Fix: address code review feedback on table gateways

- Document sqlite3.Row row_factory precondition in module docstring - Strengthen test_create_is_idempotent to verify populated table survives - Collapse three roundtrip tests into one @pytest.mark.parametrize Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Feature: add table-gateway module for document_chunks/document_meta/index_meta
2026-07-30 23:55:59 +00:00 · 2026-07-30 14:29:34 -07:00 · 2026-07-30 14:29:34 -07:00 · 2026-07-30 14:28:21 -07:00 · 2026-07-30 08:53:01 -07:00 · 2026-07-30 14:39:47 +00:00
45 changed files with 1352 additions and 1510 deletions
@@ -1,187 +0,0 @@
---
-name: paperless-benchmarking
-description: Use when profiling paperless-ngx performance, running `manage.py benchmark`, investigating a slow query or endpoint, or deciding whether a profiling finding should become a permanent registered scenario. Covers command reference, the fork/merge-back branch workflow for perf investigations, and how to read query-plan output.
---
-
-# Paperless-ngx Benchmarking
-
-This repo has a built-in benchmarking/profiling tool: `manage.py benchmark`, in
-the `paperless_benchmark` Django app. It replaces ad hoc standalone scripts —
-use it instead of writing new one-off seed/timing scripts.
-
-## Command reference
-
-```
-manage.py benchmark seed --tier {home,medium,large} [--reset --yes-i-know-this-wipes-the-database] [--seed N]
-manage.py benchmark run --repeat 5 [--label baseline]
-manage.py benchmark profile <scenario_name> [--repeat 5] [--explain]
-manage.py benchmark list-scenarios
-```
-
- **`seed`** builds a realistic dataset at one of three scales: `home` (500
-  documents — fast, use this for iteration), `medium` (20,000 — the default
-  when `--tier` is omitted; a multi-minute seed), `large` (360,000 — matches
-  the scale reported in real large-install bug reports; slow, only use it
-  when a finding needs confirming at real scale). `--reset` wipes any
-  previously-seeded benchmark data first — but it is destructive and
-  irreversible: it deletes **all** users, **all** groups, and **all**
-  documents/tags/correspondents/document types/storage paths in the target
-  database, not just benchmark-created rows. Only run it against a disposable
-  benchmark database, never a real install. Because of that, `--reset` also
-  requires passing `--yes-i-know-this-wipes-the-database` in the same
-  invocation, or the command raises an error and does nothing. Omit both
-  flags if you want to layer more data onto an existing seed instead. `seed`
-  creates two named users, `perf_target` (mixed owned/shared documents,
-  realistic guardian permission grants) and `perf_admin` (superuser), plus a
-  general user/group pool with realistic permission-row ratios.
- **`run`** times the 3 built-in API endpoint benchmarks
-  (`/api/documents/`, `/api/documents/?page_size=50`, `/api/tags/?page_size=100000`) for both
-  `perf_target` and `perf_admin`, reporting min/median/max wall-clock and SQL
-  query count. Requires `seed` to have already run — it reuses that data, it
-  does not seed its own.
- **`profile`** times one named scenario from the registry (see
-  `list-scenarios`) via best-of-N repeat timing and SQL query count, against
-  `perf_target`. `--explain` additionally captures and prints the query plan:
-  real `EXPLAIN ANALYZE` execution stats on PostgreSQL/MariaDB, or
-  `EXPLAIN QUERY PLAN` (plan only, no real timing/row counts — clearly labeled
-  as such) on SQLite. Like `run`, `profile` requires `seed` to have already
-  run — it does not seed its own data either.
- Every `run`/`profile` invocation appends a JSON line to
-  `benchmark_results/history.jsonl` at the repo root (local-only, gitignored —
-  never commit this file). Use it to compare a `before`/`after` pair across
-  two invocations without hand-copying numbers.
- Full chain example: `seed --reset` once, then `run` and `profile` as many
-  times as you want against that same seeded data — no need to reseed between
-  them.
-
-## Adding a new scenario
-
-A "scenario" is a named, registered query/operation that `profile` can time
-and explain. To add one, edit `src/paperless_benchmark/scenarios.py`: write a
-`_<name>_run(user)` function (returns whatever `run_profile` should time) and
-optionally a `_<name>_queryset(user)` function (returns the `QuerySet` for
-`--explain` to analyze), then `register(Scenario(name=..., describe=...,
-run=..., queryset_for_explain=...))` at module level. Both functions receive
-the already-seeded `perf_target` user — they should not seed their own data.
-
-## Branch workflow
-
-`tools/benchmark-management-commands` is a **long-lived tooling branch**, not
-a feature branch that gets merged and closed:
-
-1. It is periodically brought up to date with `dev` (merge `dev` into it) so
-   the tooling doesn't drift from the schema/codebase it profiles. Do this
-   before starting a new investigation if it's been a while since the last
-   sync.
-2. **Every performance investigation forks its own branch from
-   `tools/benchmark-management-commands`** (not from `dev`). Do the
-   investigation there: write throwaway profiling code, try fixes, capture
-   before/after numbers.
-3. **That investigation branch never merges into `dev` or production.** Its
-   only job is to produce evidence and, optionally, a reusable scenario.
-4. If the investigation turns up a scenario worth keeping permanently (see
-   "When to graduate a scenario" below), open a PR that adds **just that
-   scenario** back into `tools/benchmark-management-commands` — not the rest
-   of the investigation branch's throwaway code.
-5. Any actual production fix the investigation motivates (e.g. an ORM query
-   change) goes into its own normal feature branch off `dev`, following the
-   project's regular contribution process — profiling evidence informs that
-   PR's description, but the profiling code itself does not travel with it.
-
-## Reading query-plan output
-
- **PostgreSQL** `EXPLAIN ANALYZE`: look for `Seq Scan` on a large table
-  (missing index), a large gap between `rows=N` (planner's estimate) and the
-  actual row count in parentheses (stale statistics or a bad cardinality
-  estimate), and nested-loop joins driven by an outer relation with many
-  rows (usually the N+1 pattern this tool exists to catch).
- **MariaDB**: verified against a real MariaDB 12.3 container that MariaDB
-  does NOT accept MySQL 8.0.18+'s `EXPLAIN ANALYZE` syntax (it's a 1064
-  syntax error) -- `capture_explain()` instead runs MariaDB's own
-  `ANALYZE <statement>` form (no `EXPLAIN` keyword), which returns a
-  tabular plan with real per-row execution columns: `rows` (estimate) vs.
-  `r_rows` (actual), and `filtered` vs. `r_filtered`. A large gap between
-  `rows` and `r_rows`, or `type: ALL` (full table scan) on a large table,
-  are the signals to look for -- the same underlying concerns as Postgres's
-  `Seq Scan`/estimate-vs-actual gap, just in MariaDB's column-based output
-  instead of Postgres's nested-tree text format.
- **SQLite** `EXPLAIN QUERY PLAN`: no real timing/row-count data, only the
-  chosen access path (`SCAN` vs `SEARCH`, which index if any). Useful for
-  confirming an index is even being considered, not for judging real-world
-  cost — corroborate any SQLite finding against Postgres/MariaDB before
-  trusting it, since planner behavior differs meaningfully between them.
- Compare query **count**, not just timing, between before/after: a fix that
-  keeps the same wall-clock time but drops query count from O(n) to O(1) is
-  still a real, durable improvement — timing alone is noisy and
-  environment-dependent, query count is not.
-
-## Cleaning up after an interrupted run
-
-If a `seed`/`run`/`profile` invocation gets killed mid-run (Ctrl-C, `kill -9`,
-a timed-out SSH session, etc.), check whether it left anything behind before
-trusting the next benchmark's numbers. This was verified for real: a
-`benchmark seed --tier large --reset ...` was started against both a fresh
-PostgreSQL 18 container and a fresh MariaDB 12.3 container and `kill -9`'d a
-few seconds into document seeding. In both cases, the database-side
-connection disappeared immediately -- no stuck backend, no lingering query,
-no held lock was observed in either backend once the killed process's PID
-was confirmed gone. That said, this was one interruption point (mid
-bulk-seed, between chunks); a run killed mid-query, or a driver/network
-hiccup that doesn't cleanly close the socket, could behave differently, so
-still check before trusting a number if any run in the session was
-interrupted:
-
- **PostgreSQL**: look for leftover connections against the benchmark
-  database:
-
-  ```sql
-  SELECT pid, state, query, query_start
-  FROM pg_stat_activity
-  WHERE datname = current_database() AND pid <> pg_backend_pid();
-  ```
-
-  If a stuck backend shows up, clear it with:
-
-  ```sql
-  SELECT pg_terminate_backend(pid)
-  FROM pg_stat_activity
-  WHERE datname = current_database() AND pid <> pg_backend_pid();
-  ```
-
- **MariaDB**: look for leftover connections/queries:
-
-  ```sql
-  SHOW FULL PROCESSLIST;
-  ```
-
-  If a stuck connection shows up (anything other than your current admin
-  session), clear it with:
-
-  ```sql
-  KILL <id>;
-  ```
-
-A stray connection left running concurrently with a subsequent benchmark run
-would add real, contaminating load (extra queries competing for the same
-rows, possibly held locks slowing the next run's timings) -- cheap to rule
-out, expensive to silently trust a number that was actually measured
-alongside a zombie connection.
-
-## When to graduate a one-off finding into a permanent scenario
-
-Register a scenario (rather than leaving it as throwaway code on the
-investigation branch) when **both** are true:
-
- The query pattern is one this codebase is likely to regress on again (e.g.
-  it involves a permission-check join, a bulk operation, or anything else
-  with an easy-to-reintroduce N+1) — not a one-time fluke specific to this
-  investigation.
- Re-running it later, against a fresh seed, would still produce a
-  meaningful signal (it doesn't depend on investigation-specific throwaway
-  data or a fix that's already permanently landed and can't regress the same
-  way).
-
-If a finding doesn't meet both bars, keep it as disposable code on the
-investigation branch and let the branch's evidence (captured in the PR
-description of whatever production fix it motivates) be the permanent
-record instead.
@@ -115,6 +115,3 @@ celerybeat-schedule*

 # Git worktree local folder
 .worktrees
-
-# Benchmark tooling output (local only, never committed)
-/benchmark_results/
@@ -0,0 +1,12 @@
+#!/command/with-contenv /usr/bin/bash
+# shellcheck shell=bash
+
+declare -r log_prefix="[init-llmindex-migrate]"
+
+echo "${log_prefix} Checking for pending LLM index migrations..."
+cd "${PAPERLESS_SRC_DIR}"
+if [[ -n "${USER_IS_NON_ROOT}" ]]; then
+	python3 manage.py document_llmindex migrate
+else
+	s6-setuidgid paperless python3 manage.py document_llmindex migrate
+fi
@@ -0,0 +1 @@
+oneshot
@@ -0,0 +1 @@
+/etc/s6-overlay/s6-rc.d/init-llmindex-migrate/run
@@ -212,6 +212,16 @@ following:
    This is a no-op if the index is already up to date, so it is safe to
    run on every upgrade.

