("paperless_ai.indexing.Document.objects.all", return_value=mock_qs)
-        mocker.patch(
-            "paperless_ai.indexing.get_or_create_storage_context",
-            return_value=MagicMock(),
-        )
 
         indexing.update_llm_index(rebuild=True)
 
-        mock_file_lock_cls.assert_called_once()
-        mock_lock_instance.__enter__.assert_called_once()
+        mock_store.drop_table.assert_called_once()
 
-    def test_query_similar_documents_acquires_lock(
+
+@pytest.mark.django_db
+@pytest.mark.django_db
+class TestVectorStoreIndexing:
+    def test_get_vector_store_roundtrip(
         self,
         temp_llm_index_dir: Path,
+        mock_embed_model: FakeEmbedding,
+    ) -> None:
+        with indexing.get_vector_store() as store:
+            assert isinstance(store, PaperlessSqliteVecVectorStore)
+
+    def test_add_then_remove_document(
+        self,
+        temp_llm_index_dir: Path,
+        mock_embed_model: FakeEmbedding,
+        real_document: Document,
+    ) -> None:
+        indexing.llm_index_add_or_update_document(real_document)
+        with indexing.get_vector_store() as store:
+            assert store.table_exists()
+            count_sql = "SELECT count(*) FROM documents"
+            assert store.client.execute(count_sql).fetchone()[0] >= 1
+
+            indexing.llm_index_remove_document(real_document)
+            assert store.client.execute(count_sql).fetchone()[0] == 0
+
+    def test_update_shrinks_chunks_without_orphans(
+        self,
+        temp_llm_index_dir: Path,
+        mock_embed_model: FakeEmbedding,
+        real_document: Document,
+    ) -> None:
+        real_document.content = "word " * 4000  # many chunks
+        real_document.save()
+        indexing.llm_index_add_or_update_document(real_document)
+        count_sql = "SELECT count(*) FROM documents"
+        with indexing.get_vector_store() as store:
+            big = store.client.execute(count_sql).fetchone()[0]
+
+            real_document.content = "short"  # one chunk
+            real_document.save()
+            indexing.llm_index_add_or_update_document(real_document)
+
+            rows = store.client.execute(count_sql).fetchone()[0]
+            assert rows < big
+            assert rows >= 1
+
+
+@pytest.mark.django_db
+class TestQuerySimilarDocuments:
+    def test_query_similar_documents_respects_allowed_ids(
+        self,
+        temp_llm_index_dir: Path,
+        mock_embed_model: FakeEmbedding,
+    ) -> None:
+        a = DocumentFactory.create(content="alpha shared content here")
+        b = DocumentFactory.create(content="beta shared content here")
+        c = DocumentFactory.create(content="gamma shared content here")
+        for doc in (a, b, c):
+            indexing.llm_index_add_or_update_document(doc)
+
+        results = indexing.query_similar_documents(a, document_ids=[b.id])
+
+        assert all(doc.id == b.id for doc in results)
+
+
+class TestRetrieveSimilarNodes:
+    @pytest.mark.django_db
+    def test_returns_raw_nodes_from_retriever(
+        self,
+        temp_llm_index_dir: Path,
+        real_document: Document,
         mocker: pytest_mock.MockerFixture,
     ) -> None:
-        """query_similar_documents must enter the file lock before loading the index."""
-        call_order: list[str] = []
-
-        mock_lock_instance = MagicMock()
-        mock_lock_instance.__enter__ = MagicMock(
-            side_effect=lambda *_: call_order.append("lock_acquired"),
-        )
-        mock_lock_instance.__exit__ = MagicMock(return_value=False)
-
-        mock_file_lock_cls = mocker.patch(
-            "paperless_ai.indexing.FileLock",
-            return_value=mock_lock_instance,
-        )
-
-        mocker.patch(
-            "paperless_ai.indexing.vector_store_file_exists",
-            return_value=True,
-        )
-
-        mock_index = MagicMock()
-        mock_index.docstore.docs = {}
-
-        mocker.patch(
-            "paperless_ai.indexing.load_or_build_index",
-            side_effect=lambda *_a, **_kw: (
-                call_order.append("index_loaded") or mock_index
-            ),
-        )
-
-        mock_retriever = MagicMock()
-        mock_retriever.retrieve.return_value = []
+        mocker.patch("paperless_ai.indexing.llm_index_exists", return_value=True)
+        mocker.patch("paperless_ai.indexing.load_or_build_index")
+        node1 = SimpleNamespace(metadata={"document_id": "1"})
+        node2 = SimpleNamespace(metadata={"document_id": "2"})
+        retriever = mocker.MagicMock()
+        retriever.retrieve.return_value = [node1, node2]
         mocker.patch(
             "llama_index.core.retrievers.VectorIndexRetriever",
-            return_value=mock_retriever,
+            return_value=retriever,
         )
 
-        mocker.patch("paperless_ai.indexing.truncate_content", return_value="")
+        result = indexing.retrieve_similar_nodes(real_document, top_k=3)
 
-        indexing.query_similar_documents(MagicMock(spec=Document))
+        assert result == [node1, node2]
 
-        mock_file_lock_cls.assert_called()
-        mock_lock_instance.__enter__.assert_called()
-        assert call_order.index("lock_acquired") < call_order.index("index_loaded"), (
-            "Lock must be acquired before the index is loaded"
+    @pytest.mark.django_db
+    def test_empty_allow_list_fails_closed(
+        self,
+        real_document: Document,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        load = mocker.patch("paperless_ai.indexing.load_or_build_index")
+
+        result = indexing.retrieve_similar_nodes(real_document, document_ids=[])
+
+        assert result == []
+        load.assert_not_called()
+
+    @pytest.mark.django_db
+    def test_queues_update_when_index_missing(
+        self,
+        temp_llm_index_dir: Path,
+        real_document: Document,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch("paperless_ai.indexing.llm_index_exists", return_value=False)
+        queue = mocker.patch("paperless_ai.indexing.queue_llm_index_update_if_needed")
+
+        result = indexing.retrieve_similar_nodes(real_document, top_k=2)
+
+        assert result == []
+        queue.assert_called_once_with(
+            rebuild=False,
+            reason="LLM index not found for similarity query.",
         )

src/paperless_ai/tests/test_chat.py

@@ -3,19 +3,20 @@ from unittest.mock import MagicMock
 from unittest.mock import patch
 
 import pytest
+from llama_index.core import settings as llama_settings
+from llama_index.core.embeddings.mock_embed_model import MockEmbedding
 from llama_index.core.schema import TextNode
 
+from documents.tests.factories import DocumentFactory
+from paperless_ai import chat
+from paperless_ai import indexing
 from paperless_ai.chat import CHAT_ERROR_MESSAGE
 from paperless_ai.chat import CHAT_METADATA_DELIMITER
-from paperless_ai.chat import _get_document_filtered_retriever
 from paperless_ai.chat import stream_chat_with_documents
 
 
 @pytest.fixture(autouse=True)
 def patch_embed_model():
-    from llama_index.core import settings as llama_settings
-    from llama_index.core.embeddings.mock_embed_model import MockEmbedding
-
     # Use a real BaseEmbedding subclass to satisfy llama-index 0.14 validation
     llama_settings.Settings.embed_model = MockEmbedding(embed_dim=1536)
     yield
@@ -58,91 +59,6 @@ def assert_chat_output(
     }
 
 
-def add_vector_query_results(mock_index, nodes: list[TextNode]) -> None:
-    mock_index.index_struct.nodes_dict = {
-        str(vector_id): node.node_id for vector_id, node in enumerate(nodes)
-    }
-    mock_index.docstore.docs.get.side_effect = {
-        node.node_id: node for node in nodes
-    }.get
-    mock_index.vector_store._faiss_index.ntotal = len(nodes)
-    mock_index.vector_store.query.return_value = MagicMock(
-        ids=list(mock_index.index_struct.nodes_dict),
-        similarities=[0.1] * len(nodes),
-    )
-    mock_index._embed_model.get_agg_embedding_from_queries.return_value = [0.1] * 1536
-
-
-def test_document_filtered_retriever_expands_filters_and_caches() -> None:
-    allowed_node1 = TextNode(
-        text="Allowed content 1.",
-        metadata={"document_id": "1", "title": "Allowed 1"},
-    )
-    allowed_node2 = TextNode(
-        text="Allowed content 2.",
-        metadata={"document_id": "2", "title": "Allowed 2"},
-    )
-    foreign_node = TextNode(
-        text="Foreign content.",
-        metadata={"document_id": "3", "title": "Foreign"},
-    )
-    missing_node = TextNode(
-        text="Missing content.",
-        metadata={"document_id": "1", "title": "Missing"},
-    )
-
-    mock_index = MagicMock()
-    mock_index.index_struct.nodes_dict = {
-        "0": foreign_node.node_id,
-        "1": missing_node.node_id,
-        "2": allowed_node1.node_id,
-        "3": allowed_node2.node_id,
-    }
-    mock_index.docstore.docs.get.side_effect = {
-        allowed_node1.node_id: allowed_node1,
-        allowed_node2.node_id: allowed_node2,
-        foreign_node.node_id: foreign_node,
-    }.get
-    mock_index.vector_store._faiss_index.ntotal = 4
-    mock_index.vector_store.query.side_effect = [
-        MagicMock(ids=["0", "2"], similarities=[0.9, 0.8]),
-        MagicMock(ids=["0", "1", "3"], similarities=[0.9, 0.7, 0.6]),
-    ]
-    mock_index._embed_model.get_agg_embedding_from_queries.return_value = [0.1] * 1536
-
-    retriever = _get_document_filtered_retriever(
-        mock_index,
-        {"1", "2"},
-        similarity_top_k=2,
-    )
-
-    nodes = retriever.retrieve("question")
-    cached_nodes = retriever.retrieve("question")
-
-    assert [node.node.node_id for node in nodes] == [
-        allowed_node1.node_id,
-        allowed_node2.node_id,
-    ]
-    assert cached_nodes == nodes
-    assert mock_index.vector_store.query.call_count == 2
-    assert mock_index._embed_model.get_agg_embedding_from_queries.call_count == 1
-
-
-def test_document_filtered_retriever_handles_empty_faiss_index() -> None:
-    mock_index = MagicMock()
-    mock_index.vector_store._faiss_index.ntotal = 0
-    mock_index._embed_model.get_agg_embedding_from_queries.return_value = [0.1] * 1536
-
-    retriever = _get_document_filtered_retriever(
-        mock_index,
-        {"1"},
-        similarity_top_k=2,
-    )
-
-    assert retriever.retrieve("question") == []
-    mock_index.vector_store.query.assert_not_called()
-
-
 @pytest.mark.django_db
 def test_stream_chat_with_one_document_retrieval(
     mock_document,
@@ -164,17 +80,31 @@ def test_stream_chat_with_one_document_retrieval(
             metadata={"document_id": str(mock_document.pk), "title": "Test Document"},
         )
         mock_index = MagicMock()
-        mock_index.docstore.docs.values.return_value = [mock_node]
-        add_vector_query_results(mock_index, [mock_node])
+        # Simulate get_nodes returning nodes (content exists)
+        mock_index.vector_store.get_nodes.return_value = [mock_node]
         mock_load_index.return_value = mock_index
 
+        mock_retriever_instance = MagicMock()
+        mock_retriever_instance.retrieve.return_value = [
+            MagicMock(
+                metadata={
+                    "document_id": str(mock_document.pk),
+                    "title": "Test Document",
+                },
+            ),
+        ]
+
         mock_response_stream = MagicMock()
         mock_response_stream.response_gen = iter(["chunk1", "chunk2"])
         mock_query_engine = MagicMock()
         mock_query_engine_cls.return_value = mock_query_engine
         mock_query_engine.query.return_value = mock_response_stream
 
-        output = list(stream_chat_with_documents("What is this?", [mock_document]))
+        with patch(
+            "llama_index.core.retrievers.VectorIndexRetriever",
+            return_value=mock_retriever_instance,
+        ):
+            output = list(stream_chat_with_documents("What is this?", [mock_document]))
 
         mock_query_engine.query.assert_called_once_with("What is this?")
         patch_embed_nodes.assert_not_called()
@@ -196,12 +126,10 @@ def test_stream_chat_with_multiple_documents_retrieval(patch_embed_nodes) -> Non
             "llama_index.core.query_engine.RetrieverQueryEngine.from_args",
         ) as mock_query_engine_cls,
     ):
-        # Mock AIClient and LLM
         mock_client = MagicMock()
         mock_client_cls.return_value = mock_client
         mock_client.llm = MagicMock()
 
-        # Create two real TextNodes
         mock_node1 = TextNode(
             text="Content for doc 1.",
             metadata={"document_id": "1", "title": "Document 1"},
@@ -210,41 +138,32 @@ def test_stream_chat_with_multiple_documents_retrieval(patch_embed_nodes) -> Non
             text="Content for doc 2.",
             metadata={"document_id": "2", "title": "Document 2"},
         )
-        mock_duplicate_node = TextNode(
-            text="More content for doc 1.",
-            metadata={"document_id": "1", "title": "Document 1 Duplicate"},
-        )
-        mock_foreign_node = TextNode(
-            text="Content for doc 3.",
-            metadata={"document_id": "3", "title": "Document 3"},
-        )
         mock_index = MagicMock()
-        mock_index.docstore.docs.values.return_value = [
-            mock_node1,
-            mock_node2,
-            mock_duplicate_node,
-            mock_foreign_node,
-        ]
-        add_vector_query_results(
-            mock_index,
-            [mock_node1, mock_duplicate_node, mock_node2, mock_foreign_node],
-        )
+        # Simulate get_nodes returning nodes (content exists)
+        mock_index.vector_store.get_nodes.return_value = [mock_node1, mock_node2]
         mock_load_index.return_value = mock_index
 