+5.  Migrate the LLM index if needed.
+
+    ```shell-session
+    cd src
+    python3 manage.py document_llmindex migrate
+    ```
+
+    This is a no-op if the index schema is already current, so it is safe
+    to run on every upgrade.
+
 ### Database Upgrades

 Paperless-ngx is compatible with Django-supported versions of PostgreSQL and MariaDB and it is generally
@@ -532,7 +542,7 @@ index is updated automatically on the schedule set by
 can manage it manually:

 ```
-document_llmindex {rebuild,update,compact}
+document_llmindex {rebuild,update,compact,migrate}
 ```

 Specify `rebuild` to build the index from scratch from all documents in the database. Use
@@ -1849,7 +1849,7 @@
        </context-group>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">154</context>
+          <context context-type="linenumber">159</context>
        </context-group>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/filterable-dropdown/filterable-dropdown.component.html</context>
@@ -3972,11 +3972,11 @@
        </context-group>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">96</context>
+          <context context-type="linenumber">101</context>
        </context-group>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">102</context>
+          <context context-type="linenumber">107</context>
        </context-group>
      </trans-unit>
      <trans-unit id="3800326155195149498" datatype="html">
@@ -3987,11 +3987,11 @@
        </context-group>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">97</context>
+          <context context-type="linenumber">102</context>
        </context-group>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">103</context>
+          <context context-type="linenumber">108</context>
        </context-group>
      </trans-unit>
      <trans-unit id="7551700625201096185" datatype="html">
@@ -4002,14 +4002,14 @@
        </context-group>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">119</context>
+          <context context-type="linenumber">124</context>
        </context-group>
      </trans-unit>
      <trans-unit id="3184700926171002527" datatype="html">
        <source>Any</source>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">152</context>
+          <context context-type="linenumber">157</context>
        </context-group>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/filterable-dropdown/filterable-dropdown.component.html</context>
@@ -4020,21 +4020,21 @@
        <source>Not</source>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">157</context>
+          <context context-type="linenumber">162</context>
        </context-group>
      </trans-unit>
      <trans-unit id="6548676277933116532" datatype="html">
        <source>Add query</source>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">176</context>
+          <context context-type="linenumber">181</context>
        </context-group>
      </trans-unit>
      <trans-unit id="5599577087865387184" datatype="html">
        <source>Add expression</source>
        <context-group purpose="location">
          <context context-type="sourcefile">src/app/components/common/custom-fields-query-dropdown/custom-fields-query-dropdown.component.html</context>
-          <context context-type="linenumber">179</context>
+          <context context-type="linenumber">184</context>
        </context-group>
      </trans-unit>
      <trans-unit id="6312759212949884929" datatype="html">
@@ -182,7 +182,7 @@
                    container="body" triggers="mouseenter:mouseleave" popoverClass="popover-slim">
                    <i-bs class="me-2" name="stack"></i-bs><span><ng-container i18n>Attributes</ng-container></span>
                  </a>
-                  @if (!slimSidebarEnabled) {
+                  @if (!slimSidebarEnabled && canSaveSettings) {
                    <button
                      type="button"
                      class="btn btn-link btn-sm text-muted p-0 me-3 attributes-expand-btn"
@@ -68,6 +68,11 @@
    ></ng-select>
  } @else if (getCustomFieldByID(atom.field)?.data_type === CustomFieldDataType.DocumentLink) {
    <pngx-input-document-link [(ngModel)]="atom.value" class="w-25 form-select doc-link-select p-0" placeholder="Search docs..." i18n-placeholder [minimal]="true"></pngx-input-document-link>
+  } @else if (getCustomFieldByID(atom.field)?.data_type === CustomFieldDataType.Monetary) {
+    <input class="w-25 form-control rounded-end" type="text" inputmode="decimal"
+      [ngModel]="atom.value"
+      (ngModelChange)="setMonetaryValue(atom, $event)"
+      [disabled]="disabled">
  } @else {
    <input class="w-25 form-control rounded-end" type="text" [(ngModel)]="atom.value" [disabled]="disabled">
  }
@@ -1,5 +1,6 @@
 import { provideHttpClient, withInterceptorsFromDi } from '@angular/common/http'
 import { provideHttpClientTesting } from '@angular/common/http/testing'
+import { LOCALE_ID } from '@angular/core'
 import { ComponentFixture, TestBed } from '@angular/core/testing'
 import { FormsModule, ReactiveFormsModule } from '@angular/forms'
 import { NgbDropdownModule } from '@ng-bootstrap/ng-bootstrap'
@@ -41,6 +42,12 @@ const customFields = [
      ],
    },
  },
+  {
+    id: 3,
+    name: 'Test Monetary Field',
+    data_type: CustomFieldDataType.Monetary,
+    extra_data: { default_currency: 'EUR' },
+  },
 ]

 describe('CustomFieldsQueryDropdownComponent', () => {
@@ -61,6 +68,7 @@ describe('CustomFieldsQueryDropdownComponent', () => {
      providers: [
        provideHttpClient(withInterceptorsFromDi()),
        provideHttpClientTesting(),
+        { provide: LOCALE_ID, useValue: 'de' },
      ],
    }).compileComponents()

@@ -150,6 +158,22 @@ describe('CustomFieldsQueryDropdownComponent', () => {
    expect(options2).toEqual([])
  })

+  it('should normalize localized monetary comparison values', () => {
+    const atom = new CustomFieldQueryAtom([3, 'exact', null])
+
+    component.setMonetaryValue(atom, '1.234,56')
+
+    expect(atom.value).toEqual('1234.56')
+  })
+
+  it('should preserve API-formatted monetary comparison values', () => {
+    const atom = new CustomFieldQueryAtom([3, 'exact', null])
+
+    component.setMonetaryValue(atom, '1234.56')
+
+    expect(atom.value).toEqual('1234.56')
+  })
+
  it('should remove an element from the selection model', () => {
    const expression = new CustomFieldQueryExpression()
    const atom = new CustomFieldQueryAtom()
@@ -1,9 +1,14 @@
-import { NgTemplateOutlet } from '@angular/common'
+import {
+  getLocaleNumberSymbol,
+  NgTemplateOutlet,
+  NumberSymbol,
+} from '@angular/common'
 import {
  Component,
  EventEmitter,
  inject,
  Input,
+  LOCALE_ID,
  Output,
  QueryList,
  signal,
@@ -212,6 +217,7 @@ export class CustomFieldQueriesModel {
 })
 export class CustomFieldsQueryDropdownComponent extends LoadingComponentWithPermissions {
  protected customFieldsService = inject(CustomFieldsService)
+  private readonly locale = inject(LOCALE_ID)

  public CustomFieldQueryComponentType = CustomFieldQueryElementType
  public CustomFieldQueryOperator = CustomFieldQueryOperator
@@ -376,4 +382,18 @@ export class CustomFieldsQueryDropdownComponent extends LoadingComponentWithPerm
    }
    return []
  }
+
+  setMonetaryValue(atom: CustomFieldQueryAtom, value: string) {
+    // Normalize the decimal symbol e.g. . vs , by locale
+    const decimalSymbol = getLocaleNumberSymbol(
+      this.locale,
+      NumberSymbol.Decimal
+    )
+    if (decimalSymbol !== '.' && value.includes(decimalSymbol)) {
+      const groupSymbol = getLocaleNumberSymbol(this.locale, NumberSymbol.Group)
+      value = value.split(groupSymbol).join('').split(decimalSymbol).join('.')
+    }
+
+    atom.value = value
+  }
 }
@@ -51,8 +51,8 @@
      *pngxIfPermissions="{ action: PermissionAction.Add, type: activeManagementList.permissionType }">
      <i-bs name="plus-circle" class="me-1"></i-bs><ng-container i18n>Create</ng-container>
    </button>
-  } @else if (activeCustomFields) {
-    <button type="button" class="btn btn-sm btn-outline-primary" (click)="activeCustomFields.editField()"
+  } @else if (customFieldsActive) {
+    <button type="button" class="btn btn-sm btn-outline-primary" (click)="addCustomField()"
      *pngxIfPermissions="{ action: PermissionAction.Add, type: PermissionType.CustomField }">
      <i-bs name="plus-circle" class="me-1"></i-bs><ng-container i18n>Add Field</ng-container>
    </button>
@@ -18,6 +18,7 @@ import {
  DocumentAttributesComponent,
  DocumentAttributesSectionKind,
 } from './document-attributes.component'
+import { CustomFieldsComponent } from './custom-fields/custom-fields.component'
 import { ManagementListComponent } from './management-list/management-list.component'

@Component({
@@ -207,6 +208,29 @@ describe('DocumentAttributesComponent', () => {
    expect(component.activeSection.kind).toBe(
      DocumentAttributesSectionKind.CustomFields
    )
-    expect(component.activeCustomFields).toBeDefined()
+    const customFields = Object.create(CustomFieldsComponent.prototype)
+    customFields.editField = jest.fn()
+    component.activeOutlet = {
+      componentInstance: customFields,
+    } as any
+    expect(component.activeCustomFields).toBe(customFields)
+
+    component.addCustomField()
+    expect(customFields.editField).toHaveBeenCalled()
+  })
+
+  it('should show the add field button before the custom fields instance is available', async () => {
+    jest.spyOn(permissionsService, 'currentUserCan').mockReturnValue(true)
+
+    fixture.detectChanges()
+    component.activeNavID.set(2)
+    await fixture.whenStable()
+
+    expect(component.activeCustomFields).toBeNull()
+    expect(
+      fixture.nativeElement.querySelector(
+        'pngx-page-header .btn-outline-primary'
+      )?.textContent
+    ).toContain('Add Field')
  })
 })
@@ -163,12 +163,17 @@ export class DocumentAttributesComponent implements OnInit, OnDestroy {
  }

  get activeCustomFields(): CustomFieldsComponent | null {
-    if (this.activeSection?.kind !== DocumentAttributesSectionKind.CustomFields)
-      return null
+    if (!this.customFieldsActive) return null
    const instance = this.activeOutlet?.componentInstance
    return instance instanceof CustomFieldsComponent ? instance : null
  }

+  get customFieldsActive(): boolean {
+    return (
+      this.activeSection?.kind === DocumentAttributesSectionKind.CustomFields
+    )
+  }
+
  get activeTabLabel(): string {
    return this.activeSection?.label ?? ''
  }
@@ -224,6 +229,10 @@ export class DocumentAttributesComponent implements OnInit, OnDestroy {
    this.router.navigate(['attributes', nextSection])
  }

+  addCustomField(): void {
+    this.activeCustomFields?.editField(null)
+  }
+
  private getDefaultNavID(): DocumentAttributesNavIDs | null {
    return this.visibleSections[0]?.id ?? null
  }
@@ -57,9 +57,7 @@ from paperless.models import ArchiveFileGenerationChoices
 from paperless.parsers import ParserContext
 from paperless.parsers import ParserProtocol
 from paperless.parsers.registry import get_parser_registry
-from paperless.parsers.utils import PDF_TEXT_MIN_LENGTH
-from paperless.parsers.utils import extract_pdf_text
-from paperless.parsers.utils import is_tagged_pdf
+from paperless.parsers.utils import pdf_born_digital_text

 LOGGING_NAME: Final[str] = "paperless.consumer"

@@ -138,53 +136,45 @@ def should_produce_archive(

    # Must produce a PDF so the frontend can display the original format at all.
    if parser.requires_pdf_rendition:
-        _log.debug("Archive: yes — parser requires PDF rendition for frontend display")
+        _log.debug("Archive: yes - parser requires PDF rendition for frontend display")
        return True

    # Parser cannot produce an archive (e.g. TextDocumentParser).
    if not parser.can_produce_archive:
-        _log.debug("Archive: no — parser cannot produce archives")
+        _log.debug("Archive: no - parser cannot produce archives")
        return False

    generation = OcrConfig().archive_file_generation

    if generation == ArchiveFileGenerationChoices.ALWAYS:
-        _log.debug("Archive: yes — ARCHIVE_FILE_GENERATION=always")
+        _log.debug("Archive: yes - ARCHIVE_FILE_GENERATION=always")
        return True
    if generation == ArchiveFileGenerationChoices.NEVER:
-        _log.debug("Archive: no — ARCHIVE_FILE_GENERATION=never")
+        _log.debug("Archive: no - ARCHIVE_FILE_GENERATION=never")
        return False

    # auto: produce archives for scanned/image documents; skip for born-digital PDFs.
    if mime_type.startswith("image/"):
-        _log.debug("Archive: yes — image document, ARCHIVE_FILE_GENERATION=auto")
+        _log.debug("Archive: yes - image document, ARCHIVE_FILE_GENERATION=auto")
        return True
    if mime_type == "application/pdf":
-        text = extract_pdf_text(document_path)
-        has_text = text is not None and len(text) > 0
-        if has_text and is_tagged_pdf(document_path):
+        text, born_digital = pdf_born_digital_text(document_path, log=_log)
+        text_length = len(text) if text else 0
+        if born_digital:
            _log.debug(
-                "Archive: no — born-digital PDF (structure tags detected),"
+                "Archive: no - born-digital PDF (text_length=%d),"
                " ARCHIVE_FILE_GENERATION=auto",
+                text_length,
            )
            return False
-        if text is None or len(text) <= PDF_TEXT_MIN_LENGTH:
-            _log.debug(
-                "Archive: yes — scanned PDF (text_length=%d ≤ %d),"
-                " ARCHIVE_FILE_GENERATION=auto",
-                len(text) if text else 0,
-                PDF_TEXT_MIN_LENGTH,
-            )
-            return True
        _log.debug(
-            "Archive: no — born-digital PDF (text_length=%d > %d),"
+            "Archive: yes - scanned/textless PDF (text_length=%d),"
            " ARCHIVE_FILE_GENERATION=auto",
-            len(text),
-            PDF_TEXT_MIN_LENGTH,
+            text_length,
        )
-        return False
+        return True
    _log.debug(
-        "Archive: no — MIME type %r not eligible for auto archive generation",
+        "Archive: no - MIME type %r not eligible for auto archive generation",
        mime_type,
    )
    return False
@@ -3,6 +3,7 @@ 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
+from paperless_ai.indexing import llm_index_migrate


 class Command(PaperlessCommand):
@@ -13,12 +14,18 @@ class Command(PaperlessCommand):

    def add_arguments(self, parser: Any) -> None:
        super().add_arguments(parser)
-        parser.add_argument("command", choices=["rebuild", "update", "compact"])
+        parser.add_argument(
+            "command",
+            choices=["rebuild", "update", "compact", "migrate"],
+        )

    def handle(self, *args: Any, **options: Any) -> None:
        if options["command"] == "compact":
            llm_index_compact()
            return
+        if options["command"] == "migrate":
+            llm_index_migrate()
+            return
        llmindex_index(
            rebuild=options["command"] == "rebuild",
            iter_wrapper=lambda docs: self.track(
@@ -61,7 +61,7 @@ class Command(PaperlessCommand):
        )
        table.add_column("Level", width=7, no_wrap=True)
        table.add_column("Document", min_width=20)
-        table.add_column("Issue", ratio=1)
+        table.add_column("Issue", ratio=1, overflow="fold")

        for doc_pk, doc_messages in messages.iter_messages():
            if doc_pk is not None:
@@ -9,6 +9,7 @@ if TYPE_CHECKING:

 _COMPACT = "documents.management.commands.document_llmindex.llm_index_compact"
 _INDEX = "documents.management.commands.document_llmindex.llmindex_index"
+_MIGRATE = "documents.management.commands.document_llmindex.llm_index_migrate"


 class TestDocumentLlmindexCommand:
@@ -17,6 +18,11 @@ class TestDocumentLlmindexCommand:
        call_command("document_llmindex", "compact")
        mock_compact.assert_called_once_with()

+    def test_migrate_calls_llm_index_migrate(self, mocker: MockerFixture) -> None:
+        mock_migrate = mocker.patch(_MIGRATE)
+        call_command("document_llmindex", "migrate")
+        mock_migrate.assert_called_once_with()
+
    def test_rebuild_calls_llmindex_index_with_rebuild_true(
        self,
        mocker: MockerFixture,
@@ -1329,7 +1329,7 @@ class PreConsumeTestCase(DirectoriesMixin, GetConsumerMixin, TestCase):
        with self.get_consumer(self.test_file) as c:
            c.run()
            # Verify no pre-consume script subprocess was invoked
-            # (run_subprocess may still be called by _extract_text_for_archive_check)
+            # (run_subprocess may still be called by pdf_born_digital_text via pdftotext)
            script_calls = [
                call
                for call in m.call_args_list
@@ -1354,7 +1354,7 @@ class PreConsumeTestCase(DirectoriesMixin, GetConsumerMixin, TestCase):
                    self.assertTrue(m.called)

                    # Find the call that invoked the pre-consume script
-                    # (run_subprocess may also be called by _extract_text_for_archive_check)
+                    # (run_subprocess may also be called by pdf_born_digital_text via pdftotext)
                    script_call = next(
                        call
                        for call in m.call_args_list
@@ -134,60 +134,32 @@ class TestShouldProduceArchive:
        assert should_produce_archive(parser, mime, Path("/tmp/doc")) is expected

    @pytest.mark.parametrize(
-        ("extracted_text", "expected"),
+        ("born_digital", "expected"),
        [
-            pytest.param(
-                "This is a born-digital PDF with lots of text content. " * 10,
-                False,
-                id="born-digital-long-text-skips-archive",
-            ),
-            pytest.param(None, True, id="no-text-scanned-produces-archive"),
-            pytest.param("tiny", True, id="short-text-treated-as-scanned"),
+            pytest.param(True, False, id="born-digital-skips-archive"),
+            pytest.param(False, True, id="not-born-digital-produces-archive"),
        ],
    )
    def test_auto_pdf_archive_decision(
        self,
        mocker: MockerFixture,
        settings,
-        extracted_text: str | None,
+        born_digital: bool,  # noqa: FBT001
        expected: bool,  # noqa: FBT001
    ) -> None:
+        """Archive decision tracks pdf_born_digital_text()'s verdict exactly.
+
+        should_produce_archive() defers entirely to pdf_born_digital_text()
+        for the has-real-text decision, so both callers of that predicate
+        (this function and RasterisedDocumentParser.parse()) always agree.
+        """
        settings.ARCHIVE_FILE_GENERATION = "auto"
-        mocker.patch("documents.consumer.is_tagged_pdf", return_value=False)
-        mocker.patch("documents.consumer.extract_pdf_text", return_value=extracted_text)
+        mocker.patch(
+            "documents.consumer.pdf_born_digital_text",
+            return_value=("some text", born_digital),
+        )
        parser = _parser_instance(can_produce=True, requires_rendition=False)
        assert (
            should_produce_archive(parser, "application/pdf", Path("/tmp/doc.pdf"))
            is expected
        )
-
-    def test_tagged_pdf_skips_archive_in_auto_mode(
-        self,
-        mocker: MockerFixture,
-        settings,
-    ) -> None:
-        """Tagged PDFs (e.g. Word exports) with real text are treated as born-digital, even below PDF_TEXT_MIN_LENGTH."""
-        settings.ARCHIVE_FILE_GENERATION = "auto"
-        mocker.patch("documents.consumer.is_tagged_pdf", return_value=True)
-        mocker.patch("documents.consumer.extract_pdf_text", return_value="tiny")
-        parser = _parser_instance(can_produce=True, requires_rendition=False)
-        assert (
-            should_produce_archive(parser, "application/pdf", Path("/tmp/doc.pdf"))
-            is False
-        )
-
-    def test_tagged_pdf_without_text_produces_archive(
-        self,
-        mocker: MockerFixture,
-        settings,
-    ) -> None:
-        """A tagged PDF with no actual extractable text (e.g. some scanner firmware) is not
-        trusted as born-digital — the tag alone must not bypass OCR."""
-        settings.ARCHIVE_FILE_GENERATION = "auto"
-        mocker.patch("documents.consumer.is_tagged_pdf", return_value=True)
-        mocker.patch("documents.consumer.extract_pdf_text", return_value=None)
-        parser = _parser_instance(can_produce=True, requires_rendition=False)
-        assert (
-            should_produce_archive(parser, "application/pdf", Path("/tmp/doc.pdf"))
-            is True
-        )
@@ -3,7 +3,6 @@ from __future__ import annotations
 import importlib.resources
 import logging
 import os
-import re
 import shutil
 import tempfile
 from pathlib import Path
@@ -25,9 +24,9 @@ from paperless.config import OcrConfig
 from paperless.models import CleanChoices
 from paperless.models import ModeChoices
 from paperless.models import OutputTypeChoices
-from paperless.parsers.utils import PDF_TEXT_MIN_LENGTH
 from paperless.parsers.utils import extract_pdf_text
-from paperless.parsers.utils import is_tagged_pdf
+from paperless.parsers.utils import is_born_digital_text
+from paperless.parsers.utils import post_process_text
 from paperless.parsers.utils import read_file_handle_unicode_errors
 from paperless.version import __full_version_str__

@@ -510,10 +509,10 @@ class RasterisedDocumentParser:

        if mime_type == "application/pdf":
            text_original = self.extract_text(None, document_path)
-            has_text = text_original is not None and len(text_original) > 0
-            original_has_text = has_text and (
-                is_tagged_pdf(document_path, log=self.log)
-                or len(text_original) > PDF_TEXT_MIN_LENGTH
+            original_has_text = is_born_digital_text(
+                text_original,
+                document_path,
+                log=self.log,
            )
        else:
            text_original = None
@@ -658,17 +657,3 @@ class RasterisedDocumentParser:
                    f"No text was found in {document_path}, the content will be empty.",
                )
                self.text = ""
-
-
-def post_process_text(text: str | None) -> str | None:
-    if not text:
-        return None
-
-    collapsed_spaces = re.sub(r"([^\S\r\n]+)", " ", text)
-    no_leading_whitespace = re.sub(r"([\n\r]+)([^\S\n\r]+)", "\\1", collapsed_spaces)
-    no_trailing_whitespace = re.sub(r"([^\S\n\r]+)$", "", no_leading_whitespace)
-
-    # TODO: this needs a rework
-    # replace \0 prevents issues with saving to postgres.
-    # text may contain \0 when this character is present in PDF files.
-    return no_trailing_whitespace.strip().replace("\0", " ")
@@ -111,6 +111,88 @@ def extract_pdf_text(
        return None


+def post_process_text(text: str | None) -> str | None:
+    """Normalize extracted PDF/OCR text: collapse whitespace, strip padding.
+
+    Returns ``None`` for ``None`` or whitespace-only input, so callers can
+    treat "no text" and "only layout padding" the same way.
+    """
+    if not text:
+        return None
+
+    collapsed_spaces = re.sub(r"([^\S\r\n]+)", " ", text)
+    no_leading_whitespace = re.sub(r"([\n\r]+)([^\S\n\r]+)", "\\1", collapsed_spaces)
+    no_trailing_whitespace = re.sub(r"([^\S\n\r]+)$", "", no_leading_whitespace)
+
+    # replace \0 prevents issues with saving to postgres.
+    # text may contain \0 when this character is present in PDF files.
+    result = no_trailing_whitespace.strip().replace("\0", " ")
+    return result or None
+
+
+def is_born_digital_text(
+    text: str | None,
+    path: Path,
+    log: logging.Logger | None = None,
+) -> bool:
+    """Decide whether already-extracted, normalized PDF text counts as born-digital.
+
+    This is the single source of truth for "does this PDF already have real
+    text", used both to decide whether to produce an archive file and to
+    decide whether OCR can be skipped. Both decisions must agree, or a
+    tagged-but-textless PDF can end up with no archive AND a forced OCR pass
+    (see GH #13387): raw ``pdftotext -layout`` output can be non-empty
+    (whitespace/form-feed padding) even when there is no real content, so
+    *text* must already be normalized via :func:`post_process_text`, not the
+    raw extraction.
+
+    Parameters
+    ----------
+    text:
+        The normalized extracted text (or ``None``) to evaluate.
+    path:
+        Absolute path to the PDF file, used for the tagged-PDF check.
+    log:
+        Logger for warnings.  Falls back to the module-level logger when omitted.
+
+    Returns
+    -------
+    bool
+        Whether the PDF counts as born-digital (has real text, and is either
+        tagged or exceeds ``PDF_TEXT_MIN_LENGTH``).
+    """
+    if not text:
+        return False
+    return is_tagged_pdf(path, log=log) or len(text) > PDF_TEXT_MIN_LENGTH
+
+
+def pdf_born_digital_text(
+    path: Path,
+    log: logging.Logger | None = None,
+) -> tuple[str | None, bool]:
+    """Extract a PDF's text and decide whether it should be treated as born-digital.
+
+    Convenience wrapper around :func:`is_born_digital_text` for callers that
+    don't already have the PDF's text extracted (e.g. the archive-generation
+    decision, which runs before any parser has touched the file).
+
+    Parameters
+    ----------
+    path:
+        Absolute path to the PDF file.
+    log:
+        Logger for warnings.  Falls back to the module-level logger when omitted.
+
+    Returns
+    -------
+    tuple[str | None, bool]
+        The normalized extracted text (or ``None``), and whether the PDF
+        counts as born-digital.
+    """
+    text = post_process_text(extract_pdf_text(path, log=log))
+    return text, is_born_digital_text(text, path, log=log)
+
+
 def read_file_handle_unicode_errors(
    filepath: Path,
    log: logging.Logger | None = None,
@@ -150,7 +150,6 @@ INSTALLED_APPS = [
    "drf_spectacular",
    "drf_spectacular_sidecar",
    "treenode",
-    "paperless_benchmark.apps.PaperlessBenchmarkConfig",
    *env_apps,
 ]