-        # Mock response stream
+        mock_retriever_instance = MagicMock()
+        mock_retriever_instance.retrieve.return_value = [
+            MagicMock(metadata={"document_id": "1", "title": "Document 1"}),
+            MagicMock(metadata={"document_id": "2", "title": "Document 2"}),
+        ]
+
         mock_response_stream = MagicMock()
         mock_response_stream.response_gen = iter(["chunk1", "chunk2"])
 
-        # Mock RetrieverQueryEngine
         mock_query_engine = MagicMock()
         mock_query_engine_cls.return_value = mock_query_engine
         mock_query_engine.query.return_value = mock_response_stream
 
-        # Fake documents
         doc1 = MagicMock(pk=1, title="Document 1", filename="doc1.pdf")
         doc2 = MagicMock(pk=2, title="Document 2", filename="doc2.pdf")
 
-        output = list(stream_chat_with_documents("What's up?", [doc1, doc2]))
+        with patch(
+            "llama_index.core.retrievers.VectorIndexRetriever",
+            return_value=mock_retriever_instance,
+        ):
+            output = list(stream_chat_with_documents("What's up?", [doc1, doc2]))
 
         mock_query_engine.query.assert_called_once_with("What's up?")
         patch_embed_nodes.assert_not_called()
@@ -258,8 +177,16 @@ def test_stream_chat_with_multiple_documents_retrieval(patch_embed_nodes) -> Non
         )
 
 
+def test_stream_chat_empty_document_list() -> None:
+    with patch("paperless_ai.chat.load_or_build_index") as mock_load_index:
+        output = list(stream_chat_with_documents("Any info?", []))
+        mock_load_index.assert_not_called()
+        assert output == ["Sorry, I couldn't find any content to answer your question."]
+
+
 def test_stream_chat_no_matching_nodes() -> None:
     with (
+        patch("paperless_ai.chat.AIConfig"),
         patch("paperless_ai.chat.AIClient") as mock_client_cls,
         patch("paperless_ai.chat.load_or_build_index") as mock_load_index,
     ):
@@ -268,8 +195,8 @@ def test_stream_chat_no_matching_nodes() -> None:
         mock_client.llm = MagicMock()
 
         mock_index = MagicMock()
-        # No matching nodes
-        mock_index.docstore.docs.values.return_value = []
+        # No matching nodes in the store
+        mock_index.vector_store.get_nodes.return_value = []
         mock_load_index.return_value = mock_index
 
         output = list(stream_chat_with_documents("Any info?", [MagicMock(pk=1)]))
@@ -279,30 +206,83 @@ def test_stream_chat_no_matching_nodes() -> None:
 
 def test_stream_chat_unexpected_failure_returns_generic_error(caplog) -> None:
     with (
+        patch("paperless_ai.chat.AIConfig"),
         patch("paperless_ai.chat.AIClient") as mock_client_cls,
         patch("paperless_ai.chat.load_or_build_index") as mock_load_index,
-        patch(
-            "paperless_ai.chat._get_document_filtered_retriever",
-        ) as mock_get_retriever,
     ):
         mock_client = MagicMock()
         mock_client_cls.return_value = mock_client
         mock_client.llm = MagicMock()
 
-        mock_node = TextNode(
-            text="This is node content.",
-            metadata={"document_id": "1", "title": "Test Document"},
-        )
         mock_index = MagicMock()
-        mock_index.docstore.docs.values.return_value = [mock_node]
+        # Nodes found so we get past the pre-check
+        mock_index.vector_store.get_nodes.return_value = [MagicMock()]
         mock_load_index.return_value = mock_index
 
-        mock_retriever = MagicMock()
-        mock_retriever.retrieve.side_effect = RuntimeError("private provider detail")
-        mock_get_retriever.return_value = mock_retriever
+        with patch(
+            "llama_index.core.retrievers.VectorIndexRetriever",
+        ) as mock_retriever_cls:
+            mock_retriever = MagicMock()
+            mock_retriever.retrieve.side_effect = RuntimeError(
+                "private provider detail",
+            )
+            mock_retriever_cls.return_value = mock_retriever
 
-        output = list(stream_chat_with_documents("Any info?", [MagicMock(pk=1)]))
+            output = list(stream_chat_with_documents("Any info?", [MagicMock(pk=1)]))
 
         assert output == [CHAT_ERROR_MESSAGE]
         assert "Failed to stream document chat response" in caplog.text
         assert "private provider detail" in caplog.text
+
+
+@pytest.mark.django_db
+class TestStreamChatRetrieval:
+    def test_no_nodes_yields_no_content_message(
+        self,
+        temp_llm_index_dir,
+        mock_embed_model,
+    ) -> None:
+        doc = DocumentFactory.create(content="hello world")
+        # Nothing indexed for this document yet.
+        out = list(chat.stream_chat_with_documents("question?", [doc]))
+        assert chat.CHAT_NO_CONTENT_MESSAGE in out
+
+    def test_chat_filter_contains_only_requested_document_ids(
+        self,
+        temp_llm_index_dir,
+        mock_embed_model,
+        mocker,
+    ) -> None:
+        """The MetadataFilter passed to the retriever must be scoped to the
+        requested documents only — content from other indexed documents must
+        not be surfaced.
+        """
+        included = DocumentFactory.create(content="included document content")
+        excluded = DocumentFactory.create(content="excluded document content")
+        indexing.llm_index_add_or_update_document(included)
+        indexing.llm_index_add_or_update_document(excluded)
+
+        # VectorIndexRetriever is imported inside _stream_chat_with_documents;
+        # patch it at the llama_index source so the lazy import picks it up.
+        captured_filters = []
+        mock_retriever = mocker.MagicMock()
+        mock_retriever.retrieve.return_value = []
+
+        def capture_retriever(*args, **kwargs):
+            captured_filters.append(kwargs.get("filters"))
+            return mock_retriever
+
+        mocker.patch("paperless_ai.chat.AIClient")
+        mocker.patch(
+            "llama_index.core.retrievers.VectorIndexRetriever",
+            side_effect=capture_retriever,
+        )
+
+        list(chat.stream_chat_with_documents("question?", [included]))
+
+        assert captured_filters, "VectorIndexRetriever was never constructed"
+        filt = captured_filters[0]
+        assert filt is not None, "Retriever must receive a MetadataFilters"
+        filter_values = filt.filters[0].value
+        assert str(included.pk) in filter_values
+        assert str(excluded.pk) not in filter_values

src/paperless_ai/tests/test_embedding.py

@@ -1,4 +1,3 @@
-import json
 from unittest.mock import ANY
 from unittest.mock import MagicMock
 from unittest.mock import patch
@@ -10,7 +9,7 @@ from documents.models import Document
 from paperless.models import LLMEmbeddingBackend
 from paperless_ai.embedding import _normalize_llm_index_text
 from paperless_ai.embedding import build_llm_index_text
-from paperless_ai.embedding import get_embedding_dim
+from paperless_ai.embedding import get_configured_model_name
 from paperless_ai.embedding import get_embedding_model
 
 
@@ -67,7 +66,7 @@ def test_get_embedding_model_openai(mock_ai_config):
     with patch(
         "llama_index.embeddings.openai_like.OpenAILikeEmbedding",
     ) as MockOpenAIEmbedding:
-        model = get_embedding_model()
+        model = get_embedding_model(mock_ai_config.return_value)
         MockOpenAIEmbedding.assert_called_once_with(
             model_name="text-embedding-3-small",
             api_key="test_api_key",
@@ -88,7 +87,7 @@ def test_get_embedding_model_openai_prefers_embedding_endpoint(mock_ai_config):
     with patch(
         "llama_index.embeddings.openai_like.OpenAILikeEmbedding",
     ) as MockOpenAIEmbedding:
-        model = get_embedding_model()
+        model = get_embedding_model(mock_ai_config.return_value)
         MockOpenAIEmbedding.assert_called_once_with(
             model_name="text-embedding-3-small",
             api_key="test_api_key",
@@ -109,7 +108,7 @@ def test_get_embedding_model_openai_blocks_internal_endpoint_when_disallowed(
     mock_ai_config.return_value.llm_allow_internal_endpoints = False
 
     with pytest.raises(ValueError, match="non-public address"):
-        get_embedding_model()
+        get_embedding_model(mock_ai_config.return_value)
 
 
 def test_get_embedding_model_huggingface(mock_ai_config):
@@ -121,7 +120,7 @@ def test_get_embedding_model_huggingface(mock_ai_config):
     with patch(
         "llama_index.embeddings.huggingface.HuggingFaceEmbedding",
     ) as MockHuggingFaceEmbedding:
-        model = get_embedding_model()
+        model = get_embedding_model(mock_ai_config.return_value)
         MockHuggingFaceEmbedding.assert_called_once_with(
             model_name="sentence-transformers/all-MiniLM-L6-v2",
             cache_folder=str(settings.DATA_DIR / "hf_cache"),
@@ -137,7 +136,7 @@ def test_get_embedding_model_ollama(mock_ai_config):
     with patch(
         "llama_index.embeddings.ollama.OllamaEmbedding",
     ) as MockOllamaEmbedding:
-        model = get_embedding_model()
+        model = get_embedding_model(mock_ai_config.return_value)
         MockOllamaEmbedding.assert_called_once_with(
             model_name="embeddinggemma",
             base_url="http://test-url",
@@ -155,7 +154,7 @@ def test_get_embedding_model_ollama_prefers_embedding_endpoint(mock_ai_config):
     with patch(
         "llama_index.embeddings.ollama.OllamaEmbedding",
     ) as MockOllamaEmbedding:
-        model = get_embedding_model()
+        model = get_embedding_model(mock_ai_config.return_value)
         MockOllamaEmbedding.assert_called_once_with(
             model_name="embeddinggemma",
             base_url="http://embedding-url",
@@ -173,7 +172,7 @@ def test_get_embedding_model_ollama_blocks_internal_endpoint_when_disallowed(
     mock_ai_config.return_value.llm_allow_internal_endpoints = False
 
     with pytest.raises(ValueError, match="non-public address"):
-        get_embedding_model()
+        get_embedding_model(mock_ai_config.return_value)
 
 
 def test_get_embedding_model_invalid_backend(mock_ai_config):
@@ -183,55 +182,37 @@ def test_get_embedding_model_invalid_backend(mock_ai_config):
         ValueError,
         match="Unsupported embedding backend: INVALID_BACKEND",
     ):
-        get_embedding_model()
+        get_embedding_model(mock_ai_config.return_value)
 
 
-def test_get_embedding_dim_infers_and_saves(temp_llm_index_dir, mock_ai_config):
-    mock_ai_config.return_value.llm_embedding_backend = "openai-like"
-    mock_ai_config.return_value.llm_embedding_model = None
-
-    class DummyEmbedding:
-        def get_text_embedding(self, text):
-            return [0.0] * 7
-
-    with patch(
-        "paperless_ai.embedding.get_embedding_model",
-        return_value=DummyEmbedding(),
-    ) as mock_get:
-        dim = get_embedding_dim()
-        mock_get.assert_called_once()
-
-    assert dim == 7
-    meta = json.loads((temp_llm_index_dir / "meta.json").read_text())
-    assert meta == {"embedding_model": "text-embedding-3-small", "dim": 7}
+@pytest.mark.parametrize(
+    ("backend", "expected_default"),
+    [
+        (LLMEmbeddingBackend.OPENAI_LIKE, "text-embedding-3-small"),
+        (LLMEmbeddingBackend.HUGGINGFACE, "sentence-transformers/all-MiniLM-L6-v2"),
+        (LLMEmbeddingBackend.OLLAMA, "embeddinggemma"),
+    ],
+)
+def test_get_configured_model_name_falls_back_to_backend_default(
+    mock_ai_config,
+    backend,
+    expected_default,
+):
+    """When no model is explicitly configured, each backend has a distinct default."""
+    config = mock_ai_config.return_value
+    config.llm_embedding_backend = backend
+    config.llm_embedding_model = None
+    assert get_configured_model_name(config) == expected_default
 
 
-def test_get_embedding_dim_reads_existing_meta(temp_llm_index_dir, mock_ai_config):
-    mock_ai_config.return_value.llm_embedding_backend = "openai-like"
-    mock_ai_config.return_value.llm_embedding_model = None
-
-    (temp_llm_index_dir / "meta.json").write_text(
-        json.dumps({"embedding_model": "text-embedding-3-small", "dim": 11}),
-    )
-
-    with patch("paperless_ai.embedding.get_embedding_model") as mock_get:
-        assert get_embedding_dim() == 11
-        mock_get.assert_not_called()
-
-
-def test_get_embedding_dim_raises_on_model_change(temp_llm_index_dir, mock_ai_config):
-    mock_ai_config.return_value.llm_embedding_backend = "openai-like"
-    mock_ai_config.return_value.llm_embedding_model = None
-
-    (temp_llm_index_dir / "meta.json").write_text(
-        json.dumps({"embedding_model": "old", "dim": 11}),
-    )
-
-    with pytest.raises(
-        RuntimeError,
-        match="Embedding model changed from old to text-embedding-3-small",
-    ):
-        get_embedding_dim()
+def test_get_configured_model_name_explicit_overrides_default(mock_ai_config):
+    """An explicit model name overrides the backend default for all backends."""
+    config = mock_ai_config.return_value
+    config.llm_embedding_backend = LLMEmbeddingBackend.OPENAI_LIKE
+    config.llm_embedding_model = "my-custom-model"
+    # The backend default for OPENAI_LIKE is "text-embedding-3-small", so if
+    # the explicit name was ignored we'd get the wrong result.
+    assert get_configured_model_name(config) == "my-custom-model"
 
 
 def test_build_llm_index_text(mock_document):
@@ -243,12 +224,17 @@ def test_build_llm_index_text(mock_document):
 
         result = build_llm_index_text(mock_document)
 