@@ -36,6 +36,23 @@ def samples_dir() -> Path:
    return (Path(__file__).parent / "samples").resolve()


+@pytest.fixture(scope="session")
+def tagged_no_text_pdf_file(samples_dir: Path) -> Path:
+    """Path to a tagged PDF whose only "text" is pdftotext layout padding.
+
+    Reproduces GH #13387: ``/MarkInfo /Marked true`` is set, but the only
+    extractable content is a form-feed byte, not real text. Lives here
+    rather than in parsers/conftest.py so both parser tests and
+    paperless/tests/test_parser_utils.py can use it.
+
+    Returns
+    -------
+    Path
+        Absolute path to ``tesseract/tagged-but-no-text.pdf``.
+    """
+    return samples_dir / "tesseract" / "tagged-but-no-text.pdf"
+
+
@pytest.fixture(autouse=True)
 def clean_registry() -> Generator[None, None, None]:
    """Reset the parser registry before and after every test.
@@ -21,7 +21,7 @@ from documents.parsers import run_convert
 from paperless.models import ModeChoices
 from paperless.parsers import ParserProtocol
 from paperless.parsers.tesseract import RasterisedDocumentParser
-from paperless.parsers.tesseract import post_process_text
+from paperless.parsers.utils import is_tagged_pdf

 if TYPE_CHECKING:
    from pathlib import Path
@@ -151,36 +151,6 @@ class TestRasterisedDocumentParserLifecycle:
        assert tempdir is not None and not tempdir.exists()


-# ---------------------------------------------------------------------------
-# post_process_text
-# ---------------------------------------------------------------------------
-
-
-class TestPostProcessText:
-    @pytest.mark.parametrize(
-        ("source", "expected"),
-        [
-            pytest.param(
-                "simple     string",
-                "simple string",
-                id="collapse-spaces",
-            ),
-            pytest.param(
-                "simple    newline\n   testing string",
-                "simple newline\ntesting string",
-                id="preserve-newline",
-            ),
-            pytest.param(
-                "utf-8   строка с пробелами в конце  ",  # noqa: RUF001
-                "utf-8 строка с пробелами в конце",  # noqa: RUF001
-                id="utf8-trailing-spaces",
-            ),
-        ],
-    )
-    def test_post_process_text(self, source: str, expected: str) -> None:
-        assert post_process_text(source) == expected
-
-
 # ---------------------------------------------------------------------------
 # Page count
 # ---------------------------------------------------------------------------
@@ -910,25 +880,25 @@ class TestSkipArchive:
        self,
        mocker: MockerFixture,
        tesseract_parser: RasterisedDocumentParser,
-        tesseract_samples_dir: Path,
+        tagged_no_text_pdf_file: Path,
    ) -> None:
        """
        GIVEN:
-            - A PDF that reports itself as tagged (/MarkInfo /Marked true) but
-              has no actual extractable text (some scanner firmware produces
-              this — see GitHub issue #13349)
+            - A real PDF that reports itself as tagged (/MarkInfo /Marked
+              true) but whose only pdftotext output is layout padding (a
+              lone form-feed byte), not real text (see GitHub issue #13387,
+              originally reported against #13349's tagged-PDF handling)
            - Mode: auto, produce_archive=False
        WHEN:
            - Document is parsed
        THEN:
            - The tag alone is not trusted as "has text"; OCRmyPDF still runs
        """
+        assert is_tagged_pdf(tagged_no_text_pdf_file) is True
        tesseract_parser.settings.mode = ModeChoices.AUTO
-        mocker.patch("paperless.parsers.tesseract.is_tagged_pdf", return_value=True)
-        mocker.patch.object(tesseract_parser, "extract_text", return_value=None)
        mock_ocr = mocker.patch("ocrmypdf.ocr")
        tesseract_parser.parse(
-            tesseract_samples_dir / "multi-page-images.pdf",
+            tagged_no_text_pdf_file,
            "application/pdf",
            produce_archive=False,
        )
@@ -4,10 +4,18 @@ from __future__ import annotations

 import codecs
 from pathlib import Path
+from typing import TYPE_CHECKING
+
+import pytest

 from paperless.parsers.utils import is_tagged_pdf
+from paperless.parsers.utils import pdf_born_digital_text
+from paperless.parsers.utils import post_process_text
 from paperless.parsers.utils import read_file_handle_unicode_errors

+if TYPE_CHECKING:
+    from pytest_mock import MockerFixture
+
 SAMPLES = Path(__file__).parent / "samples" / "tesseract"


@@ -60,3 +68,105 @@ class TestIsTaggedPdf:
        bad = tmp_path / "bad.pdf"
        bad.write_bytes(b"not a pdf")
        assert is_tagged_pdf(bad) is False
+
+
+class TestPostProcessText:
+    @pytest.mark.parametrize(
+        ("source", "expected"),
+        [
+            pytest.param(
+                "simple     string",
+                "simple string",
+                id="collapse-spaces",
+            ),
+            pytest.param(
+                "simple    newline\n   testing string",
+                "simple newline\ntesting string",
+                id="preserve-newline",
+            ),
+            pytest.param(
+                "utf-8   строка с пробелами в конце  ",  # noqa: RUF001
+                "utf-8 строка с пробелами в конце",  # noqa: RUF001
+                id="utf8-trailing-spaces",
+            ),
+            pytest.param(None, None, id="none-input"),
+            pytest.param("", None, id="empty-string"),
+            pytest.param("   \n\x0c  \n ", None, id="whitespace-and-formfeed-only"),
+        ],
+    )
+    def test_post_process_text(
+        self,
+        source: str | None,
+        expected: str | None,
+    ) -> None:
+        assert post_process_text(source) == expected
+
+
+class TestPdfBornDigitalText:
+    """Regression coverage for GH #13387.
+
+    should_produce_archive() and RasterisedDocumentParser.parse() must agree
+    on whether a PDF has real text, so both go through this one function.
+    """
+
+    @pytest.mark.parametrize(
+        ("extracted", "tagged", "expected_text", "expected_born_digital"),
+        [
+            pytest.param("tiny", True, "tiny", True, id="tagged-with-real-text"),
+            pytest.param("tiny", False, "tiny", False, id="untagged-below-min-length"),
+            pytest.param(
+                "x" * 51,
+                False,
+                "x" * 51,
+                True,
+                id="untagged-above-min-length",
+            ),
+            pytest.param(None, True, None, False, id="tagged-but-no-text"),
+        ],
+    )
+    def test_born_digital_decision(
+        self,
+        mocker: MockerFixture,
+        tmp_path: Path,
+        extracted: str | None,
+        tagged: bool,  # noqa: FBT001
+        expected_text: str | None,
+        expected_born_digital: bool,  # noqa: FBT001
+    ) -> None:
+        """
+        GIVEN:
+            - A PDF whose pdftotext output and /MarkInfo tag status vary
+        WHEN:
+            - pdf_born_digital_text() is called
+        THEN:
+            - The normalized text and born-digital verdict match; the tag
+              alone never counts as "has text"
+        """
+        mocker.patch(
+            "paperless.parsers.utils.extract_pdf_text",
+            return_value=extracted,
+        )
+        mocker.patch("paperless.parsers.utils.is_tagged_pdf", return_value=tagged)
+        text, born_digital = pdf_born_digital_text(tmp_path / "doc.pdf")
+        assert text == expected_text
+        assert born_digital is expected_born_digital
+
+    def test_tagged_but_textless_pdf_is_not_born_digital(
+        self,
+        tagged_no_text_pdf_file: Path,
+    ) -> None:
+        """
+        GIVEN:
+            - A real PDF that is tagged (/MarkInfo /Marked true) but whose
+              only "text" is layout padding (a stray form-feed byte)
+        WHEN:
+            - pdf_born_digital_text() is called with no mocking
+        THEN:
+            - The normalized text is None and the PDF is not treated as
+              born-digital. The raw, unnormalized pdftotext output is
+              non-empty for this file, which is exactly what caused the
+              archive decision to disagree with the OCR decision in #13387.
+        """
+        text, born_digital = pdf_born_digital_text(tagged_no_text_pdf_file)
+        assert text is None
+        assert born_digital is False
@@ -1,5 +1,6 @@
 import logging
 from collections.abc import Iterable
+from collections.abc import Iterator
 from contextlib import contextmanager
 from datetime import timedelta
 from typing import TYPE_CHECKING
@@ -22,6 +23,7 @@ from paperless_ai.embedding import get_configured_model_name
 from paperless_ai.embedding import get_embedding_model

 if TYPE_CHECKING:
+    from django.db.models import QuerySet
    from llama_index.core.schema import BaseNode

    from paperless_ai.vector_store import PaperlessSqliteVecVectorStore
@@ -32,6 +34,35 @@ logger = logging.getLogger("paperless_ai.indexing")
 RAG_NUM_OUTPUT = 512
 RAG_CHUNK_OVERLAP = 200

+# update_llm_index(): row count per .iterator() batch when streaming
+# documents for a rebuild/update, matching _DocumentViewerStream's chunk
+# size in documents/search/_backend.py.
+_INDEX_STREAM_CHUNK_SIZE = 1000
+
+
+class _StreamedDocuments:
+    """A thin QuerySet wrapper that streams via ``.iterator()`` instead of
+    materializing every row (plus its ``content`` and prefetch caches) into
+    memory at once, while still supporting ``len()`` so ``iter_wrapper``'s
+    progress bar shows a real total instead of falling back to indeterminate.
+    Same shape as ``documents/search/_backend.py``'s ``_DocumentViewerStream``,
+    just without that class's extra per-batch permission lookup -- nothing
+    here needs one.
+    """
+
+    def __init__(self, documents: "QuerySet[Document]") -> None:
+        self._documents = documents
+
+    def __len__(self) -> int:
+        return self._documents.count()
+
+    def __iter__(self) -> Iterator[Document]:
+        # iterator(chunk_size=...) streams from a server-side cursor instead
+        # of materializing the whole queryset in memory; since Django 4.1 it
+        # still honours prefetch_related, running the prefetches one batch
+        # at a time.
+        return iter(self._documents.iterator(chunk_size=_INDEX_STREAM_CHUNK_SIZE))
+

 def queue_llm_index_update_if_needed(*, rebuild: bool, reason: str) -> bool:
    # NOTE: The check-then-enqueue sequence below is non-atomic (TOCTOU): two
@@ -144,6 +175,24 @@ def _exclude_readers():
        lock.close()


+def _with_exclusive_access(operation: str, fn):
+    """Run ``fn()`` with exclusive index access (see ``_exclude_readers()``),
+    for compaction/migration file swaps that must not run while readers are
+    active. Returns ``fn()``'s result, or None (after logging) if active
+    readers do not drain within ``LLM_INDEX_COMPACTION_LOCK_TIMEOUT`` --
+    callers skip the operation this run; it retries next time.
+    """
+    try:
+        with _exclude_readers():
+            return fn()
+    except Timeout:
+        logger.info(
+            "Skipping LLM index %s: index readers are active; will retry next run.",
+            operation,
+        )
+        return None
+
+
@contextmanager
 def write_store(embed_model_name: str | None = None):
    """Acquire the write lock and yield the vector store.
@@ -168,6 +217,21 @@ def write_store(embed_model_name: str | None = None):
        yield store


+def _check_and_run_migrations(store: "PaperlessSqliteVecVectorStore") -> bool:
+    """Run any pending structural migrations, returning True if a pending
+    re-embed migration needs the caller to force a rebuild -- never
+    triggered automatically here. Safe to call before any write, including
+    delete()/upsert_document(): has_pending_migration() (see its docstring)
+    keeps this a no-op, with no exclusive access taken, once the store is
+    current.
+    """
+    if not store.has_pending_migration():
+        return False
+    return bool(
+        _with_exclusive_access("migration check", store.check_and_run_migrations),
+    )
+
+
 def _safe_related_name(document: Document, field: str) -> str | None:
    """
    Returns the ``name`` of a related object (correspondent, document_type,
@@ -339,15 +403,7 @@ def update_llm_index(
    happens, since a rebuild always covers the whole library regardless.
    """
    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
+        needs_reembed = _check_and_run_migrations(store)
        if needs_reembed:
            logger.warning(
                "LLM index migration requires re-embedding; forcing rebuild.",
@@ -385,7 +441,7 @@ def update_llm_index(
        if rebuild or not store.table_exists():
            logger.info("Rebuilding LLM index.")
            store.drop_table()
-            for document in iter_wrapper(documents):
+            for document in iter_wrapper(_StreamedDocuments(documents)):
                nodes = build_document_node(document, chunk_size=chunk_size)
                _embed_nodes(nodes, embed_model)
                store.add(nodes)
@@ -398,7 +454,7 @@ def update_llm_index(
            )
            existing = store.get_modified_times()
            changed = 0
-            for document in iter_wrapper(scoped_documents):
+            for document in iter_wrapper(_StreamedDocuments(scoped_documents)):
                doc_id = str(document.id)
                if existing.get(doc_id) == document.modified.isoformat():
                    continue
@@ -412,14 +468,7 @@ def update_llm_index(
                else "No changes detected in LLM index."
            )

-        try:
-            with _exclude_readers():
-                store.compact()
-        except Timeout:
-            logger.info(
-                "Skipping LLM index compaction: index readers are active; "
-                "will retry next run.",
-            )
+        _with_exclusive_access("compaction", store.compact)
    return msg


@@ -434,25 +483,60 @@ def llm_index_add_or_update_document(document: Document):
        _embed_nodes(new_nodes, get_embedding_model(config))

    with write_store(embed_model_name=get_configured_model_name(config)) as store:
+        needs_reembed = _check_and_run_migrations(store)
+        if needs_reembed:
+            logger.warning(
+                "Skipping incremental LLM index update for document %s: the "
+                "index requires re-embedding first. Run 'document_llmindex "
+                "rebuild' to resolve.",
+                document.id,
+            )
+            return
        store.upsert_document(str(document.id), new_nodes)


+def llm_index_migrate() -> None:
+    """Apply any pending LLM index schema migrations, with no reindex.
+
+    Intended to run unconditionally on every startup (see the
+    init-llmindex-migrate container step and the bare-metal upgrade docs):
+    has_pending_migration() short-circuits to a metadata-only read once the
+    store is current, so a healthy install pays almost nothing here. Only
+    ever applies structural migrations -- a pending re-embed migration is
+    left for the explicit, deliberate rebuild path (``document_llmindex
+    update``/``rebuild``) to resolve, since re-embedding can be slow and,
+    for a metered embedding backend, cost money.
+    """
+    if not AIConfig().llm_index_enabled:
+        return
+    with write_store() as store:
+        needs_reembed = _check_and_run_migrations(store)
+    if needs_reembed:
+        logger.warning(
+            "LLM index requires re-embedding, which this automatic migration "
+            "check will not do on its own -- it can be slow and, for a "
+            "metered embedding backend, cost money. Run "
+            "'document_llmindex rebuild' manually when ready.",
+        )
+
+
 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.",
-            )
+        _with_exclusive_access("compaction", lambda: store.compact(force=True))


 def llm_index_remove_document(document: Document):
    """Remove a document's chunks from the LLM index."""
    with write_store() as store:
+        if _check_and_run_migrations(store):
+            logger.warning(
+                "Skipping removal of document %s from the LLM index: the "
+                "index requires re-embedding first. Run 'document_llmindex "
+                "rebuild' to resolve.",
+                document.id,
+            )
+            return
        store.delete(str(document.id))


@@ -0,0 +1,60 @@
+"""Schema migrations for the sqlite-vec vector store.
+
+Each migration lives in its own module here, named ``mNNNN_description.py``
+(e.g. ``m0001_v1_to_v2.py`` -- a leading digit isn't a valid Python
+identifier, hence the ``m`` prefix, unlike Django's own numbered migrations,
+which load via a dynamic ``importlib.import_module()`` call rather than a
+static import statement), and registers itself into ``MIGRATIONS`` at import
+time. ``vector_store.py`` imports those modules at the bottom of the file,
+purely for that registration side effect, after ``PaperlessSqliteVecVectorStore``
+is fully defined -- migrations need it to implement ``apply()`` (see
+``Migration`` below).
+
+To add a new migration: add a new ``mNNNN_description.py`` module here that
+imports ``PaperlessSqliteVecVectorStore`` from ``paperless_ai.vector_store``,
+defines its ``apply()``, and appends a ``Migration`` to ``MIGRATIONS``; then
+import that module at the bottom of ``vector_store.py`` and bump
+``SCHEMA_VERSION`` there. A migration must freeze its own historical DDL for
+any side table its target version depends on (``DROP TABLE IF EXISTS`` +
+its own literal ``CREATE TABLE``/``CREATE INDEX`` statements) rather than
+delegating to any "current schema" helper -- see ``m0001_v1_to_v2.py`` for
+why and the worked example.
+"""
+
+import sqlite3
+from collections.abc import Callable
+from dataclasses import dataclass
+from dataclasses import field
+from typing import Literal
+
+
+@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 every table its target schema needs in ``dst_conn`` and copy
+    ``src_conn``'s rows and relevant ``index_meta`` keys into it.
+    ``schema_version`` is written by the migration runner after ``apply``
+    returns, not by ``apply`` itself.
+
+    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, populated by each migration
+# module's import-time registration (see the module docstring above).
+MIGRATIONS: list[Migration] = []
@@ -0,0 +1,229 @@
+"""Thin gateways over the plain relational side tables that sit alongside the
+vec0 table. Each method takes the sqlite3.Connection to operate on
+explicitly, rather than owning one -- the store swaps connections during
+compact()/migration, and migrations always work across two connections
+(src_conn, dst_conn) at once.
+
+PRECONDITION: Callers must set conn.row_factory = sqlite3.Row before passing a
+connection to any of these gateways' read methods. The read methods across all
+three classes (DocumentChunksTable.chunk_ids_for_document, IndexMetaTable._get,
+DocumentMetaTable.all_modified_times, DocumentMetaTable.copy_all) use
+row["column_name"] dictionary-style indexing, which requires sqlite3.Row as the
+row factory -- without it, sqlite3.Row is not set, rows are returned as plain
+tuples, and tuple indices must be integers, raising TypeError.
+"""
+
+import sqlite3
+from collections.abc import Iterable
+from typing import NamedTuple
+
+
+class ChunkRow(NamedTuple):
+    chunk_id: str
+    document_id: int
+
+
+class DocumentMetaRow(NamedTuple):
+    document_id: int
+    modified: str
+
+
+class DocumentChunksTable:
+    """chunk_id -> document_id, indexed by document_id. Gives O(1)
+    per-document chunk lookup that vec0's own document_id metadata column
+    cannot (see PaperlessSqliteVecVectorStore._delete_chunks_by_document_id).
+    """
+
+    @staticmethod
+    def create(conn: sqlite3.Connection) -> None:
+        conn.execute(
+            "CREATE TABLE IF NOT EXISTS document_chunks "
+            "(chunk_id TEXT PRIMARY KEY, document_id INTEGER NOT NULL)",
+        )
+        conn.execute(
+            "CREATE INDEX IF NOT EXISTS idx_document_chunks_document_id "
+            "ON document_chunks (document_id)",
+        )
+
+    @staticmethod
+    def insert_many(conn: sqlite3.Connection, rows: Iterable[ChunkRow]) -> None:
+        """rows must already be batch-bounded by the caller (e.g. vec0's own
+        fetchmany() loop) -- this never reads, so it can't itself introduce
+        an unbounded scan, but a whole-table iterable defeats the point."""
+        conn.executemany(
+            "INSERT INTO document_chunks (chunk_id, document_id) VALUES (?, ?)",
+            rows,
+        )
+
+    @staticmethod
+    def chunk_ids_for_document(
+        conn: sqlite3.Connection,
+        document_id: int,
+    ) -> list[str]:
+        return [
+            row["chunk_id"]
+            for row in conn.execute(
+                "SELECT chunk_id FROM document_chunks WHERE document_id = ?",
+                (document_id,),
+            ).fetchall()
+        ]
+
+    @staticmethod
+    def delete_for_document(conn: sqlite3.Connection, document_id: int) -> None:
+        conn.execute(
+            "DELETE FROM document_chunks WHERE document_id = ?",
+            (document_id,),
+        )
+
+    @staticmethod
+    def delete_all(conn: sqlite3.Connection) -> None:
+        conn.execute("DELETE FROM document_chunks")
+
+    @staticmethod
+    def count(conn: sqlite3.Connection) -> int:
+        """Cheap stand-in for vec0's own row count -- see compact()."""
+        return conn.execute("SELECT count(*) FROM document_chunks").fetchone()[0]
+
+
+class DocumentMetaTable:
+    """document_id -> modified, one row per document. Lives outside vec0
+    because vec0 only inlines TEXT metadata up to 12 bytes and `modified`
+    (an ISO timestamp) is always longer.
+    """
+
+    @staticmethod
+    def create(conn: sqlite3.Connection) -> None:
+        conn.execute(
+            "CREATE TABLE IF NOT EXISTS document_meta "
+            "(document_id INTEGER PRIMARY KEY, modified TEXT NOT NULL)",
+        )
+
+    @staticmethod
+    def upsert_many(
+        conn: sqlite3.Connection,
+        rows: Iterable[DocumentMetaRow],
+    ) -> None:
+        conn.executemany(
+            "INSERT INTO document_meta (document_id, modified) VALUES (?, ?) "
+            "ON CONFLICT(document_id) DO UPDATE SET modified = excluded.modified",
+            rows,
+        )
+
+    @staticmethod
+    def delete_for_document(conn: sqlite3.Connection, document_id: int) -> None:
+        conn.execute(
+            "DELETE FROM document_meta WHERE document_id = ?",
+            (document_id,),
+        )
+
+    @staticmethod
+    def delete_all(conn: sqlite3.Connection) -> None:
+        conn.execute("DELETE FROM document_meta")
+
+    @staticmethod
+    def copy_all(
+        src_conn: sqlite3.Connection,
+        dst_conn: sqlite3.Connection,
+        batch_size: int,
+    ) -> None:
+        """Stream document_meta from src_conn into dst_conn in bounded
+        batches. The *only* sanctioned way to move this table across
+        connections (compact()/migrations) -- an unbounded fetchall here
+        would defeat the same OOM-avoidance the vec0 row copy already relies
+        on. batch_size has no default: forces the call site to think about
+        it (pass COMPACT_BATCH_SIZE)."""
+        cursor = src_conn.execute(
+            "SELECT document_id, modified FROM document_meta",
+        )
+        while batch := cursor.fetchmany(batch_size):
+            DocumentMetaTable.upsert_many(
+                dst_conn,
+                (DocumentMetaRow(r["document_id"], r["modified"]) for r in batch),
+            )
+
+    @staticmethod
+    def all_modified_times(conn: sqlite3.Connection) -> dict[str, str]:
+        """Full document_id -> modified map, for get_modified_times()'s
+        public API only. One unbounded read by design (existing behavior).
+        Never use this for cross-connection copying; see copy_all()."""
+        return {
+            str(row["document_id"]): str(row["modified"] or "")
+            for row in conn.execute(
+                "SELECT document_id, modified FROM document_meta",
+            )
+        }
+
+
+class IndexMetaTable:
+    """Typed accessors over index_meta's key/value rows -- replaces
+    PaperlessSqliteVecVectorStore._meta_get_on/_meta_set_on, which returned
+    untyped str | None regardless of whether the key held an int (dim,
+    schema_version, total_inserts) or a string (embed_model).
+    """
+
+    @staticmethod
+    def create(conn: sqlite3.Connection) -> None:
+        conn.execute(
+            "CREATE TABLE IF NOT EXISTS index_meta (key TEXT PRIMARY KEY, value TEXT)",
+        )
+
+    @staticmethod
+    def _get(conn: sqlite3.Connection, key: str) -> str | None:
+        row = conn.execute(
+            "SELECT value FROM index_meta WHERE key = ?",
+            (key,),
+        ).fetchone()
+        return row["value"] if row else None
+
+    @staticmethod
+    def _set(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),
+        )
+
+    @staticmethod
+    def get_dim(conn: sqlite3.Connection) -> int | None:
+        value = IndexMetaTable._get(conn, "dim")
+        return int(value) if value is not None else None
+
+    @staticmethod
+    def set_dim(conn: sqlite3.Connection, dim: int) -> None:
+        IndexMetaTable._set(conn, "dim", str(dim))
+
+    @staticmethod
+    def get_embed_model(conn: sqlite3.Connection) -> str | None:
+        return IndexMetaTable._get(conn, "embed_model")
+
+    @staticmethod
+    def set_embed_model(conn: sqlite3.Connection, name: str) -> None:
+        IndexMetaTable._set(conn, "embed_model", name)
+
+    @staticmethod
+    def get_schema_version(conn: sqlite3.Connection) -> int | None:
+        value = IndexMetaTable._get(conn, "schema_version")
+        return int(value) if value is not None else None
+
+    @staticmethod
+    def set_schema_version(conn: sqlite3.Connection, version: int) -> None:
+        IndexMetaTable._set(conn, "schema_version", str(version))
+
+    @staticmethod
+    def get_total_inserts(conn: sqlite3.Connection) -> int:
+        value = IndexMetaTable._get(conn, "total_inserts")
+        return int(value) if value is not None else 0
+
+    @staticmethod
+    def increment_total_inserts(conn: sqlite3.Connection, count: int) -> None:
+        current = IndexMetaTable.get_total_inserts(conn)
+        IndexMetaTable._set(conn, "total_inserts", str(current + count))
+
+    @staticmethod
+    def reset_total_inserts(conn: sqlite3.Connection, count: int) -> None:
+        """Set total_inserts to an absolute value -- distinct from
+        increment_total_inserts(): used by compact()'s rebuild and by
+        m0001_v1_to_v2 after copying live rows into a fresh file, where
+        total_inserts must become exactly the live row count, not add to
+        whatever the source file's counter held."""
+        IndexMetaTable._set(conn, "total_inserts", str(count))
@@ -1,3 +1,4 @@
+import logging
 from pathlib import Path
 from unittest.mock import MagicMock
 from unittest.mock import patch
@@ -186,6 +187,45 @@ def test_truncate_embedding_query_returns_single_chunk() -> None:
    assert "word199" not in result