-        assert "Title: Test Title" in result
-        assert "Filename: test_file.pdf" in result
-        assert "Created: 2023-01-01" in result
-        assert "Tags: Tag1, Tag2" in result
-        assert "Document Type: Invoice" in result
-        assert "Correspondent: Test Correspondent" in result
+        # Structured fields live in node.metadata for LLM context -- not body text
+        assert "Title: Test Title" not in result
+        assert "Created: 2023-01-01" not in result
+        assert "Tags: Tag1, Tag2" not in result
+        assert "Document Type: Invoice" not in result
+        assert "Correspondent: Test Correspondent" not in result
+        assert "Filename:" not in result
+        assert "Storage Path:" not in result
+        assert "Archive Serial Number:" not in result
+
+        # Fields without a metadata equivalent stay in body text
         assert "Notes: Note1,Note2" in result
         assert "Content:\n\nThis is the document content." in result
         assert "Custom Field - Field1: Value1\nCustom Field - Field2: Value2" in result

src/paperless_ai/tests/test_index_locking.py

@@ -0,0 +1,134 @@
+import logging
+import sqlite3
+import threading
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pytest
+from django.conf import settings
+from filelock import ReadWriteLock
+from llama_index.core.schema import TextNode
+from pytest_django.fixtures import SettingsWrapper
+
+from paperless_ai import indexing
+from paperless_ai.vector_store import PaperlessSqliteVecVectorStore
+
+DIM = 8
+
+
+def _node(node_id: str, document_id: str, *, seed: float = 0.0) -> TextNode:
+    node = TextNode(
+        id_=node_id,
+        text="chunk",
+        metadata={"document_id": document_id, "modified": "2026-06-01T00:00:00"},
+    )
+    node.relationships = {}
+    node.embedding = [seed + i / 100 for i in range(DIM)]
+    return node
+
+
+def _seed_bloated_index(index_dir: Path) -> None:
+    """Create an index whose cumulative inserts far exceed live rows."""
+    store = PaperlessSqliteVecVectorStore(uri=str(index_dir))
+    store.add([_node(f"d{j}", str(j), seed=float(j)) for j in range(20)])
+    for cycle in range(6):
+        for j in range(20):
+            store.upsert_document(
+                str(j),
+                [_node(f"d{j}-c{cycle}", str(j), seed=float(j))],
+            )
+    store.client.close()
+
+
+def _bloat_ratio(index_dir: Path) -> float:
+    store = PaperlessSqliteVecVectorStore(uri=str(index_dir))
+    live = store.client.execute("SELECT count(*) FROM documents").fetchone()[0]
+    row = store.client.execute(
+        "SELECT value FROM index_meta WHERE key = 'total_inserts'",
+    ).fetchone()
+    total = int(row["value"]) if row else live
+    store.client.close()
+    return total / max(live, 1)
+
+
+def _integrity_ok(index_dir: Path) -> bool:
+    store = PaperlessSqliteVecVectorStore(uri=str(index_dir))
+    result = store.client.execute("PRAGMA integrity_check").fetchone()[0]
+    rows = store.client.execute("SELECT count(*) FROM documents").fetchone()[0]
+    store.client.close()
+    return result == "ok" and rows == 20
+
+
+def _reader_lock() -> ReadWriteLock:
+    # A distinct instance simulates a reader in another process: it coordinates
+    # with the production lock purely through SQLite, never reentrant upgrade.
+    return ReadWriteLock(str(settings.LLM_INDEX_RWLOCK), is_singleton=False)
+
+
+class TestCompactionLock:
+    def test_compaction_skips_when_a_reader_holds_the_lock(
+        self,
+        temp_llm_index_dir: Path,
+        settings: SettingsWrapper,
+        caplog: pytest.LogCaptureFixture,
+    ) -> None:
+        _seed_bloated_index(temp_llm_index_dir)
+        settings.LLM_INDEX_COMPACTION_LOCK_TIMEOUT = 0.3
+
+        lock = _reader_lock()
+        with lock.read_lock(), caplog.at_level(logging.INFO):
+            indexing.llm_index_compact()  # must not raise
+        lock.close()
+
+        # Swap was skipped: bloat remains, nothing corrupted, data intact.
+        assert _integrity_ok(temp_llm_index_dir)
+        assert _bloat_ratio(temp_llm_index_dir) > 2
+        assert "Skipping LLM index compaction" in caplog.text
+
+    def test_compaction_runs_when_no_reader_holds_the_lock(
+        self,
+        temp_llm_index_dir: Path,
+    ) -> None:
+        _seed_bloated_index(temp_llm_index_dir)
+        assert _bloat_ratio(temp_llm_index_dir) > 2
+
+        indexing.llm_index_compact()
+
+        assert _bloat_ratio(temp_llm_index_dir) == pytest.approx(1.0)
+        assert _integrity_ok(temp_llm_index_dir)
+
+    def test_normal_write_is_not_gated_by_the_compaction_lock(
+        self,
+        temp_llm_index_dir: Path,
+    ) -> None:
+        """A held exclusive lock must not block ordinary writes (WAL handles them)."""
+        _seed_bloated_index(temp_llm_index_dir)
+        done = threading.Event()
+
+        def remove() -> None:
+            indexing.llm_index_remove_document(MagicMock(id=999))
+            done.set()
+
+        holder = _reader_lock()
+        with holder.write_lock():
+            t = threading.Thread(target=remove)
+            t.start()
+            finished = done.wait(timeout=5)
+        t.join(timeout=2)
+        holder.close()
+        assert finished, "a normal write blocked on the compaction lock"
+
+
+class TestReadStore:
+    def test_closes_connection_on_exit(self, temp_llm_index_dir: Path) -> None:
+        with indexing.read_store() as store:
+            conn = store.client
+            assert conn.execute("SELECT 1").fetchone()[0] == 1
+        with pytest.raises(sqlite3.ProgrammingError):
+            conn.execute("SELECT 1")
+
+    def test_concurrent_readers_do_not_block(self, temp_llm_index_dir: Path) -> None:
+        _seed_bloated_index(temp_llm_index_dir)
+        with indexing.read_store() as a, indexing.read_store() as b:
+            assert a.table_exists()
+            assert b.table_exists()

src/paperless_ai/tests/test_lazy_imports.py

@@ -0,0 +1,25 @@
+import subprocess
+import sys
+from pathlib import Path
+
+_SRC_DIR = Path(__file__).parent.parent.parent
+
+
+class TestLazyAiImports:
+    def test_importing_tasks_does_not_load_ai_libraries(self) -> None:
+        code = (
+            "import os, django, sys\n"
+            "os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'paperless.settings')\n"
+            "django.setup()\n"
+            "import documents.tasks  # noqa: F401\n"
+            "leaked = [m for m in ('lancedb', 'pyarrow', 'llama_index', 'sqlite_vec') "
+            "if m in sys.modules]\n"
+            "assert not leaked, f'AI libraries leaked into the light path: {leaked}'\n"
+        )
+        result = subprocess.run(
+            [sys.executable, "-c", code],
+            capture_output=True,
+            text=True,
+            cwd=_SRC_DIR,
+        )
+        assert result.returncode == 0, result.stdout + result.stderr

src/paperless_ai/tests/test_matching.py

@@ -1,12 +1,15 @@
+import difflib
 from unittest.mock import patch
 
 import pytest
+import pytest_mock
 from django.test import TestCase
 
 from documents.models import Correspondent
 from documents.models import DocumentType
 from documents.models import StoragePath
 from documents.models import Tag
+from documents.tests.factories import TagFactory
 from paperless_ai.matching import extract_unmatched_names
 from paperless_ai.matching import match_correspondents_by_name
 from paperless_ai.matching import match_document_types_by_name
@@ -87,6 +90,95 @@ class TestAIMatching(TestCase):
         self.assertEqual(result[1].name, "Test Tag 2")
 
 
+class TestHintedMatching:
+    def test_hinted_verbatim_skips_fuzzy(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.matching.get_objects_for_user_owner_aware",
+            return_value=[TagFactory.build(name="Bloodwork")],
+        )
+        spy = mocker.spy(difflib, "get_close_matches")
+
+        result = match_tags_by_name(
+            ["Bloodwork"],
+            user=None,
+            hinted_names={"Bloodwork"},
+        )
+
+        assert [t.name for t in result] == ["Bloodwork"]
+        spy.assert_not_called()
+
+    def test_unhinted_name_still_fuzzy_matches(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.matching.get_objects_for_user_owner_aware",
+            return_value=[TagFactory.build(name="Bloodwork")],
+        )
+
+        # "Bloodwrok" is a typo not in hints -> fuzzy still maps it to Bloodwork.
+        result = match_tags_by_name(
+            ["Bloodwrok"],
+            user=None,
+            hinted_names={"Taxes"},
+        )
+
+        assert [t.name for t in result] == ["Bloodwork"]
+
+    def test_hinted_name_with_whitespace_exact_matches(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.matching.get_objects_for_user_owner_aware",
+            return_value=[TagFactory.build(name="Bloodwork")],
+        )
+        spy = mocker.spy(difflib, "get_close_matches")
+
+        result = match_tags_by_name(
+            ["Bloodwork "],
+            user=None,
+            hinted_names={"Bloodwork"},
+        )
+
+        assert [t.name for t in result] == ["Bloodwork"]
+        spy.assert_not_called()
+
+    def test_hinted_name_absent_from_queryset_is_skipped_not_fuzzed(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        # A hint with no exact object must not fall through to fuzzy.
+        mocker.patch(
+            "paperless_ai.matching.get_objects_for_user_owner_aware",
+            return_value=[TagFactory.build(name="Bloodwork")],
+        )
+
+        result = match_tags_by_name(
+            ["Bloodwrok"],
+            user=None,
+            hinted_names={"Bloodwrok"},
+        )
+
+        assert result == []
+
+    def test_backward_compatible_without_kwarg(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.matching.get_objects_for_user_owner_aware",
+            return_value=[TagFactory.build(name="Test Tag 1")],
+        )
+
+        result = match_tags_by_name(["Test Tag 1", "Nonexistent"], user=None)
+
+        assert [t.name for t in result] == ["Test Tag 1"]
+
+
 @pytest.mark.django_db
 class TestExtractUnmatchedNamesNormalization:
     def test_punctuated_name_already_matched_is_not_returned_as_unmatched(

src/paperless_ai/tests/test_taxonomy.py

@@ -0,0 +1,220 @@
+from types import SimpleNamespace
+
+import pytest_mock
+
+from documents.tests.factories import DocumentFactory
+from paperless_ai.taxonomy import TaxonomyHints
+from paperless_ai.taxonomy import build_taxonomy_hints_from_nodes
+from paperless_ai.taxonomy import format_hints_for_prompt
+from paperless_ai.taxonomy import get_taxonomy_hints_for_document
+
+
+def make_node(**metadata: object) -> SimpleNamespace:
+    """A stand-in for NodeWithScore: only ``.metadata`` is accessed."""
+    return SimpleNamespace(metadata=metadata)
+
+
+class TestBuildTaxonomyHintsFromNodes:
+    def test_returns_all_four_keys(self) -> None:
+        hints = build_taxonomy_hints_from_nodes([])
+        assert set(hints.keys()) == {
+            "tags",
+            "document_types",
+            "correspondents",
+            "storage_paths",
+        }
+
+    def test_collects_and_sorts_values(self) -> None:
+        nodes = [
+            make_node(
+                tags=["Taxes", "Bloodwork"],
+                document_type="Invoice",
+                correspondent="IRS",
+                storage_path="Financial",
+            ),
+        ]
+        hints = build_taxonomy_hints_from_nodes(nodes)
+        assert hints["tags"] == ["Bloodwork", "Taxes"]
+        assert hints["document_types"] == ["Invoice"]
+        assert hints["correspondents"] == ["IRS"]
+        assert hints["storage_paths"] == ["Financial"]
+
+    def test_deduplicates_across_nodes(self) -> None:
+        nodes = [
+            make_node(tags=["Taxes"], document_type="Invoice"),
+            make_node(tags=["Taxes", "Medical"], document_type="Invoice"),
+        ]
+        hints = build_taxonomy_hints_from_nodes(nodes)
+        assert hints["tags"] == ["Medical", "Taxes"]
+        assert hints["document_types"] == ["Invoice"]
+
+    def test_none_values_skipped(self) -> None:
+        nodes = [
+            make_node(
+                tags=["Taxes", None, ""],
+                document_type=None,
+                correspondent=None,
+                storage_path=None,
+            ),
+        ]
+        hints = build_taxonomy_hints_from_nodes(nodes)
+        assert hints["tags"] == ["Taxes"]
+        assert hints["document_types"] == []
+        assert hints["correspondents"] == []
+        assert hints["storage_paths"] == []
+
+    def test_missing_storage_path_key_handled(self) -> None:
+        # Pre-enrichment nodes have no storage_path key at all.
+        nodes = [make_node(tags=["Taxes"], document_type="Invoice")]
+        hints = build_taxonomy_hints_from_nodes(nodes)
+        assert hints["storage_paths"] == []
+
+    def test_empty_node_list_all_empty(self) -> None:
+        hints = build_taxonomy_hints_from_nodes([])
+        assert hints == {
+            "tags": [],
+            "document_types": [],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+
+    def test_output_stable_across_calls(self) -> None:
+        nodes = [make_node(tags=["b", "a", "c"])]
+        assert build_taxonomy_hints_from_nodes(
+            nodes,
+        ) == build_taxonomy_hints_from_nodes(nodes)
+
+
+class TestFormatHintsForPrompt:
+    def test_all_blocks_present_when_all_categories_nonempty(self) -> None:
+        hints: TaxonomyHints = {
+            "tags": ["Bloodwork"],
+            "document_types": ["Invoice"],
+            "correspondents": ["IRS"],
+            "storage_paths": ["Financial"],
+        }
+        result = format_hints_for_prompt(hints)
+        assert "Available tags:" in result
+        assert "Available document types:" in result
+        assert "Available correspondents:" in result
+        assert "Available storage paths:" in result
+        assert "- Bloodwork" in result
+
+    def test_empty_category_produces_no_block(self) -> None:
+        hints: TaxonomyHints = {
+            "tags": ["Bloodwork"],
+            "document_types": [],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+        result = format_hints_for_prompt(hints)
+        assert "Available tags:" in result
+        assert "Available document types:" not in result
+        assert "Available correspondents:" not in result
+        assert "Available storage paths:" not in result
+
+    def test_all_empty_produces_empty_string(self) -> None:
+        hints: TaxonomyHints = {
+            "tags": [],
+            "document_types": [],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+        assert format_hints_for_prompt(hints) == ""
+
+    def test_instruction_line_appears_once(self) -> None:
+        hints: TaxonomyHints = {
+            "tags": ["Bloodwork"],
+            "document_types": ["Invoice"],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+        result = format_hints_for_prompt(hints)
+        assert result.count("Prefer existing names from these lists verbatim") == 1
+
+
+class TestGetTaxonomyHintsForDocument:
+    def test_returns_none_when_embedding_backend_off(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.taxonomy.AIConfig",
+            return_value=SimpleNamespace(llm_embedding_backend=None),
+        )
+        retrieve = mocker.patch("paperless_ai.taxonomy.retrieve_similar_nodes")
+
+        result = get_taxonomy_hints_for_document(DocumentFactory.build(), user=None)
+
+        assert result is None
+        retrieve.assert_not_called()
+
+    def test_passes_owner_aware_ids_when_user_present(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.taxonomy.AIConfig",
+            return_value=SimpleNamespace(llm_embedding_backend="huggingface"),
+        )
+        mocker.patch(
+            "paperless_ai.taxonomy.visible_document_ids_for_user",
+            return_value=[1, 2, 3],
+        )
+        retrieve = mocker.patch(
+            "paperless_ai.taxonomy.retrieve_similar_nodes",
+            return_value=[],
+        )
+        document = DocumentFactory.build()
+        user = mocker.MagicMock()
+
+        get_taxonomy_hints_for_document(document, user=user)
+
+        retrieve.assert_called_once_with(
+            document=document,
+            document_ids=[1, 2, 3],
+        )
+
+    def test_returns_populated_hints_when_nodes_found(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.taxonomy.AIConfig",
+            return_value=SimpleNamespace(llm_embedding_backend="huggingface"),
+        )
+        mocker.patch(
+            "paperless_ai.taxonomy.retrieve_similar_nodes",
+            return_value=[make_node(tags=["Taxes"], document_type="Invoice")],
+        )
+
+        result = get_taxonomy_hints_for_document(DocumentFactory.build(), user=None)
+
+        assert result == {
+            "tags": ["Taxes"],
+            "document_types": ["Invoice"],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+
+    def test_returns_empty_hints_not_none_when_no_nodes(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        mocker.patch(
+            "paperless_ai.taxonomy.AIConfig",
+            return_value=SimpleNamespace(llm_embedding_backend="huggingface"),
+        )
+        mocker.patch(
+            "paperless_ai.taxonomy.retrieve_similar_nodes",
+            return_value=[],
+        )
+
+        result = get_taxonomy_hints_for_document(DocumentFactory.build(), user=None)
+
+        assert result == {
+            "tags": [],
+            "document_types": [],
+            "correspondents": [],
+            "storage_paths": [],
+        }