+class TestStreamedDocuments:
+    """_StreamedDocuments streams via .iterator() instead of materializing
+    the whole queryset (plus its content and prefetch caches) in memory at
+    once, while still supporting len() so a progress bar wrapped around it
+    shows a real total.
+    """
+
+    def test_len_and_iter_delegate_to_streaming_queryset_methods(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        """
+        GIVEN:
+            - A mock queryset
+        WHEN:
+            - A _StreamedDocuments wrapping it is measured and iterated
+        THEN:
+            - len() uses count() (not a materializing len()), and iteration
+              uses .iterator(chunk_size=...) (not plain iteration, which
+              would materialize prefetches for the whole queryset at once)
+        """
+        mock_queryset = mocker.MagicMock()
+        mock_queryset.count.return_value = 42
+        mock_queryset.iterator.return_value = iter(["doc-1", "doc-2"])
+        streamed = indexing._StreamedDocuments(mock_queryset)
+
+        assert len(streamed) == 42
+        assert list(streamed) == ["doc-1", "doc-2"]
+        # count.call_count isn't asserted exactly: list()'s own size-hint
+        # optimization calls len(streamed) again internally, on top of the
+        # explicit len() call above -- both legitimately delegate to
+        # count(), so only the delegation itself (not the call count) is
+        # the thing being verified here.
+        mock_queryset.count.assert_called_with()
+        mock_queryset.iterator.assert_called_once_with(
+            chunk_size=indexing._INDEX_STREAM_CHUNK_SIZE,
+        )
+
+
@pytest.mark.django_db
 def test_update_llm_index(
    temp_llm_index_dir: Path,
@@ -737,6 +777,7 @@ class TestLlmIndexLocking:
        mocker: pytest_mock.MockerFixture,
    ) -> None:
        mock_store = MagicMock()
+        mock_store.has_pending_migration.return_value = False
        mocker.patch(
            "paperless_ai.indexing.write_store",
            return_value=mocker.MagicMock(
@@ -757,12 +798,45 @@ class TestLlmIndexLocking:

        mock_store.upsert_document.assert_called_once()

+    def test_add_or_update_document_skips_write_when_reembed_pending(
+        self,
+        temp_llm_index_dir: Path,
+        mock_embed_model: FakeEmbedding,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        """A pending re-embed migration must block the incremental write,
+        not let it proceed against a schema that just changed underneath it.
+        """
+        mock_store = MagicMock()
+        mock_store.has_pending_migration.return_value = True
+        mock_store.check_and_run_migrations.return_value = True
+        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=[mock_node],
+        )
+
+        doc = MagicMock(spec=Document)
+        doc.id = 1
+        indexing.llm_index_add_or_update_document(doc)
+
+        mock_store.upsert_document.assert_not_called()
+
    def test_remove_document_uses_write_store(
        self,
        temp_llm_index_dir: Path,
        mocker: pytest_mock.MockerFixture,
    ) -> None:
        mock_store = MagicMock()
+        mock_store.has_pending_migration.return_value = False
        mocker.patch(
            "paperless_ai.indexing.write_store",
            return_value=mocker.MagicMock(
@@ -777,6 +851,31 @@ class TestLlmIndexLocking:

        mock_store.delete.assert_called_once_with("1")

+    def test_remove_document_skips_write_when_reembed_pending(
+        self,
+        temp_llm_index_dir: Path,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        """A pending re-embed migration must block the delete too, for the
+        same consistency reason as the incremental-update path.
+        """
+        mock_store = MagicMock()
+        mock_store.has_pending_migration.return_value = True
+        mock_store.check_and_run_migrations.return_value = True
+        mocker.patch(
+            "paperless_ai.indexing.write_store",
+            return_value=mocker.MagicMock(
+                __enter__=mocker.MagicMock(return_value=mock_store),
+                __exit__=mocker.MagicMock(return_value=False),
+            ),
+        )
+
+        doc = MagicMock(spec=Document)
+        doc.id = 1
+        indexing.llm_index_remove_document(doc)
+
+        mock_store.delete.assert_not_called()
+
    def test_update_llm_index_rebuild_uses_write_store(
        self,
        temp_llm_index_dir: Path,
@@ -849,6 +948,76 @@ class TestVectorStoreIndexing:
            assert rows >= 1