src/paperless_ai/tests/test_vector_store.py

@@ -0,0 +1,606 @@
+import sqlite3
+from collections.abc import Generator
+from pathlib import Path
+
+import pytest
+from llama_index.core.schema import TextNode
+from llama_index.core.vector_stores.types import FilterOperator
+from llama_index.core.vector_stores.types import MetadataFilter
+from llama_index.core.vector_stores.types import MetadataFilters
+from llama_index.core.vector_stores.types import VectorStoreQuery
+
+from paperless_ai.vector_store import DB_FILENAME
+from paperless_ai.vector_store import DEFAULT_TABLE_NAME
+from paperless_ai.vector_store import MIGRATIONS
+from paperless_ai.vector_store import SCHEMA_VERSION
+from paperless_ai.vector_store import Migration
+from paperless_ai.vector_store import PaperlessSqliteVecVectorStore
+from paperless_ai.vector_store import _build_where
+
+DIM = 16
+
+
+def make_node(
+    node_id: str,
+    document_id: str,
+    *,
+    modified: str = "2026-06-10T00:00:00",
+    seed: float = 0.0,
+    text: str = "some text",
+) -> TextNode:
+    node = TextNode(
+        id_=node_id,
+        text=text,
+        metadata={"document_id": document_id, "modified": modified},
+    )
+    node.relationships = {}
+    node.embedding = [seed + i / 100 for i in range(DIM)]
+    return node
+
+
+@pytest.fixture
+def store(tmp_path: Path) -> Generator[PaperlessSqliteVecVectorStore, None, None]:
+    with PaperlessSqliteVecVectorStore(uri=str(tmp_path)) as store:
+        yield store
+
+
+def _query(
+    store: PaperlessSqliteVecVectorStore,
+    embedding: list[float],
+    top_k: int = 5,
+    filters=None,
+):
+    return store.query(
+        VectorStoreQuery(
+            query_embedding=embedding,
+            similarity_top_k=top_k,
+            filters=filters,
+        ),
+    )
+
+
+def _eq_filter(key: str, value: str):
+    return MetadataFilters(
+        filters=[MetadataFilter(key=key, operator=FilterOperator.EQ, value=value)],
+    )
+
+
+def _in_filter(document_ids: list[str]):
+    return MetadataFilters(
+        filters=[
+            MetadataFilter(
+                key="document_id",
+                operator=FilterOperator.IN,
+                value=document_ids,
+            ),
+        ],
+    )
+
+
+class TestCrud:
+    def test_add_then_query_returns_node(self, store) -> None:
+        node = make_node("n1", "1")
+        assert store.add([node]) == ["n1"]
+        result = _query(store, node.embedding, top_k=1)
+        assert result.ids == ["n1"]
+        assert result.nodes[0].metadata["document_id"] == "1"
+        # cosine distance of the identical vector is 0 -> similarity 1
+        assert result.similarities[0] == pytest.approx(1.0)
+
+    def test_query_empty_store_returns_empty_no_raise(self, store) -> None:
+        result = _query(store, [0.0] * DIM)
+        assert result.ids == [] and result.nodes == [] and result.similarities == []
+
+    def test_add_empty_list_is_noop(self, store) -> None:
+        assert store.add([]) == []
+        assert not store.table_exists()
+
+    def test_delete_removes_all_chunks_of_document(self, store) -> None:
+        store.add([make_node("a1", "1"), make_node("a2", "1"), make_node("b1", "2")])
+        store.delete("1")
+        result = _query(store, [0.0] * DIM, top_k=10)
+        assert result.ids == ["b1"]
+
+    def test_query_with_in_filter_scopes_results(self, store) -> None:
+        store.add(
+            [
+                make_node("a1", "1", seed=0.0),
+                make_node("b1", "2", seed=1.0),
+                make_node("c1", "3", seed=2.0),
+            ],
+        )
+        result = _query(store, [0.0] * DIM, top_k=10, filters=_in_filter(["2", "3"]))
+        assert sorted(result.ids) == ["b1", "c1"]
+
+    def test_query_respects_top_k_with_filter(self, store) -> None:
+        # k semantics: global top-k even with IN filters (document_id is a
+        # metadata column, not a partition key -- see design doc).
+        store.add(
+            [make_node(f"n{i}", str(i % 4), seed=float(i)) for i in range(12)],
+        )
+        result = _query(
+            store,
+            [0.0] * DIM,
+            top_k=3,
+            filters=_in_filter(["0", "1", "2", "3"]),
+        )
+        assert len(result.ids) == 3
+        assert result.similarities == sorted(result.similarities, reverse=True)
+
+    def test_get_nodes_filter_and_empty_paths(self, store) -> None:
+        assert store.get_nodes(filters=_in_filter(["1"])) == []  # no table yet
+        store.add([make_node("a1", "1"), make_node("b1", "2")])
+        nodes = store.get_nodes(filters=_in_filter(["1"]))
+        assert [n.node_id for n in nodes] == ["a1"]
+        assert nodes[0].embedding is not None
+        assert store.get_nodes(filters=_in_filter(["999"])) == []
+
+    def test_query_with_eq_filter_scopes_results(self, store) -> None:
+        store.add(
+            [
+                make_node("a1", "1", seed=0.0),
+                make_node("b1", "2", seed=1.0),
+                make_node("c1", "3", seed=2.0),
+            ],
+        )
+        result = _query(
+            store,
+            [0.0] * DIM,
+            top_k=10,
+            filters=_eq_filter("document_id", "2"),
+        )
+        assert result.ids == ["b1"]
+
+    def test_get_nodes_node_ids_not_implemented(self, store) -> None:
+        with pytest.raises(NotImplementedError):
+            store.get_nodes(node_ids=["x"])
+
+    def test_fresh_instance_sees_existing_table(self, store, tmp_path: Path) -> None:
+        store.add([make_node("a1", "1")])
+        with PaperlessSqliteVecVectorStore(uri=str(tmp_path)) as reopened:
+            assert reopened.table_exists()
+            assert reopened.vector_dim() == DIM
+            assert _query(reopened, [0.0] * DIM, top_k=1).ids == ["a1"]
+
+    def test_table_exists_and_drop(self, store) -> None:
+        assert not store.table_exists()
+        store.add([make_node("a1", "1")])
+        assert store.table_exists()
+        store.drop_table()
+        assert not store.table_exists()
+        assert store.vector_dim() is None
+
+
+class TestBuildWhere:
+    def test_fails_closed_when_no_filter_is_translatable(self) -> None:
+        # A nested MetadataFilters is not a MetadataFilter, so it is skipped.
+        # With no translatable clauses, the function must fail closed rather
+        # than emit "()" (invalid SQL) and never widen document access.
+        nested = MetadataFilters(
+            filters=[
+                MetadataFilter(
+                    key="document_id",
+                    operator=FilterOperator.EQ,
+                    value="1",
+                ),
+            ],
+        )
+        where, params = _build_where(MetadataFilters(filters=[nested]))
+        assert where == "1 = 0"
+        assert params == []
+
+    def test_query_with_untranslatable_filter_returns_no_rows(self, store) -> None:
+        store.add([make_node("a1", "1"), make_node("b1", "2")])
+        nested = MetadataFilters(
+            filters=[
+                MetadataFilter(
+                    key="document_id",
+                    operator=FilterOperator.EQ,
+                    value="1",
+                ),
+            ],
+        )
+        filters = MetadataFilters(filters=[nested])
+        # Must not raise (no "WHERE ()") and must return nothing (fail closed).
+        assert _query(store, [0.0] * DIM, top_k=5, filters=filters).ids == []
+        assert store.get_nodes(filters=filters) == []
+
+
+class TestUpsert:
+    def test_upsert_replaces_and_prunes_stale_chunks(self, store) -> None:
+        store.add(
+            [make_node("d1c1", "1"), make_node("d1c2", "1"), make_node("d2c1", "2")],
+        )
+        store.upsert_document("1", [make_node("d1new", "1")])
+        result = _query(store, [0.0] * DIM, top_k=10)
+        assert sorted(result.ids) == ["d1new", "d2c1"]
+
+    def test_upsert_creates_table_when_missing(self, store) -> None:
+        store.upsert_document("1", [make_node("a1", "1")])
+        assert _query(store, [0.0] * DIM, top_k=1).ids == ["a1"]
+
+    def test_upsert_empty_nodes_removes_document(self, store) -> None:
+        store.add([make_node("a1", "1"), make_node("b1", "2")])
+        store.upsert_document("1", [])
+        assert _query(store, [0.0] * DIM, top_k=10).ids == ["b1"]
+
+    def test_upsert_is_atomic_for_concurrent_readers(
+        self,
+        store,
+        tmp_path: Path,
+    ) -> None:
+        """A second connection must never observe document 1 half-replaced."""
+        store.add([make_node("a1", "1"), make_node("a2", "1")])
+        with PaperlessSqliteVecVectorStore(uri=str(tmp_path)) as reader:
+            store.upsert_document("1", [make_node("a3", "1")])
+            ids = [n.node_id for n in reader.get_nodes(filters=_in_filter(["1"]))]
+            assert ids == ["a3"]
+
+
+class TestMetadataCoercion:
+    def test_none_metadata_values_become_empty_strings(self, store) -> None:
+        node = make_node("a1", "1")
+        node.metadata["modified"] = None
+        store.add([node])  # must not raise (vec0 rejects NULL metadata)
+        assert store.get_modified_times() == {"1": ""}
+
+
+class TestModelNameTracking:
+    def test_stored_model_name_none_without_table(self, tmp_path: Path) -> None:
+        with PaperlessSqliteVecVectorStore(
+            uri=str(tmp_path),
+            embed_model_name="model-a",
+        ) as store:
+            assert store.stored_model_name() is None
+
+    def test_model_name_stored_after_add_and_persists(self, tmp_path: Path) -> None:
+        with PaperlessSqliteVecVectorStore(
+            uri=str(tmp_path),
+            embed_model_name="model-a",
+        ) as store:
+            store.add([make_node("a1", "1")])
+            assert store.stored_model_name() == "model-a"
+        with PaperlessSqliteVecVectorStore(uri=str(tmp_path)) as reopened:
+            assert reopened.stored_model_name() == "model-a"
+
+    def test_config_mismatch_semantics(self, tmp_path: Path) -> None:
+        with PaperlessSqliteVecVectorStore(
+            uri=str(tmp_path),
+            embed_model_name="model-a",
+        ) as store:
+            assert not store.config_mismatch("anything")  # no table yet
+            store.add([make_node("a1", "1")])
+            assert not store.config_mismatch("model-a")
+            assert store.config_mismatch("model-b")
+
+    def test_config_mismatch_false_when_table_predates_tracking(
+        self,
+        tmp_path: Path,
+    ) -> None:
+        with PaperlessSqliteVecVectorStore(uri=str(tmp_path)) as store:  # no model name
+            store.add([make_node("a1", "1")])
+            assert not store.config_mismatch("model-a")
+
+
+class TestGetModifiedTimes:
+    def test_empty_store_returns_empty_dict(self, store) -> None:
+        assert store.get_modified_times() == {}
+
+    def test_returns_one_entry_per_document(self, store) -> None:
+        store.add(
+            [
+                make_node("a1", "1", modified="2026-01-01T00:00:00"),
+                make_node("a2", "1", modified="2026-01-01T00:00:00"),
+                make_node("b1", "2", modified="2026-02-02T00:00:00"),
+            ],
+        )
+        assert store.get_modified_times() == {
+            "1": "2026-01-01T00:00:00",
+            "2": "2026-02-02T00:00:00",
+        }
+
+
+class TestCompact:
+    def _bloat_ratio(self, store) -> float:
+        live = store.client.execute(
+            "SELECT count(*) FROM documents",
+        ).fetchone()[0]
+        # vec0 0.1.9 does not accumulate deleted rows in the _rowids shadow
+        # table, so we track cumulative inserts in index_meta instead.
+        row = store.client.execute(
+            "SELECT value FROM index_meta WHERE key = 'total_inserts'",
+        ).fetchone()
+        total = int(row["value"]) if row else live
+        return total / max(live, 1)
+
+    def _churn(self, store, cycles: int) -> None:
+        for i in range(cycles):
+            store.upsert_document(
+                "1",
+                [make_node(f"gen{i}-{j}", "1", seed=float(j)) for j in range(20)],
+            )
+
+    def test_compact_noop_below_threshold(self, store) -> None:
+        store.add([make_node("a1", "1")])
+        store.compact()
+        assert _query(store, [0.0] * DIM, top_k=1).ids == ["a1"]
+
+    def test_force_compact_preserves_rows_and_metadata(self, store) -> None:
+        store.add([make_node("a1", "1"), make_node("b1", "2", seed=3.0)])
+        self._churn(store, 5)
+        before = {
+            n.node_id: n.metadata
+            for n in store.get_nodes(filters=_in_filter(["1", "2"]))
+        }
+        store.compact(force=True)
+        after = {
+            n.node_id: n.metadata
+            for n in store.get_nodes(filters=_in_filter(["1", "2"]))