+class TestLlmIndexMigrate:
+    def test_noop_when_ai_disabled(self, mocker: pytest_mock.MockerFixture) -> None:
+        """
+        GIVEN:
+            - AI/LLM index support is disabled in configuration
+        WHEN:
+            - llm_index_migrate() is called
+        THEN:
+            - No store is opened and no migration check runs
+        """
+        mocker.patch(
+            "paperless_ai.indexing.AIConfig",
+            return_value=mocker.Mock(llm_index_enabled=False),
+        )
+        write_store_mock = mocker.patch("paperless_ai.indexing.write_store")
+        indexing.llm_index_migrate()
+        write_store_mock.assert_not_called()
+
+    def test_runs_pending_migration_when_enabled(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        """
+        GIVEN:
+            - AI/LLM index support is enabled
+        WHEN:
+            - llm_index_migrate() is called
+        THEN:
+            - The store is opened for write and a migration check runs
+        """
+        mocker.patch(
+            "paperless_ai.indexing.AIConfig",
+            return_value=mocker.Mock(llm_index_enabled=True),
+        )
+        store_mock = mocker.MagicMock()
+        store_mock.has_pending_migration.return_value = False
+        write_store_cm = mocker.patch("paperless_ai.indexing.write_store")
+        write_store_cm.return_value.__enter__.return_value = store_mock
+        indexing.llm_index_migrate()
+        store_mock.has_pending_migration.assert_called_once()
+
+    def test_logs_warning_when_reembed_needed(
+        self,
+        mocker: pytest_mock.MockerFixture,
+        caplog: pytest.LogCaptureFixture,
+    ) -> None:
+        """
+        GIVEN:
+            - AI/LLM index support is enabled
+            - A pending migration requires re-embedding
+        WHEN:
+            - llm_index_migrate() is called
+        THEN:
+            - A warning directs the operator to run a manual rebuild, since
+              this automatic check must never re-embed on its own
+        """
+        mocker.patch(
+            "paperless_ai.indexing.AIConfig",
+            return_value=mocker.Mock(llm_index_enabled=True),
+        )
+        store_mock = mocker.MagicMock()
+        store_mock.has_pending_migration.return_value = True
+        store_mock.check_and_run_migrations.return_value = True
+        write_store_cm = mocker.patch("paperless_ai.indexing.write_store")
+        write_store_cm.return_value.__enter__.return_value = store_mock
+        with caplog.at_level(logging.WARNING, logger="paperless_ai.indexing"):
+            indexing.llm_index_migrate()
+        assert "requires re-embedding" in caplog.text
+
+
@pytest.mark.django_db
 class TestQuerySimilarDocuments:
    def test_query_similar_documents_respects_allowed_ids(
@@ -0,0 +1,306 @@
+import sqlite3
+from collections.abc import Generator
+
+import pytest
+
+from paperless_ai.tables import ChunkRow
+from paperless_ai.tables import DocumentChunksTable
+from paperless_ai.tables import DocumentMetaRow
+from paperless_ai.tables import DocumentMetaTable
+from paperless_ai.tables import IndexMetaTable
+
+
+@pytest.fixture
+def conn() -> Generator[sqlite3.Connection, None, None]:
+    connection = sqlite3.connect(":memory:")
+    connection.row_factory = sqlite3.Row
+    try:
+        yield connection
+    finally:
+        connection.close()
+
+
+class TestDocumentChunksTable:
+    def test_create_is_idempotent(self, conn: sqlite3.Connection) -> None:
+        """
+        GIVEN:
+            - A bare sqlite3 connection
+        WHEN:
+            - create() is called, a row is inserted, then create() is called again
+        THEN:
+            - No error is raised and the row survives uncorrupted
+        """
+        DocumentChunksTable.create(conn)
+        DocumentChunksTable.insert_many(conn, [ChunkRow("c1", 1)])
+        DocumentChunksTable.create(conn)
+        assert DocumentChunksTable.chunk_ids_for_document(conn, 1) == ["c1"]
+
+    def test_insert_many_then_lookup_by_document_id(
+        self,
+        conn: sqlite3.Connection,
+    ) -> None:
+        """
+        GIVEN:
+            - An empty document_chunks table
+        WHEN:
+            - Two chunks for document 1 and one for document 2 are inserted
+        THEN:
+            - chunk_ids_for_document returns exactly the matching chunk ids
+        """
+        DocumentChunksTable.create(conn)
+        DocumentChunksTable.insert_many(
+            conn,
+            [ChunkRow("c1", 1), ChunkRow("c2", 1), ChunkRow("c3", 2)],
+        )
+        assert sorted(DocumentChunksTable.chunk_ids_for_document(conn, 1)) == [
+            "c1",
+            "c2",
+        ]
+        assert DocumentChunksTable.chunk_ids_for_document(conn, 2) == ["c3"]
+        assert DocumentChunksTable.chunk_ids_for_document(conn, 999) == []
+
+    def test_delete_for_document_removes_only_that_document(
+        self,
+        conn: sqlite3.Connection,
+    ) -> None:
+        """
+        GIVEN:
+            - Chunks for two different documents
+        WHEN:
+            - delete_for_document() is called for one of them
+        THEN:
+            - Only that document's chunks are removed
+        """
+        DocumentChunksTable.create(conn)
+        DocumentChunksTable.insert_many(
+            conn,
+            [ChunkRow("c1", 1), ChunkRow("c2", 2)],
+        )
+        DocumentChunksTable.delete_for_document(conn, 1)
+        assert DocumentChunksTable.chunk_ids_for_document(conn, 1) == []
+        assert DocumentChunksTable.chunk_ids_for_document(conn, 2) == ["c2"]
+
+    def test_delete_all_clears_every_row(self, conn: sqlite3.Connection) -> None:
+        """
+        GIVEN:
+            - Chunks for multiple documents
+        WHEN:
+            - delete_all() is called
+        THEN:
+            - count() returns 0
+        """
+        DocumentChunksTable.create(conn)
+        DocumentChunksTable.insert_many(
+            conn,
+            [ChunkRow("c1", 1), ChunkRow("c2", 2)],
+        )
+        DocumentChunksTable.delete_all(conn)
+        assert DocumentChunksTable.count(conn) == 0
+
+    def test_count_reflects_live_rows(self, conn: sqlite3.Connection) -> None:
+        """
+        GIVEN:
+            - An empty document_chunks table
+        WHEN:
+            - Rows are inserted then one document's rows are deleted
+        THEN:
+            - count() reflects the remaining row count
+        """
+        DocumentChunksTable.create(conn)
+        DocumentChunksTable.insert_many(
+            conn,
+            [ChunkRow("c1", 1), ChunkRow("c2", 1), ChunkRow("c3", 2)],
+        )
+        assert DocumentChunksTable.count(conn) == 3
+        DocumentChunksTable.delete_for_document(conn, 1)
+        assert DocumentChunksTable.count(conn) == 1
+
+
+class TestDocumentMetaTable:
+    def test_upsert_many_then_all_modified_times(
+        self,
+        conn: sqlite3.Connection,
+    ) -> None:
+        """
+        GIVEN:
+            - An empty document_meta table
+        WHEN:
+            - Two documents' modified timestamps are upserted
+        THEN:
+            - all_modified_times() returns both, keyed by str(document_id)
+        """
+        DocumentMetaTable.create(conn)
+        DocumentMetaTable.upsert_many(
+            conn,
+            [
+                DocumentMetaRow(1, "2026-01-01T00:00:00"),
+                DocumentMetaRow(2, "2026-02-02T00:00:00"),
+            ],
+        )
+        assert DocumentMetaTable.all_modified_times(conn) == {
+            "1": "2026-01-01T00:00:00",
+            "2": "2026-02-02T00:00:00",
+        }
+
+    def test_upsert_many_overwrites_existing_value(
+        self,
+        conn: sqlite3.Connection,
+    ) -> None:
+        """
+        GIVEN:
+            - A document_meta row for document 1
+        WHEN:
+            - upsert_many() is called again with a new modified value for
+              the same document_id
+        THEN:
+            - The stored value is replaced, not duplicated
+        """
+        DocumentMetaTable.create(conn)
+        DocumentMetaTable.upsert_many(conn, [DocumentMetaRow(1, "old")])
+        DocumentMetaTable.upsert_many(conn, [DocumentMetaRow(1, "new")])
+        assert DocumentMetaTable.all_modified_times(conn) == {"1": "new"}
+
+    def test_delete_for_document_removes_only_that_row(
+        self,
+        conn: sqlite3.Connection,
+    ) -> None:
+        """
+        GIVEN:
+            - document_meta rows for two documents
+        WHEN:
+            - delete_for_document() is called for one of them
+        THEN:
+            - Only that document's row is removed
+        """
+        DocumentMetaTable.create(conn)
+        DocumentMetaTable.upsert_many(
+            conn,
+            [DocumentMetaRow(1, "a"), DocumentMetaRow(2, "b")],
+        )
+        DocumentMetaTable.delete_for_document(conn, 1)
+        assert DocumentMetaTable.all_modified_times(conn) == {"2": "b"}
+
+    def test_delete_all_clears_every_row(self, conn: sqlite3.Connection) -> None:
+        """
+        GIVEN:
+            - document_meta rows for multiple documents
+        WHEN:
+            - delete_all() is called
+        THEN:
+            - all_modified_times() returns an empty dict
+        """
+        DocumentMetaTable.create(conn)
+        DocumentMetaTable.upsert_many(
+            conn,
+            [DocumentMetaRow(1, "a"), DocumentMetaRow(2, "b")],
+        )
+        DocumentMetaTable.delete_all(conn)
+        assert DocumentMetaTable.all_modified_times(conn) == {}
+
+    def test_copy_all_streams_every_row_to_destination(
+        self,
+        conn: sqlite3.Connection,
+    ) -> None:
+        """
+        GIVEN:
+            - A source connection with document_meta rows for 5 documents
+            - A separate, empty destination connection
+        WHEN:
+            - copy_all() is called with a batch size smaller than the row
+              count, forcing multiple fetchmany() cycles
+        THEN:
+            - Every row is present on the destination connection
+        """
+        DocumentMetaTable.create(conn)
+        DocumentMetaTable.upsert_many(
+            conn,
+            [DocumentMetaRow(i, f"modified-{i}") for i in range(5)],
+        )
+        dst_conn = sqlite3.connect(":memory:")
+        dst_conn.row_factory = sqlite3.Row
+        try:
+            DocumentMetaTable.create(dst_conn)
+            DocumentMetaTable.copy_all(conn, dst_conn, batch_size=2)
+            assert DocumentMetaTable.all_modified_times(dst_conn) == {
+                str(i): f"modified-{i}" for i in range(5)
+            }
+        finally:
+            dst_conn.close()
+
+
+class TestIndexMetaTable:
+    @pytest.mark.parametrize(
+        ("setter_name", "getter_name", "value"),
+        [
+            ("set_dim", "get_dim", 384),
+            ("set_embed_model", "get_embed_model", "model-a"),
+            ("set_schema_version", "get_schema_version", 2),
+        ],
+    )
+    def test_typed_accessor_roundtrip(
+        self,
+        conn: sqlite3.Connection,
+        setter_name: str,
+        getter_name: str,
+        value: int | str,
+    ) -> None:
+        """
+        GIVEN:
+            - An empty index_meta table
+        WHEN:
+            - A typed accessor's setter is called then the getter is read back
+        THEN:
+            - The same value is returned, correctly typed (int or str)
+        """
+        IndexMetaTable.create(conn)
+        getter = getattr(IndexMetaTable, getter_name)
+        setter = getattr(IndexMetaTable, setter_name)
+        assert getter(conn) is None
+        setter(conn, value)
+        assert getter(conn) == value
+
+    def test_total_inserts_starts_at_zero(self, conn: sqlite3.Connection) -> None:
+        """
+        GIVEN:
+            - An empty index_meta table
+        WHEN:
+            - get_total_inserts() is read before anything is set
+        THEN:
+            - 0 is returned
+        """
+        IndexMetaTable.create(conn)
+        assert IndexMetaTable.get_total_inserts(conn) == 0
+
+    def test_increment_total_inserts_accumulates(
+        self,
+        conn: sqlite3.Connection,
+    ) -> None:
+        """
+        GIVEN:
+            - An empty index_meta table
+        WHEN:
+            - increment_total_inserts() is called twice
+        THEN:
+            - get_total_inserts() returns the running sum
+        """
+        IndexMetaTable.create(conn)
+        IndexMetaTable.increment_total_inserts(conn, 5)
+        IndexMetaTable.increment_total_inserts(conn, 3)
+        assert IndexMetaTable.get_total_inserts(conn) == 8
+
+    def test_reset_total_inserts_sets_absolute_value(
+        self,
+        conn: sqlite3.Connection,
+    ) -> None:
+        """
+        GIVEN:
+            - A total_inserts counter already at a high value
+        WHEN:
+            - reset_total_inserts() is called with a lower value
+        THEN:
+            - get_total_inserts() returns exactly that value, not a sum
+        """
+        IndexMetaTable.create(conn)
+        IndexMetaTable.increment_total_inserts(conn, 100)
+        IndexMetaTable.reset_total_inserts(conn, 7)
+        assert IndexMetaTable.get_total_inserts(conn) == 7
@@ -9,11 +9,11 @@ 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.migrations import MIGRATIONS
+from paperless_ai.migrations import Migration
 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

@@ -646,3 +646,50 @@ class TestMigrations:

        assert result is True
        assert self._schema_version(store) == 2
+
+    def test_has_pending_migration_false_when_no_table(
+        self,
+        store: PaperlessSqliteVecVectorStore,
+    ) -> None:
+        """
+        GIVEN:
+            - A vector store with no table created yet
+        WHEN:
+            - has_pending_migration() is checked
+        THEN:
+            - False is returned (nothing to migrate before anything exists)
+        """
+        assert store.has_pending_migration() is False
+
+    def test_has_pending_migration_false_at_current_version(
+        self,
+        store: PaperlessSqliteVecVectorStore,
+    ) -> None:
+        """
+        GIVEN:
+            - A store at the current SCHEMA_VERSION
+        WHEN:
+            - has_pending_migration() is checked
+        THEN:
+            - False is returned
+        """
+        store.add([make_node("a1", "1")])
+        assert store.has_pending_migration() is False
+
+    def test_has_pending_migration_true_when_behind(
+        self,
+        store: PaperlessSqliteVecVectorStore,
+    ) -> None:
+        """
+        GIVEN:
+            - A store whose schema_version has been forced behind SCHEMA_VERSION
+        WHEN:
+            - has_pending_migration() is checked
+        THEN:
+            - True is returned
+        """
+        store.add([make_node("a1", "1")])
+        store.client.execute(
+            "UPDATE index_meta SET value = '0' WHERE key = 'schema_version'",
+        )
+        assert store.has_pending_migration() is True
@@ -2,16 +2,12 @@ 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
@@ -26,6 +22,9 @@ 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

+from paperless_ai.migrations import MIGRATIONS
+from paperless_ai.migrations import Migration
+
 logger = logging.getLogger("paperless_ai.vector_store")

 DB_FILENAME = "llmindex.db"
@@ -53,38 +52,6 @@ COMPACT_BATCH_SIZE = 500
 _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)

@@ -551,6 +518,31 @@ class PaperlessSqliteVecVectorStore(BasePydanticVectorStore):
        Path(compact_path).replace(db_path)
        self._conn = self._open_connection(db_path)

+    def _stored_schema_version(self) -> int | None:
+        """The schema_version recorded in index_meta, or None if no table
+        exists. A missing key (a store predating version tracking) is
+        treated as SCHEMA_VERSION -- i.e. already current -- since no
+        migration in MIGRATIONS targets a version before tracking began.
+        """
+        if not self.table_exists():
+            return None
+        raw = self._meta_get("schema_version")
+        return int(raw) if raw is not None else SCHEMA_VERSION
+
+    def has_pending_migration(self) -> bool:
+        """Cheaply check whether a migration is pending, with no exclusive
+        access needed -- just a metadata read under the connection callers
+        already hold via the write FileLock.
+
+        Callers should only pay for check_and_run_migrations()'s exclusive
+        access (a structural migration's file swap must not run while
+        readers are active) when this returns True, so that the common
+        case -- already at SCHEMA_VERSION -- never contends with readers
+        or a concurrent compaction.
+        """
+        current = self._stored_schema_version()
+        return current is not None and current < SCHEMA_VERSION
+
    def check_and_run_migrations(self) -> bool:
        """Apply any pending schema migrations to the store.