+        }
+        assert after == before
+        assert self._bloat_ratio(store) == pytest.approx(1.0)
+        # store remains fully usable after the rebuild; use a seed far from all
+        # existing nodes (gen4-0..gen4-19 have seeds 0..19) so cosine KNN is
+        # unambiguous at top_k=1.
+        store.upsert_document("3", [make_node("c1", "3", seed=100.0)])
+        assert "c1" in _query(store, [100.0] * DIM, top_k=1).ids
+
+    def test_auto_compact_triggers_on_churn(self, store) -> None:
+        store.add([make_node(f"s{j}", "1", seed=float(j)) for j in range(20)])
+        self._churn(store, 5)
+        assert self._bloat_ratio(store) > 2
+        store.compact()
+        assert self._bloat_ratio(store) == pytest.approx(1.0)
+
+    def test_compact_on_missing_table_is_noop(self, store) -> None:
+        store.compact()
+        store.compact(force=True)
+
+    def test_failed_compact_removes_temp_wal_and_shm(
+        self,
+        store,
+        tmp_path: Path,
+        monkeypatch,
+    ) -> None:
+        """A compact() that raises mid-rebuild must leave no .compact* files.
+
+        Normally the sole connection's close() checkpoints the temp WAL away,
+        but a concurrent reader keeps -wal/-shm alive, so the cleanup must
+        unlink them explicitly (as the structural-migration path does).
+        """
+        store.add([make_node("a1", "1")])
+        compact_path = str(tmp_path / DB_FILENAME) + ".compact"
+        held: list[sqlite3.Connection] = []
+
+        def boom(conn: sqlite3.Connection, dim: int) -> None:
+            # Hold an extra connection so close() of the rebuild connection is
+            # not the last one -> the temp -wal/-shm survive the checkpoint.
+            extra = sqlite3.connect(compact_path)
+            extra.execute("SELECT 1").fetchall()
+            held.append(extra)
+            raise RuntimeError("boom")
+
+        monkeypatch.setattr(
+            PaperlessSqliteVecVectorStore,
+            "_create_vec_table",
+            staticmethod(boom),
+        )
+        try:
+            with pytest.raises(RuntimeError):
+                store.compact(force=True)
+            assert sorted(p.name for p in tmp_path.glob("*.compact*")) == []
+        finally:
+            for c in held:
+                c.close()
+
+
+class TestDbFile:
+    def test_single_db_file_in_index_dir(self, store, tmp_path: Path) -> None:
+        store.add([make_node("a1", "1")])
+        assert (tmp_path / DB_FILENAME).exists()
+
+    def test_wal_mode_enabled(self, store) -> None:
+        assert (
+            store.client.execute("PRAGMA journal_mode").fetchone()[0].lower() == "wal"
+        )
+
+
+class TestMigrations:
+    """Tests for the schema migration machinery."""
+
+    def _schema_version(self, store: PaperlessSqliteVecVectorStore) -> int | None:
+        row = store.client.execute(
+            "SELECT value FROM index_meta WHERE key = 'schema_version'",
+        ).fetchone()
+        return int(row[0]) if row else None
+
+    def test_new_table_records_schema_version(self, store) -> None:
+        store.add([make_node("a1", "1")])
+        assert self._schema_version(store) == SCHEMA_VERSION
+
+    def test_check_migrations_no_table_returns_false(self, store) -> None:
+        assert store.check_and_run_migrations() is False
+
+    def test_check_migrations_current_version_returns_false(self, store) -> None:
+        store.add([make_node("a1", "1")])
+        assert store.check_and_run_migrations() is False
+
+    def test_reembed_migration_returns_true(self, store, tmp_path: Path) -> None:
+        store.add([make_node("a1", "1")])
+        migration = Migration(
+            from_version=1,
+            to_version=2,
+            kind="re-embed",
+            description="test re-embed",
+        )
+        MIGRATIONS.append(migration)
+        try:
+            from paperless_ai import vector_store as vs_mod
+
+            original = vs_mod.SCHEMA_VERSION
+            vs_mod.SCHEMA_VERSION = 2
+            result = store.check_and_run_migrations()
+        finally:
+            MIGRATIONS.remove(migration)
+            vs_mod.SCHEMA_VERSION = original
+        assert result is True
+
+    def test_structural_migration_copies_rows_and_updates_version(
+        self,
+        store,
+        tmp_path: Path,
+    ) -> None:
+        store.add([make_node("a1", "1"), make_node("b1", "2")])
+
+        def apply(
+            src: sqlite3.Connection,
+            dst: sqlite3.Connection,
+            dim: int,
+        ) -> None:
+            dst.execute(  # nosemgrep
+                f"CREATE VIRTUAL TABLE {DEFAULT_TABLE_NAME} USING vec0("
+                "id TEXT PRIMARY KEY, document_id TEXT, modified TEXT,"
+                f" +node_content TEXT, embedding float[{dim}] distance_metric=cosine"
+                ")",
+            )
+            dst.execute(
+                "INSERT INTO index_meta (key, value) VALUES ('dim', ?) "
+                "ON CONFLICT(key) DO UPDATE SET value = excluded.value",
+                (str(dim),),
+            )
+            rows = src.execute(
+                "SELECT id, document_id, modified, node_content, embedding "
+                f"FROM {DEFAULT_TABLE_NAME}",
+            ).fetchall()
+            dst.execute("BEGIN IMMEDIATE")
+            dst.executemany(
+                f"INSERT INTO {DEFAULT_TABLE_NAME} "
+                "(id, document_id, modified, node_content, embedding) "
+                "VALUES (?, ?, ?, ?, ?)",
+                [
+                    (
+                        r["id"],
+                        r["document_id"],
+                        r["modified"],
+                        r["node_content"],
+                        bytes(r["embedding"]),
+                    )
+                    for r in rows
+                ],
+            )
+            dst.execute(
+                "INSERT INTO index_meta (key, value) VALUES ('total_inserts', ?) "
+                "ON CONFLICT(key) DO UPDATE SET value = excluded.value",
+                (str(len(rows)),),
+            )
+            dst.execute("COMMIT")
+
+        migration = Migration(
+            from_version=1,
+            to_version=2,
+            kind="structural",
+            description="test structural",
+            apply=apply,
+        )
+        MIGRATIONS.append(migration)
+        try:
+            from paperless_ai import vector_store as vs_mod
+
+            original = vs_mod.SCHEMA_VERSION
+            vs_mod.SCHEMA_VERSION = 2
+            result = store.check_and_run_migrations()
+        finally:
+            MIGRATIONS.remove(migration)
+            vs_mod.SCHEMA_VERSION = original
+
+        assert result is False
+        assert self._schema_version(store) == 2
+        ids = {n.node_id for n in store.get_nodes()}
+        assert ids == {"a1", "b1"}
+
+    def test_compact_preserves_schema_version(self, store) -> None:
+        store.add([make_node("a1", "1")])
+        assert self._schema_version(store) == SCHEMA_VERSION
+        store.compact(force=True)
+        assert self._schema_version(store) == SCHEMA_VERSION
+
+    def test_stop_at_reembed_boundary(self, store) -> None:
+        # Registry: structural v2, re-embed v3, structural v4.
+        # Only v2 should apply; the re-embed boundary must stop execution
+        # before v4 runs, and the stored version must stay at 2.
+        store.add([make_node("a1", "1"), make_node("b1", "2")])
+
+        def copy_apply(
+            src: sqlite3.Connection,
+            dst: sqlite3.Connection,
+            dim: int,
+        ) -> None:
+            dst.execute(  # nosemgrep
+                f"CREATE VIRTUAL TABLE {DEFAULT_TABLE_NAME} USING vec0("
+                "id TEXT PRIMARY KEY, document_id TEXT, modified TEXT,"
+                f" +node_content TEXT, embedding float[{dim}] distance_metric=cosine"
+                ")",
+            )
+            dst.execute(
+                "INSERT INTO index_meta (key, value) VALUES ('dim', ?) "
+                "ON CONFLICT(key) DO UPDATE SET value = excluded.value",
+                (str(dim),),
+            )
+            rows = src.execute(
+                "SELECT id, document_id, modified, node_content, embedding "
+                f"FROM {DEFAULT_TABLE_NAME}",
+            ).fetchall()
+            dst.execute("BEGIN IMMEDIATE")
+            dst.executemany(
+                f"INSERT INTO {DEFAULT_TABLE_NAME} "
+                "(id, document_id, modified, node_content, embedding) "
+                "VALUES (?, ?, ?, ?, ?)",
+                [
+                    (
+                        r["id"],
+                        r["document_id"],
+                        r["modified"],
+                        r["node_content"],
+                        bytes(r["embedding"]),
+                    )
+                    for r in rows
+                ],
+            )
+            dst.execute("COMMIT")
+
+        migrations = [
+            Migration(
+                from_version=1,
+                to_version=2,
+                kind="structural",
+                description="v2 structural",
+                apply=copy_apply,
+            ),
+            Migration(
+                from_version=2,
+                to_version=3,
+                kind="re-embed",
+                description="v3 re-embed boundary",
+            ),
+            Migration(
+                from_version=3,
+                to_version=4,
+                kind="structural",
+                description="v4 structural - must not run",
+                apply=copy_apply,
+            ),
+        ]
+        MIGRATIONS.extend(migrations)
+        try:
+            from paperless_ai import vector_store as vs_mod
+
+            original = vs_mod.SCHEMA_VERSION
+            vs_mod.SCHEMA_VERSION = 4
+            result = store.check_and_run_migrations()
+        finally:
+            for m in migrations:
+                MIGRATIONS.remove(m)
+            vs_mod.SCHEMA_VERSION = original
+
+        assert result is True
+        assert self._schema_version(store) == 2

src/paperless_ai/tests/test_views_suggestions.py

@@ -0,0 +1,77 @@
+from types import SimpleNamespace
+
+import pytest
+import pytest_mock
+from django.contrib.auth.models import User
+from rest_framework.test import APIClient
+
+from documents.models import Document
+from documents.tests.factories import DocumentFactory
+
+
+@pytest.mark.django_db
+class TestSuggestionsHintWiring:
+    @pytest.fixture
+    def document(self) -> Document:
+        return DocumentFactory()  # type: ignore[return-value]
+
+    @pytest.fixture
+    def api_client(self, admin_user: User) -> APIClient:
+        client = APIClient()
+        client.force_authenticate(user=admin_user)
+        return client
+
+    def test_hints_passed_to_classifier_and_matchers(
+        self,
+        api_client: APIClient,
+        document: Document,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        hints = {
+            "tags": ["Bloodwork"],
+            "document_types": [],
+            "correspondents": [],
+            "storage_paths": [],
+        }
+        mocker.patch(
+            "documents.views.get_taxonomy_hints_for_document",
+            return_value=hints,
+        )
+        mocker.patch(
+            "documents.views.AIConfig",
+            return_value=SimpleNamespace(
+                ai_enabled=True,
+                llm_backend="ollama",
+                llm_output_language=None,
+            ),
+        )
+        # No cached suggestion -> the view reaches the classifier path.
+        mocker.patch(
+            "documents.views.get_llm_suggestion_cache",
+            return_value=None,
+        )
+        mocker.patch("documents.views.set_llm_suggestions_cache")
+        classify = mocker.patch(
+            "documents.views.get_ai_document_classification",
+            return_value={
+                "title": "Doc",
+                "tags": ["Bloodwork"],
+                "correspondents": [],
+                "document_types": [],
+                "storage_paths": [],
+                "dates": [],
+            },
+        )
+        match_tags = mocker.patch(
+            "documents.views.match_tags_by_name",
+            return_value=[],
+        )
+        mocker.patch("documents.views.match_correspondents_by_name", return_value=[])
+        mocker.patch("documents.views.match_document_types_by_name", return_value=[])
+        mocker.patch("documents.views.match_storage_paths_by_name", return_value=[])
+
+        response = api_client.get(f"/api/documents/{document.pk}/ai_suggestions/")
+
+        assert response.status_code == 200
+        assert classify.call_args.kwargs["hints"] == hints
+        assert match_tags.call_args.kwargs["hinted_names"] == {"Bloodwork"}

src/paperless_ai/vector_store.py

@@ -0,0 +1,604 @@
+import json
+import logging
+import sqlite3
+import struct
+from collections.abc import Callable
+from collections.abc import Iterator
+from collections.abc import Sequence
+from contextlib import contextmanager
+from dataclasses import dataclass
+from dataclasses import field
+from pathlib import Path
+from types import TracebackType
+from typing import Any
+from typing import Literal
+
+import sqlite_vec
+from llama_index.core.bridge.pydantic import PrivateAttr
+from llama_index.core.schema import BaseNode
+from llama_index.core.vector_stores.types import BasePydanticVectorStore
+from llama_index.core.vector_stores.types import FilterCondition
+from llama_index.core.vector_stores.types import FilterOperator
+from llama_index.core.vector_stores.types import MetadataFilter
+from llama_index.core.vector_stores.types import MetadataFilters
+from llama_index.core.vector_stores.types import VectorStoreQuery
+from llama_index.core.vector_stores.types import VectorStoreQueryResult
+from llama_index.core.vector_stores.utils import metadata_dict_to_node
+from llama_index.core.vector_stores.utils import node_to_metadata_dict
+
+logger = logging.getLogger("paperless_ai.vector_store")
+
+DB_FILENAME = "llmindex.db"
+DEFAULT_TABLE_NAME = "documents"
+
+# Current schema version. Written to index_meta at table creation and bumped
+# whenever a Migration is added to MIGRATIONS. check_and_run_migrations() uses
+# this to decide which migrations to run on an existing store.
+SCHEMA_VERSION = 1
+
+# compact(): rebuild when the cumulative rowid count exceeds this multiple of
+# the live row count. DELETEs on vec0 tables never reclaim space (upstream
+# asg017/sqlite-vec#54), so per-document re-index churn grows the file until
+# a rebuild copies the live rows into a fresh table.
+COMPACT_BLOAT_RATIO = 2.0
+
+# Filterable vec0 metadata columns. _build_where() only ever receives filter
+# keys we construct ourselves, but allowlisting keeps SQL identifiers safe by
+# construction.
+_FILTER_COLUMNS = frozenset({"document_id", "modified"})
+
+
+@dataclass
+class Migration:
+    """A schema migration for the sqlite-vec vector store.
+
+    kind="structural": rows are copied into a new-schema file with no
+    re-embedding needed.  Supply ``apply(src_conn, dst_conn, dim)`` which
+    must create the vec0 table in ``dst_conn``, copy all rows from
+    ``src_conn``, and write ``dim`` / ``embed_model`` / ``total_inserts`` to
+    ``dst_conn``'s ``index_meta``.  ``schema_version`` is written by the
+    migration runner after ``apply`` returns.
+
+    kind="re-embed": the new schema requires fresh embeddings.
+    ``check_and_run_migrations()`` returns True when it encounters one of
+    these so the caller can force a full rebuild (which recreates the table
+    at the current SCHEMA_VERSION).
+    """
+
+    from_version: int
+    to_version: int
+    kind: Literal["structural", "re-embed"]
+    description: str
+    apply: Callable[[sqlite3.Connection, sqlite3.Connection, int], None] | None = field(
+        default=None,
+        repr=False,
+    )
+
+
+# Registry of all schema migrations in order. Empty at v1 -- this is the
+# baseline. Add entries here (and bump SCHEMA_VERSION) when the schema changes.
+MIGRATIONS: list[Migration] = []
+
+
+def _pack(embedding: Sequence[float]) -> bytes:
+    return struct.pack(f"{len(embedding)}f", *embedding)
+
+
+def _unpack(blob: bytes) -> list[float]:
+    return list(struct.unpack(f"{len(blob) // 4}f", blob))
+
+
+def _build_where(filters: MetadataFilters | None) -> tuple[str, list[str]]:
+    """Translate the EQ / IN filters we use into a parameterized SQL clause
+    on vec0 metadata columns. Returns ("", []) when there is nothing to filter.
+    """
+    if filters is None or not filters.filters:
+        return "", []
+    clauses: list[str] = []
+    params: list[str] = []
+    for f in filters.filters:
+        # filters.filters is Union[MetadataFilter, ExactMatchFilter, MetadataFilters];
+        # we only build MetadataFilter entries, so skip anything else at runtime.
+        if not isinstance(f, MetadataFilter):
+            continue
+        if f.key not in _FILTER_COLUMNS:  # pragma: no cover - we build the keys
+            raise NotImplementedError(f"Unsupported filter column: {f.key}")
+        if f.operator == FilterOperator.IN:
+            values = [str(v) for v in f.value]  # type: ignore[union-attr]  # value is list when operator is IN
+            if not values:  # pragma: no cover
+                clauses.append("1 = 0")
+                continue
+            placeholders = ",".join("?" for _ in values)
+            clauses.append(f"{f.key} IN ({placeholders})")
+            params.extend(values)
+        elif f.operator == FilterOperator.EQ:
+            clauses.append(f"{f.key} = ?")
+            params.append(str(f.value))
+        else:  # pragma: no cover - we only ever build EQ/IN filters
+            raise NotImplementedError(f"Unsupported filter operator: {f.operator}")
+    if not clauses:
+        # Filters were requested but none could be translated. Fail closed
+        # rather than emit "()" (invalid SQL): filters scope document access,
+        # so an empty translation must match no rows, never widen the scope.
+        return "1 = 0", []
+    joiner = " OR " if filters.condition == FilterCondition.OR else " AND "
+    return "(" + joiner.join(clauses) + ")", params
+
+
+class PaperlessSqliteVecVectorStore(BasePydanticVectorStore):
+    """A llama-index vector store backed by a sqlite-vec vec0 table.
+
+    Stores one row per node: the node id (TEXT primary key), its document id
+    (metadata column, used for EQ/IN filtering and per-document delete), the
+    document's modified timestamp, the embedding (float32, cosine metric), and
+    the serialized node (text + metadata) as JSON in an auxiliary column.
+    ``stores_text`` lets llama-index run off this store alone, with no
+    separate docstore or index store.
+
+    Everything lives in one SQLite database file (``DB_FILENAME``) inside the
+    directory given as ``uri`` (kept as a directory for compatibility with the
+    previous LanceDB layout). WAL mode allows readers in other processes to
+    proceed while the (FileLock-serialized) writer holds a transaction.
+
+    Implemented surface of ``BasePydanticVectorStore``
+    ---------------------------------------------------
+    Only the methods actively used by this codebase are implemented.
+    ``delete_nodes`` and the ``node_ids`` lookup path of ``get_nodes`` are
+    part of the llama-index interface contract and may be needed if a future
+    retriever or extension invokes them — add them then, with tests.
+    """
+
+    stores_text: bool = True
+    flat_metadata: bool = False
+
+    _uri: str = PrivateAttr()
+    _embed_model_name: str | None = PrivateAttr()
+    _conn: Any = PrivateAttr()
+
+    def __init__(
+        self,
+        uri: str,
+        embed_model_name: str | None = None,
+    ) -> None:
+        super().__init__(stores_text=True, flat_metadata=False)
+        self._uri = uri
+        self._embed_model_name = embed_model_name
+        self._conn = self._open_connection(str(Path(uri) / DB_FILENAME))
+
+    @staticmethod
+    def _open_connection(db_path: str) -> sqlite3.Connection:
+        conn = sqlite3.connect(
+            db_path,
+            timeout=30,
+            isolation_level=None,  # autocommit; explicit transactions below
+        )
+        conn.row_factory = sqlite3.Row
+        conn.enable_load_extension(True)  # noqa: FBT003
+        sqlite_vec.load(conn)
+        conn.enable_load_extension(False)  # noqa: FBT003
+        conn.execute("PRAGMA journal_mode=WAL")
+        conn.execute("PRAGMA synchronous=NORMAL")
+        conn.execute(
+            "CREATE TABLE IF NOT EXISTS index_meta (key TEXT PRIMARY KEY, value TEXT)",
+        )
+        return conn
+
+    @property
+    def client(self) -> Any:
+        return self._conn
+
+    def close(self) -> None:
+        """Close the underlying SQLite connection (idempotent)."""
+        self._conn.close()
+
+    def __enter__(self) -> "PaperlessSqliteVecVectorStore":
+        return self
+
+    def __exit__(
+        self,
+        exc_type: type[BaseException] | None,
+        exc_val: BaseException | None,
+        exc_tb: TracebackType | None,
+    ) -> None:
+        # Deterministically release the connection (and its WAL/SHM handles) so
+        # it is never left open across a compaction/migration file swap.
+        self.close()
+
+    @contextmanager
+    def _transaction(self) -> Iterator[None]:
+        self._conn.execute("BEGIN IMMEDIATE")
+        try:
+            yield
+        except BaseException:  # pragma: no cover
+            self._conn.execute("ROLLBACK")
+            raise
+        else:
+            self._conn.execute("COMMIT")
+
+    def _meta_get(self, key: str) -> str | None:
+        row = self._conn.execute(
+            "SELECT value FROM index_meta WHERE key = ?",
+            (key,),
+        ).fetchone()
+        return row["value"] if row else None
+
+    @staticmethod
+    def _meta_set_on(conn: sqlite3.Connection, key: str, value: str) -> None:
+        conn.execute(
+            "INSERT INTO index_meta (key, value) VALUES (?, ?) "
+            "ON CONFLICT(key) DO UPDATE SET value = excluded.value",
+            (key, value),
+        )
+
+    def _meta_set(self, key: str, value: str) -> None:
+        self._meta_set_on(self._conn, key, value)
+
+    def table_exists(self) -> bool:
+        return (
+            self._conn.execute(
+                "SELECT 1 FROM sqlite_master WHERE type = 'table' AND name = ?",
+                (DEFAULT_TABLE_NAME,),
+            ).fetchone()
+            is not None
+        )
+
+    def vector_dim(self) -> int | None:
+        if not self.table_exists():
+            return None
+        value = self._meta_get("dim")
+        return int(value) if value else None
+
+    def drop_table(self) -> None:
+        self._conn.execute("DROP TABLE IF EXISTS " + DEFAULT_TABLE_NAME)
+        self._conn.execute("DELETE FROM index_meta")
+
+    def stored_model_name(self) -> str | None:
+        """Return the embedding model name recorded at table creation, or None."""
+        if not self.table_exists():
+            return None
+        return self._meta_get("embed_model")
+
+    def config_mismatch(self, model_name: str) -> bool:
+        """True when the stored model name differs from ``model_name``.
+
+        Returns False when no table exists or when the table predates
+        model-name tracking — conservative default avoids spurious rebuilds.
+        """
+        stored = self.stored_model_name()
+        if stored is None:
+            return False
+        return stored != model_name
+
+    @staticmethod
+    def _create_vec_table(conn: sqlite3.Connection, dim: int) -> None:
+        # document_id is deliberately a metadata column, NOT a partition key:
+        # partition keys change KNN `k` to per-partition semantics under IN
+        # filters (asg017/sqlite-vec#142); metadata columns give a correct
+        # global top-k.
+        conn.execute(  # nosemgrep: python.sqlalchemy.security.sqlalchemy-execute-raw-query.sqlalchemy-execute-raw-query
+            "CREATE VIRTUAL TABLE "
+            + DEFAULT_TABLE_NAME
+            + " USING vec0("
+            + "id TEXT PRIMARY KEY,"
+            + " document_id TEXT,"
+            + " modified TEXT,"
+            + " +node_content TEXT,"
+            + " embedding float["
+            + str(int(dim))
+            + "] distance_metric=cosine"
+            + ")",
+        )
+
+    def _create_table(self, dim: int) -> None:
+        self._create_vec_table(self._conn, dim)
+        self._meta_set("dim", str(dim))
+        self._meta_set("schema_version", str(SCHEMA_VERSION))
+        if self._embed_model_name:
+            self._meta_set("embed_model", self._embed_model_name)
+
+    def _ensure_table(self, dim: int) -> None:
+        if not self.table_exists():
+            self._create_table(dim)
+
+    def _row(self, node: BaseNode) -> tuple[str, str, str, str, bytes]:
+        meta = node_to_metadata_dict(
+            node,
+            remove_text=False,
+            flat_metadata=self.flat_metadata,
+        )
+        # vec0 metadata columns reject NULL (asg017/sqlite-vec#141): coerce
+        # every value to a string, with "" as the absent sentinel.
+        document_id = node.ref_doc_id or node.metadata.get("document_id")
+        return (
+            node.node_id,
+            str(document_id or ""),
+            str(node.metadata.get("modified") or ""),
+            json.dumps(meta),
+            _pack(node.get_embedding()),
+        )
+
+    _INSERT = (
+        "INSERT INTO "
+        + DEFAULT_TABLE_NAME
+        + " (id, document_id, modified, node_content, embedding) VALUES (?, ?, ?, ?, ?)"
+    )
+
+    def _increment_total_inserts(self, count: int) -> None:
+        """Increment the cumulative insert counter stored in index_meta.
+
+        This counter never decreases (DELETEs do not decrement it) and is
+        used by compact() to estimate the bloat ratio: when total_inserts /
+        live_rows exceeds COMPACT_BLOAT_RATIO the table has accumulated
+        enough deleted-but-not-freed rows to warrant a rebuild.
+        """
+        current = int(self._meta_get("total_inserts") or "0")
+        self._meta_set("total_inserts", str(current + count))
+
+    def add(self, nodes: Sequence[BaseNode], **add_kwargs: Any) -> list[str]:
+        if not nodes:
+            return []
+        rows = [self._row(node) for node in nodes]
+        with self._transaction():
+            self._ensure_table(len(nodes[0].get_embedding()))
+            self._conn.executemany(self._INSERT, rows)
+            self._increment_total_inserts(len(rows))
+        return [node.node_id for node in nodes]
+
+    def upsert_document(self, document_id: str, nodes: list[BaseNode]) -> list[str]:
+        """Atomically replace all stored chunks of ``document_id`` with ``nodes``.
+
+        One transaction deletes the document's existing rows and inserts the
+        new set (vec0's INSERT OR REPLACE is broken upstream, #259, so
+        delete+insert it is). WAL readers in other processes see either the
+        old or the new chunk set, never a partial state.
+        """
+        rows = [self._row(node) for node in nodes]
+        with self._transaction():
+            if nodes:
+                self._ensure_table(len(nodes[0].get_embedding()))
+            if self.table_exists():
+                self._conn.execute(
+                    "DELETE FROM " + DEFAULT_TABLE_NAME + " WHERE document_id = ?",
+                    (str(document_id),),
+                )
+            if rows:
+                self._conn.executemany(self._INSERT, rows)
+                self._increment_total_inserts(len(rows))
+        return [node.node_id for node in nodes]
+
+    def delete(self, ref_doc_id: str, **delete_kwargs: Any) -> None:
+        if self.table_exists():
+            with self._transaction():
+                self._conn.execute(
+                    "DELETE FROM " + DEFAULT_TABLE_NAME + " WHERE document_id = ?",
+                    (str(ref_doc_id),),
+                )
+
+    def _rows_to_nodes(self, rows: list[sqlite3.Row]) -> list[BaseNode]:
+        nodes: list[BaseNode] = []
+        for row in rows:
+            node = metadata_dict_to_node(json.loads(row["node_content"]))
+            node.embedding = _unpack(row["embedding"])
+            nodes.append(node)
+        return nodes
+
+    def get_nodes(
+        self,
+        node_ids: list[str] | None = None,
+        filters: MetadataFilters | None = None,
+        **kwargs: Any,
+    ) -> list[BaseNode]:
+        if node_ids is not None:  # pragma: no cover
+            # node_ids lookup is not implemented; see class docstring.
+            raise NotImplementedError(
+                "PaperlessSqliteVecVectorStore does not support node_ids lookup",
+            )
+        if not self.table_exists():
+            return []
+        where, params = _build_where(filters)
+        sql = "SELECT node_content, embedding FROM " + DEFAULT_TABLE_NAME
+        if where:
+            sql += " WHERE " + where
+        return self._rows_to_nodes(self._conn.execute(sql, params).fetchall())
+
+    def query(
+        self,
+        query: VectorStoreQuery,
+        **kwargs: Any,
+    ) -> VectorStoreQueryResult:
+        if not self.table_exists():
+            return VectorStoreQueryResult(nodes=[], similarities=[], ids=[])
+        if query.query_embedding is None:  # pragma: no cover
+            return VectorStoreQueryResult(nodes=[], similarities=[], ids=[])
+        top_k = query.similarity_top_k if query.similarity_top_k is not None else 10
+        where, params = _build_where(query.filters)
+        sql = (
+            "SELECT id, node_content, embedding, distance FROM "
+            + DEFAULT_TABLE_NAME
+            + " WHERE embedding MATCH ? AND k = ?"
+        )
+        if where:
+            sql += " AND " + where
+        rows = self._conn.execute(
+            sql,
+            [_pack(query.query_embedding), top_k, *params],
+        ).fetchall()
+        # vec0 returns rows distance-sorted ascending; slice defensively in
+        # case future schema changes alter k semantics (e.g. partition keys
+        # return k rows per partition).
+        rows = rows[:top_k]
+        nodes = self._rows_to_nodes(rows)
+        # Cosine distance in [0, 2]; map to a descending similarity.
+        # vec0 returns None distance when the query embedding is the zero vector
+        # (no meaningful cosine angle); treat that as maximum distance (1.0) so
+        # the row is included but ranked last.
+        sims = [
+            1.0 - float(row["distance"] if row["distance"] is not None else 1.0)
+            for row in rows
+        ]
+        ids = [row["id"] for row in rows]
+        return VectorStoreQueryResult(nodes=nodes, similarities=sims, ids=ids)
+
+    def get_modified_times(self) -> dict[str, str]:
+        """Return {document_id: stored_modified_isoformat} for all indexed documents.
+
+        All chunks of a document share the same ``modified`` value, so the
+        first row seen per document is sufficient.
+        """
+        if not self.table_exists():
+            return {}
+        result: dict[str, str] = {}
+        for row in self._conn.execute(
+            "SELECT document_id, modified FROM " + DEFAULT_TABLE_NAME,
+        ):
+            doc_id = str(row["document_id"])
+            if doc_id not in result:
+                result[doc_id] = str(row["modified"] or "")
+        return result
+
+    def compact(self, *, force: bool = False) -> None:
+        """Rebuild the database file to reclaim space left behind by DELETEs.
+
+        vec0 DELETE only invalidates rows; the vector data stays in the file
+        forever (asg017/sqlite-vec#54), and per-document re-indexing is a
+        delete+insert. The cumulative insert counter in ``index_meta`` tracks
+        total rows ever written; when that exceeds ``COMPACT_BLOAT_RATIO`` x
+        the live row count (or when forced), live rows are copied into a fresh
+        database file and swapped in via ``os.replace``.
+
+        Note: ``ALTER TABLE ... RENAME TO`` on vec0 virtual tables does NOT
+        rename the shadow tables (sqlite-vec upstream limitation), so
+        an in-place rename-based rebuild is not safe.  The file-swap approach
+        is the maintainer-endorsed workaround (asg017/sqlite-vec#205).
+        """
+        if not self.table_exists():
+            return
+        live = self._conn.execute(
+            "SELECT count(*) FROM " + DEFAULT_TABLE_NAME,
+        ).fetchone()[0]
+        total = int(self._meta_get("total_inserts") or str(live))
+        if not force and total <= max(live, 1) * COMPACT_BLOAT_RATIO:
+            return
+        dim = self.vector_dim()
+        if dim is None:  # pragma: no cover - dim is written at creation
+            logger.warning("Skipping compact: no stored vector dimension")
+            return
+        logger.info(
+            "Compacting LLM index (%d live rows, %d cumulative inserts)",
+            live,
+            total,
+        )
+        db_path = str(Path(self._uri) / DB_FILENAME)
+        compact_path = db_path + ".compact"
+
+        # Copy all live rows into a fresh database file.
+        new_conn = self._open_connection(compact_path)
+        try:
+            self._create_vec_table(new_conn, dim)
+            self._meta_set_on(new_conn, "dim", str(dim))
+            for key in ("embed_model", "schema_version"):
+                value = self._meta_get(key)
+                if value is not None:
+                    self._meta_set_on(new_conn, key, value)
+            rows = self._conn.execute(
+                "SELECT id, document_id, modified, node_content, embedding "
+                "FROM " + DEFAULT_TABLE_NAME,
+            ).fetchall()
+            new_conn.execute("BEGIN IMMEDIATE")
+            new_conn.executemany(
+                self._INSERT,
+                [
+                    (
+                        r["id"],
+                        r["document_id"],
+                        r["modified"],
+                        r["node_content"],
+                        bytes(r["embedding"]),
+                    )
+                    for r in rows
+                ],
+            )
+            # Reset the cumulative counter: after compact, total_inserts == live.
+            self._meta_set_on(new_conn, "total_inserts", str(live))
+            new_conn.execute("COMMIT")
+        except BaseException:
+            new_conn.close()
+            for p in [compact_path, compact_path + "-wal", compact_path + "-shm"]:
+                Path(p).unlink(missing_ok=True)
+            raise
+        new_conn.close()
+        self._swap_in_compact(compact_path, db_path)
+
+    def _swap_in_compact(self, compact_path: str, db_path: str) -> None:
+        """Atomically replace the live database with the compacted copy."""
+        self._conn.close()
+        for suffix in ["-wal", "-shm"]:
+            stale = Path(compact_path + suffix)
+            if stale.exists():  # pragma: no cover
+                stale.unlink()
+        Path(compact_path).replace(db_path)
+        self._conn = self._open_connection(db_path)
+
+    def check_and_run_migrations(self) -> bool:
+        """Apply any pending schema migrations to the store.
+
+        Structural migrations copy live rows into a new-schema file with no
+        re-embedding.  Re-embed migrations cannot be applied automatically;
+        this method returns True when one is encountered so the caller can
+        force a full rebuild (which recreates the table at SCHEMA_VERSION).
+
+        Must be called under the write FileLock.  No-op when the table does
+        not exist or is already at SCHEMA_VERSION.
+        """
+        if not self.table_exists():
+            return False
+
+        raw = self._meta_get("schema_version")
+        current = int(raw) if raw is not None else SCHEMA_VERSION
+        if current >= SCHEMA_VERSION:
+            return False
+
+        pending = sorted(
+            [m for m in MIGRATIONS if current <= m.from_version < SCHEMA_VERSION],
+            key=lambda m: m.from_version,
+        )
+
+        for migration in pending:
+            if migration.kind == "re-embed":
+                logger.warning(
+                    "LLM index schema v%d -> v%d requires re-embedding (%s); "
+                    "forcing full rebuild.",
+                    migration.from_version,
+                    migration.to_version,
+                    migration.description,
+                )
+                return True
+            logger.info(
+                "Running structural LLM index migration v%d -> v%d: %s",
+                migration.from_version,
+                migration.to_version,
+                migration.description,
+            )
+            self._run_structural_migration(migration)
+
+        return False
+
+    def _run_structural_migration(self, migration: Migration) -> None:
+        """Execute a structural migration using the same file-swap as compact()."""
+        assert migration.apply is not None, "structural migration must have apply()"
+        dim = self.vector_dim()
+        if dim is None:  # pragma: no cover
+            raise RuntimeError("Cannot migrate: no stored vector dimension")
+        db_path = str(Path(self._uri) / DB_FILENAME)
+        compact_path = db_path + ".compact"
+        new_conn = self._open_connection(compact_path)
+        try:
+            migration.apply(self._conn, new_conn, dim)
+            self._meta_set_on(new_conn, "schema_version", str(migration.to_version))
+        except BaseException:  # pragma: no cover
+            new_conn.close()
+            for p in [compact_path, compact_path + "-wal", compact_path + "-shm"]:
+                Path(p).unlink(missing_ok=True)
+            raise
+        new_conn.close()
+        self._swap_in_compact(compact_path, db_path)

src/paperless_mail/mail.py

@@ -4,6 +4,7 @@ import logging
 import ssl
 import tempfile
 import traceback
+import unicodedata
 from datetime import date
 from datetime import timedelta
 from fnmatch import fnmatch
@@ -496,10 +497,10 @@ class MailAccountHandler(LoggingMixin):
         rule: MailRule,
     ) -> str | None:
         if rule.assign_title_from == MailRule.TitleSource.FROM_SUBJECT:
-            return message.subject
+            return unicodedata.normalize("NFC", message.subject)
 
         elif rule.assign_title_from == MailRule.TitleSource.FROM_FILENAME:
-            return Path(att.filename).stem
+            return unicodedata.normalize("NFC", Path(att.filename).stem)
 
         elif rule.assign_title_from == MailRule.TitleSource.NONE:
             return None
@@ -866,7 +867,9 @@ class MailAccountHandler(LoggingMixin):
                     ),
                 )
 
-                attachment_name = pathvalidate.sanitize_filename(att.filename)
+                attachment_name = pathvalidate.sanitize_filename(
+                    unicodedata.normalize("NFC", att.filename),
+                )
                 if attachment_name:
                     temp_filename = temp_dir / attachment_name
                 else:  # pragma: no cover
@@ -882,7 +885,7 @@ class MailAccountHandler(LoggingMixin):
                 )
                 doc_overrides = DocumentMetadataOverrides(
                     title=title,
-                    filename=pathvalidate.sanitize_filename(att.filename),
+                    filename=attachment_name,
                     correspondent_id=correspondent.id if correspondent else None,
                     document_type_id=doc_type.id if doc_type else None,
                     tag_ids=tag_ids,
@@ -988,7 +991,9 @@ class MailAccountHandler(LoggingMixin):
         )
         doc_overrides = DocumentMetadataOverrides(
             title=message.subject,
-            filename=pathvalidate.sanitize_filename(f"{message.subject}.eml"),
+            filename=pathvalidate.sanitize_filename(
+                unicodedata.normalize("NFC", f"{message.subject}.eml"),
+            ),
             correspondent_id=correspondent.id if correspondent else None,
             document_type_id=doc_type.id if doc_type else None,
             tag_ids=tag_ids,

src/paperless_mail/tests/test_mail_nfc.py

@@ -0,0 +1,182 @@
+"""
+Tests that mail attachment filenames and EML subject filenames are
+normalized to NFC Unicode before being stored as document overrides.
+
+Filenames from MIME headers can arrive in NFD form (e.g. from macOS Mail),
+and must be normalized to NFC so filenames are consistent regardless of the
+sending client.
+"""
+
+import unicodedata
+from pathlib import Path
+from unittest import mock
+
+import pytest
+
+from documents.tests.utils import remove_dirs
+from documents.tests.utils import setup_directories
+from paperless_mail.models import MailRule
+from paperless_mail.tests.factories import MailAccountFactory
+from paperless_mail.tests.test_mail import MessageBuilder
+from paperless_mail.tests.test_mail import _AttachmentDef
+from paperless_mail.tests.test_mail import fake_magic_from_buffer
+
+
+@pytest.fixture()
+def directories(settings):
+    dirs = setup_directories()
+    yield dirs
+    remove_dirs(dirs)
+
+
+@pytest.fixture()
+def queue_consumption_tasks_mock():
+    with mock.patch("paperless_mail.mail.queue_consumption_tasks") as m:
+        yield m
+
+
+@pytest.fixture()
+def mail_account(db):
+    return MailAccountFactory()
+
+
+@pytest.fixture()
+def attachment_rule(mail_account):
+    rule = MailRule(
+        name="attachment rule",
+        account=mail_account,
+        assign_title_from=MailRule.TitleSource.FROM_FILENAME,
+        consumption_scope=MailRule.ConsumptionScope.ATTACHMENTS_ONLY,
+        attachment_type=MailRule.AttachmentProcessing.ATTACHMENTS_ONLY,
+    )
+    rule.save()
+    return rule
+
+
+@pytest.fixture()
+def eml_rule(mail_account):
+    rule = MailRule(
+        name="eml rule",
+        account=mail_account,
+        assign_title_from=MailRule.TitleSource.FROM_SUBJECT,
+        consumption_scope=MailRule.ConsumptionScope.EML_ONLY,
+        attachment_type=MailRule.AttachmentProcessing.ATTACHMENTS_ONLY,
+    )
+    rule.save()
+    return rule
+
+
+@pytest.fixture()
+def message_builder():
+    return MessageBuilder()
+
+
+@pytest.mark.django_db
+@mock.patch("paperless_mail.mail.magic.from_buffer", fake_magic_from_buffer)
+class TestMailNFCNormalization:
+    """Attachment filenames and EML subject filenames must be NFC-normalized."""
+
+    def test_attachment_nfd_filename_normalized_to_nfc(
+        self,
+        directories,
+        queue_consumption_tasks_mock,
+        attachment_rule,
+        mail_account_handler,
+        message_builder,
+    ):
+        """Attachment filename arriving as NFD must be stored as NFC in both
+        the overrides and the temp file written to disk.
+        """
+        nfd_filename = unicodedata.normalize("NFD", "Rechnung März.pdf")
+        nfc_filename = unicodedata.normalize("NFC", "Rechnung März.pdf")
+
+        # Confirm the fixture is actually NFD (not already NFC)
+        assert unicodedata.is_normalized("NFD", nfd_filename)
+        assert not unicodedata.is_normalized("NFC", nfd_filename)
+
+        message = message_builder.create_message(
+            subject="Test invoice",
+            from_="sender@example.com",
+            attachments=[
+                _AttachmentDef(filename=nfd_filename, content=b"%PDF-1.4 test"),
+            ],
+        )
+
+        result = mail_account_handler._handle_message(message, attachment_rule)
+
+        assert result == 1
+        queue_consumption_tasks_mock.assert_called_once()
+
+        call_kwargs = queue_consumption_tasks_mock.call_args.kwargs
+        consume_tasks = call_kwargs["consume_tasks"]
+        assert len(consume_tasks) == 1
+
+        overrides = consume_tasks[0].kwargs["overrides"]
+        assert overrides.filename == nfc_filename
+        assert unicodedata.is_normalized("NFC", overrides.filename)
+        assert unicodedata.is_normalized("NFC", overrides.title)
+
+        input_doc = consume_tasks[0].kwargs["input_doc"]
+        original_file = Path(input_doc.original_file)
+        assert original_file.exists()
+        assert original_file.name == nfc_filename
+
+    def test_eml_subject_filename_nfc(
+        self,
+        directories,
+        queue_consumption_tasks_mock,
+        eml_rule,
+        mail_account_handler,
+        message_builder,
+    ):
+        """EML filename derived from subject arriving as NFD must be stored as NFC."""
+        nfd_subject = unicodedata.normalize("NFD", "Rechnung März 2024")
+        nfc_expected_filename = unicodedata.normalize("NFC", "Rechnung März 2024.eml")
+
+        # Confirm the fixture is actually NFD
+        assert unicodedata.is_normalized("NFD", nfd_subject)
+
+        message = message_builder.create_message(
+            subject=nfd_subject,
+            from_="sender@example.com",
+            attachments=0,
+        )
+
+        mail_account_handler._handle_message(message, eml_rule)
+
+        queue_consumption_tasks_mock.assert_called_once()
+
+        call_kwargs = queue_consumption_tasks_mock.call_args.kwargs
+        consume_tasks = call_kwargs["consume_tasks"]
+        assert len(consume_tasks) == 1
+
+        overrides = consume_tasks[0].kwargs["overrides"]
+        assert overrides.filename == nfc_expected_filename
+        assert unicodedata.is_normalized("NFC", overrides.filename)
+
+    def test_already_nfc_attachment_filename_unchanged(
+        self,
+        directories,
+        queue_consumption_tasks_mock,
+        attachment_rule,
+        mail_account_handler,
+        message_builder,
+    ):
+        """An attachment filename already in NFC must pass through unchanged."""
+        nfc_filename = "Invoice_2024.pdf"
+        assert unicodedata.is_normalized("NFC", nfc_filename)
+
+        message = message_builder.create_message(
+            subject="Invoice",
+            from_="sender@example.com",
+            attachments=[
+                _AttachmentDef(filename=nfc_filename, content=b"%PDF-1.4 test"),
+            ],
+        )
+
+        mail_account_handler._handle_message(message, attachment_rule)
+
+        call_kwargs = queue_consumption_tasks_mock.call_args.kwargs
+        consume_tasks = call_kwargs["consume_tasks"]
+        overrides = consume_tasks[0].kwargs["overrides"]
+        assert overrides.filename == nfc_filename

uv.lock

@@ -1200,23 +1200,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/27/8d/2bc5f5546ff2ccb3f7de06742853483ab75bf74f36a92254702f8baecc79/factory_boy-3.3.3-py2.py3-none-any.whl", hash = "sha256:1c39e3289f7e667c4285433f305f8d506efc2fe9c73aaea4151ebd5cdea394fc", size = 37036, upload-time = "2025-02-03T09:49:01.659Z" },
 ]
 
-[[package]]
-name = "faiss-cpu"
-version = "1.13.2"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "numpy", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
-    { name = "packaging", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
-]
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/07/c9/671f66f6b31ec48e5825d36435f0cb91189fa8bb6b50724029dbff4ca83c/faiss_cpu-1.13.2-cp310-abi3-macosx_14_0_arm64.whl", hash = "sha256:a9064eb34f8f64438dd5b95c8f03a780b1a3f0b99c46eeacb1f0b5d15fc02dc1", size = 3452776, upload-time = "2025-12-24T10:27:01.419Z" },
-    { url = "https://files.pythonhosted.org/packages/5a/4a/97150aa1582fb9c2bca95bd8fc37f27d3b470acec6f0a6833844b21e4b40/faiss_cpu-1.13.2-cp310-abi3-macosx_14_0_x86_64.whl", hash = "sha256:c8d097884521e1ecaea6467aeebbf1aa56ee4a36350b48b2ca6b39366565c317", size = 7896434, upload-time = "2025-12-24T10:27:03.592Z" },
-    { url = "https://files.pythonhosted.org/packages/0b/d0/0940575f059591ca31b63a881058adb16a387020af1709dcb7669460115c/faiss_cpu-1.13.2-cp310-abi3-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:0ee330a284042c2480f2e90450a10378fd95655d62220159b1408f59ee83ebf1", size = 11485825, upload-time = "2025-12-24T10:27:05.681Z" },
-    { url = "https://files.pythonhosted.org/packages/e7/e1/a5acac02aa593809f0123539afe7b4aff61d1db149e7093239888c9053e1/faiss_cpu-1.13.2-cp310-abi3-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:ab88ee287c25a119213153d033f7dd64c3ccec466ace267395872f554b648cd7", size = 23845772, upload-time = "2025-12-24T10:27:08.194Z" },
-    { url = "https://files.pythonhosted.org/packages/9c/7b/49dcaf354834ec457e85ca769d50bc9b5f3003fab7c94a9dcf08cf742793/faiss_cpu-1.13.2-cp310-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:85511129b34f890d19c98b82a0cd5ffb27d89d1cec2ee41d2621ee9f9ef8cf3f", size = 13477567, upload-time = "2025-12-24T10:27:10.822Z" },
-    { url = "https://files.pythonhosted.org/packages/f7/6b/12bb4037921c38bb2c0b4cfc213ca7e04bbbebbfea89b0b5746248ce446e/faiss_cpu-1.13.2-cp310-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:8b32eb4065bac352b52a9f5ae07223567fab0a976c7d05017c01c45a1c24264f", size = 25102239, upload-time = "2025-12-24T10:27:13.476Z" },
-]
-
 [[package]]
 name = "faker"
 version = "40.15.0"
@@ -2280,18 +2263,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/f4/0c/fdddaee5391d915d3d568d2d8dbdb7c95647e65bb94d4ddb31d47cef5daf/llama_index_llms_openai_like-0.7.2-py3-none-any.whl", hash = "sha256:1f45a7b1cec8fb3f5997684327ffe6c19f93e789c2fff35dc5522465850faf0b", size = 6602, upload-time = "2026-04-23T23:05:31.708Z" },
 ]
 
-[[package]]
-name = "llama-index-vector-stores-faiss"
-version = "0.6.0"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "llama-index-core", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/7c/32/89a04e38fa9595b7116c61955d9a67085f0a5480738e9c14063e374724c2/llama_index_vector_stores_faiss-0.6.0.tar.gz", hash = "sha256:00bfeb6cb7571e0e856566cb4f10c89b415b6108f151d9ad48ee9c31da563f5e", size = 6045, upload-time = "2026-03-12T20:46:31.454Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/5b/85/465b4f199075ae7773c181b2f98cf689f3107a8de031e7a9d4cd5e906446/llama_index_vector_stores_faiss-0.6.0-py3-none-any.whl", hash = "sha256:d4600c60ef5411d9e35ba573b4f416a5e13ea04c6f942c8e6f49f03f2feb4f3b", size = 7739, upload-time = "2026-03-12T20:46:30.736Z" },
-]
-
 [[package]]
 name = "llama-index-workflows"
 version = "2.20.0"
@@ -2912,7 +2883,6 @@ dependencies = [
     { name = "drf-spectacular", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "drf-spectacular-sidecar", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "drf-writable-nested", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
-    { name = "faiss-cpu", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "filelock", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "flower", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "gotenberg-client", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
@@ -2927,7 +2897,6 @@ dependencies = [
     { name = "llama-index-embeddings-openai-like", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "llama-index-llms-ollama", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "llama-index-llms-openai-like", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
-    { name = "llama-index-vector-stores-faiss", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "nltk", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "ocrmypdf", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "openai", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
@@ -2944,6 +2913,7 @@ dependencies = [
     { name = "scikit-learn", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "sentence-transformers", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "setproctitle", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
+    { name = "sqlite-vec", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "tantivy", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "tika-client", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
     { name = "torch", version = "2.11.0", source = { registry = "https://download.pytorch.org/whl/cpu" }, marker = "sys_platform == 'darwin'" },
@@ -3062,7 +3032,6 @@ requires-dist = [
     { name = "drf-spectacular", specifier = "~=0.28" },
     { name = "drf-spectacular-sidecar", specifier = "~=2026.5.1" },
     { name = "drf-writable-nested", specifier = "~=0.7.1" },
-    { name = "faiss-cpu", specifier = ">=1.10" },
     { name = "filelock", specifier = "~=3.29.0" },
     { name = "flower", specifier = "~=2.0.1" },
     { name = "gotenberg-client", specifier = "~=0.14.0" },
@@ -3078,7 +3047,6 @@ requires-dist = [
     { name = "llama-index-embeddings-openai-like", specifier = ">=0.2.2" },
     { name = "llama-index-llms-ollama", specifier = ">=0.9.1" },
     { name = "llama-index-llms-openai-like", specifier = ">=0.7.1" },
-    { name = "llama-index-vector-stores-faiss", specifier = ">=0.5.2" },
     { name = "mysqlclient", marker = "extra == 'mariadb'", specifier = "~=2.2.7" },
     { name = "nltk", specifier = "~=3.9.1" },
     { name = "ocrmypdf", specifier = "~=17.4.2" },
@@ -3101,6 +3069,7 @@ requires-dist = [
     { name = "scikit-learn", specifier = "~=1.8.0" },
     { name = "sentence-transformers", specifier = ">=5.4.1" },
     { name = "setproctitle", specifier = "~=1.3.4" },
+    { name = "sqlite-vec", specifier = "==0.1.9" },
     { name = "tantivy", specifier = "~=0.26.0" },
     { name = "tika-client", specifier = "~=0.11.0" },
     { name = "torch", specifier = "~=2.11.0", index = "https://download.pytorch.org/whl/cpu" },
@@ -4699,6 +4668,17 @@ asyncio = [
     { name = "greenlet", marker = "sys_platform == 'darwin' or sys_platform == 'linux'" },
 ]
 
+[[package]]
+name = "sqlite-vec"
+version = "0.1.9"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/68/85/9fad0045d8e7c8df3e0fa5a56c630e8e15ad6e5ca2e6106fceb666aa6638/sqlite_vec-0.1.9-py3-none-macosx_10_6_x86_64.whl", hash = "sha256:1b62a7f0a060d9475575d4e599bbf94a13d85af896bc1ce86ee80d1b5b48e5fb", size = 131171, upload-time = "2026-03-31T08:02:31.717Z" },
+    { url = "https://files.pythonhosted.org/packages/a4/3d/3677e0cd2f92e5ebc43cd29fbf565b75582bff1ccfa0b8327c7508e1084f/sqlite_vec-0.1.9-py3-none-macosx_11_0_arm64.whl", hash = "sha256:1d52e30513bae4cc9778ddbf6145610434081be4c3afe57cd877893bad9f6b6c", size = 165434, upload-time = "2026-03-31T08:02:32.712Z" },
+    { url = "https://files.pythonhosted.org/packages/00/d4/f2b936d3bdc38eadcbd2a87875815db36430fab0363182ba5d12cd8e0b51/sqlite_vec-0.1.9-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4e921e592f24a5f9a18f590b6ddd530eb637e2d474e3b1972f9bbeb773aa3cb9", size = 160076, upload-time = "2026-03-31T08:02:33.796Z" },
+    { url = "https://files.pythonhosted.org/packages/6f/ad/6afd073b0f817b3e03f9e37ad626ae341805891f23c74b5292818f49ac63/sqlite_vec-0.1.9-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.manylinux1_x86_64.whl", hash = "sha256:1515727990b49e79bcaf75fdee2ffc7d461f8b66905013231251f1c8938e7786", size = 163388, upload-time = "2026-03-31T08:02:34.888Z" },
+]
+
 [[package]]
 name = "sqlparse"
 version = "0.5.5"