@@ -559,15 +551,13 @@ class PaperlessSqliteVecVectorStore(BasePydanticVectorStore):
        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.
+        Must be called under the write FileLock, with readers excluded (see
+        has_pending_migration() for a cheap pre-check that avoids paying for
+        that exclusion in the common case).  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:
+        current = self._stored_schema_version()
+        if current is None or current >= SCHEMA_VERSION:
            return False

        pending = sorted(
@@ -579,7 +569,7 @@ class PaperlessSqliteVecVectorStore(BasePydanticVectorStore):
            if migration.kind == "re-embed":
                logger.warning(
                    "LLM index schema v%d -> v%d requires re-embedding (%s); "
-                    "forcing full rebuild.",
+                    "the caller must force a rebuild.",
                    migration.from_version,
                    migration.to_version,
                    migration.description,
@@ -1,8 +0,0 @@
-from django.apps import AppConfig
-from django.utils.translation import gettext_lazy as _
-
-
-class PaperlessBenchmarkConfig(AppConfig):
-    name = "paperless_benchmark"
-
-    verbose_name = _("Paperless benchmark")
@@ -1,136 +0,0 @@
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-from django.db import connection
-
-if TYPE_CHECKING:
-    from django.db.models import QuerySet
-
-
-def _reset_table_names() -> list[str]:
-    from guardian.models import GroupObjectPermission
-    from guardian.models import UserObjectPermission
-
-    from documents.models import Correspondent
-    from documents.models import Document
-    from documents.models import DocumentType
-    from documents.models import StoragePath
-    from documents.models import Tag
-
-    return [
-        Document.tags.through._meta.db_table,
-        Document._meta.db_table,
-        Tag._meta.db_table,
-        Correspondent._meta.db_table,
-        DocumentType._meta.db_table,
-        StoragePath._meta.db_table,
-        UserObjectPermission._meta.db_table,
-        GroupObjectPermission._meta.db_table,
-    ]
-
-
-def _delete_all_users_and_groups() -> None:
-    # ASSUMPTION: this tool assumes a disposable benchmark database, never
-    # point it at a real install. This deletes EVERY user and group in the
-    # database (not just benchmark-created ones) -- there is no way to
-    # distinguish "real" users from seeded ones, so this is only safe against
-    # a database that exists solely to run this benchmarking tool. The
-    # `benchmark seed --reset` CLI path requires an explicit
-    # `--yes-i-know-this-wipes-the-database` flag before reaching here; do
-    # not remove that guard.
-    from django.contrib.auth import get_user_model
-    from django.contrib.auth.models import Group
-
-    get_user_model().objects.all().delete()
-    Group.objects.all().delete()
-
-
-def _reset_postgresql() -> None:
-    tables = _reset_table_names()
-    with connection.cursor() as cursor:
-        cursor.execute(f"TRUNCATE TABLE {', '.join(tables)} RESTART IDENTITY CASCADE;")
-    _delete_all_users_and_groups()
-
-
-def _reset_mariadb() -> None:
-    # MariaDB's TRUNCATE has no CASCADE clause and refuses to truncate a
-    # table referenced by a foreign key while checks are enabled, so
-    # checks are disabled for the duration of the reset.
-    tables = _reset_table_names()
-    with connection.cursor() as cursor:
-        cursor.execute("SET FOREIGN_KEY_CHECKS = 0;")
-        try:
-            for table in tables:
-                cursor.execute(f"TRUNCATE TABLE {table};")
-        finally:
-            cursor.execute("SET FOREIGN_KEY_CHECKS = 1;")
-    _delete_all_users_and_groups()
-
-
-def _reset_sqlite() -> None:
-    from documents.models import Correspondent
-    from documents.models import Document
-    from documents.models import DocumentType
-    from documents.models import StoragePath
-    from documents.models import Tag
-
-    Document.global_objects.all().delete()
-    Tag.objects.all().delete()
-    Correspondent.objects.all().delete()
-    DocumentType.objects.all().delete()
-    StoragePath.objects.all().delete()
-    _delete_all_users_and_groups()
-
-
-def reset_benchmark_data() -> None:
-    """
-    Remove all previously-seeded benchmark data (documents, tags,
-    correspondents, document types, storage paths, guardian permission
-    rows, users, and groups) so a fresh `benchmark seed` run starts
-    from an empty slate. Dispatches per-backend because TRUNCATE syntax
-    and cascade behavior differ across the 3 supported databases.
-    """
-    if connection.vendor == "postgresql":
-        _reset_postgresql()
-    elif connection.vendor == "mysql":
-        # MariaDB also reports vendor == "mysql" under Django's mysql backend.
-        _reset_mariadb()
-    else:
-        _reset_sqlite()
-
-
-def capture_explain(queryset: QuerySet) -> str:
-    """
-    Return the query plan for `queryset` using the current backend's
-    explain facility. PostgreSQL supports `EXPLAIN ANALYZE {sql}` (real
-    execution stats). MariaDB does NOT accept that syntax -- verified
-    against a real MariaDB 12.3 container: `EXPLAIN ANALYZE {sql}` raises a
-    1064 syntax error, while MariaDB's own `ANALYZE {sql}` form (no
-    `EXPLAIN` keyword) works and returns real per-row execution stats
-    (`r_rows`, `r_filtered`, etc. columns) -- this is MariaDB's
-    EXPLAIN-ANALYZE-equivalent, distinct from MySQL 8.0.18+'s
-    `EXPLAIN ANALYZE` syntax, which MariaDB does not implement. SQLite only
-    supports EXPLAIN QUERY PLAN (the chosen plan, not real timing/row
-    counts) -- that output is clearly labeled rather than silently looking
-    equivalent to the other two backends' output.
-    """
-    sql, params = queryset.query.sql_with_params()
-    with connection.cursor() as cursor:
-        if connection.vendor == "postgresql":
-            cursor.execute(f"EXPLAIN ANALYZE {sql}", params)
-            return "\n".join(str(row[0]) for row in cursor.fetchall())
-        if connection.vendor == "mysql":
-            # MariaDB also reports vendor == "mysql" under Django's mysql
-            # backend. Unlike MySQL 8.0.18+, MariaDB has no `EXPLAIN
-            # ANALYZE` syntax -- its equivalent is `ANALYZE <statement>`.
-            cursor.execute(f"ANALYZE {sql}", params)
-            columns = [c[0] for c in cursor.description]
-            header = " | ".join(columns)
-            rows = "\n".join(
-                " | ".join(str(c) for c in row) for row in cursor.fetchall()
-            )
-            return f"{header}\n{rows}"
-        cursor.execute(f"EXPLAIN QUERY PLAN {sql}", params)
-        rows = "\n".join(" | ".join(str(c) for c in row) for row in cursor.fetchall())
-        return f"(plan only -- no execution stats on SQLite)\n{rows}"
@@ -1,81 +0,0 @@
-# src/paperless_benchmark/endpoints.py
-from __future__ import annotations
-
-import statistics
-import time
-from dataclasses import dataclass
-from typing import TYPE_CHECKING
-
-if TYPE_CHECKING:
-    from django.contrib.auth.models import User
-    from rest_framework.test import APIClient
-
-ENDPOINTS: tuple[tuple[str, str], ...] = (
-    ("documents_default", "/api/documents/"),
-    ("documents_page50", "/api/documents/?page_size=50"),
-    ("tags_all", "/api/tags/?page_size=100000"),
-)
-
-
-@dataclass(frozen=True, slots=True)
-class EndpointTiming:
-    user_label: str
-    endpoint_name: str
-    query_count: int
-    min_ms: float
-    median_ms: float
-    max_ms: float
-
-
-def _timed_requests(client: APIClient, url: str, n: int) -> list[float]:
-    times = []
-    for _ in range(n):
-        t0 = time.perf_counter()
-        resp = client.get(url)
-        t1 = time.perf_counter()
-        if resp.status_code != 200:
-            raise RuntimeError(
-                f"GET {url} -> {resp.status_code}: {resp.content[:300]!r}",
-            )
-        times.append(t1 - t0)
-    return times
-
-
-def _query_count(client: APIClient, url: str) -> int:
-    from django.db import connection
-    from django.test.utils import CaptureQueriesContext
-
-    with CaptureQueriesContext(connection) as ctx:
-        resp = client.get(url)
-    if resp.status_code != 200:
-        raise RuntimeError(f"GET {url} -> {resp.status_code}: {resp.content[:300]!r}")
-    return len(ctx.captured_queries)
-
-
-def run_endpoint_benchmarks(
-    *,
-    perf_target: User,
-    perf_admin: User,
-    repeat: int,
-) -> list[EndpointTiming]:
-    from rest_framework.test import APIClient
-
-    results: list[EndpointTiming] = []
-    for user_label, user in (("target", perf_target), ("admin", perf_admin)):
-        client = APIClient()
-        client.force_authenticate(user=user)
-        for name, url in ENDPOINTS:
-            client.get(url)  # warm-up request, not counted
-            qcount = _query_count(client, url)
-            times_ms = [t * 1000 for t in _timed_requests(client, url, repeat)]
-            results.append(
-                EndpointTiming(
-                    user_label=user_label,
-                    endpoint_name=name,
-                    query_count=qcount,
-                    min_ms=min(times_ms),
-                    median_ms=statistics.median(times_ms),
-                    max_ms=max(times_ms),
-                ),
-            )
-    return results
@@ -1,56 +0,0 @@
-# src/paperless_benchmark/harness.py
-from __future__ import annotations
-
-import time
-from dataclasses import dataclass
-from typing import TYPE_CHECKING
-from typing import Generic
-from typing import TypeVar
-
-from django.db import connection
-from django.test.utils import CaptureQueriesContext
-
-if TYPE_CHECKING:
-    from collections.abc import Callable
-
-T = TypeVar("T")
-
-
-@dataclass(frozen=True, slots=True)
-class ProfileResult(Generic[T]):
-    best_seconds: float
-    all_seconds: tuple[float, ...]
-    query_count: int
-    result: T
-
-
-def run_profile(fn: Callable[[], T], *, repeat: int = 5) -> ProfileResult[T]:
-    """
-    Call `fn` `repeat` times, capturing wall-clock time for every call and
-    the SQL query count for the final call. Returns the best (minimum)
-    time across all repeats, since the first call(s) can be skewed by
-    connection warm-up or cold caches.
-    """
-    if repeat < 1:
-        raise ValueError("repeat must be >= 1")
-
-    all_seconds: list[float] = []
-    result: T | None = None
-    query_count = 0
-    for i in range(repeat):
-        with CaptureQueriesContext(connection) as ctx:
-            start = time.perf_counter()
-            result = fn()
-            all_seconds.append(time.perf_counter() - start)
-        if i == repeat - 1:
-            query_count = len(ctx.captured_queries)
-    # Purely a type-narrowing aid for the type checker: the `repeat < 1`
-    # guard above already turns the one case that could leave `result`
-    # unset into a clear ValueError, so this is unreachable in practice.
-    assert result is not None
-    return ProfileResult(
-        best_seconds=min(all_seconds),
-        all_seconds=tuple(all_seconds),
-        query_count=query_count,
-        result=result,
-    )
@@ -1,222 +0,0 @@
-# src/paperless_benchmark/management/commands/benchmark.py
-from __future__ import annotations
-
-from typing import Any
-
-from django.core.management.base import BaseCommand
-from django.core.management.base import CommandError
-from django.core.management.base import CommandParser
-
-
-class Command(BaseCommand):
-    help = "Seed, run, and profile paperless-ngx performance benchmarks."
-
-    def add_arguments(self, parser: CommandParser) -> None:
-        parser.add_argument(
-            "action",
-            choices=["seed", "run", "profile", "list-scenarios"],
-            help="Which benchmark action to perform.",
-        )
-        parser.add_argument(
-            "scenario",
-            nargs="?",
-            default=None,
-            help="Scenario name (required for `profile`; see `list-scenarios`).",
-        )
-        parser.add_argument(
-            "--tier",
-            choices=["home", "medium", "large"],
-            default="medium",
-            help="Dataset scale tier for `seed` (default: medium).",
-        )
-        parser.add_argument(
-            "--reset",
-            action="store_true",
-            default=False,
-            help="For `seed`: truncate existing benchmark data first.",
-        )
-        parser.add_argument(
-            "--yes-i-know-this-wipes-the-database",
-            action="store_true",
-            default=False,
-            help=(
-                "Required alongside --reset: confirms you understand `seed "
-                "--reset` deletes ALL users, ALL groups, and ALL documents/"
-                "tags/correspondents/document types/storage paths in this "
-                "database, not just benchmark-created ones."
-            ),
-        )
-        parser.add_argument(
-            "--seed",
-            type=int,
-            default=42,
-            help="RNG seed for reproducible datasets (default: 42).",
-        )
-        parser.add_argument(
-            "--repeat",
-            type=int,
-            default=5,
-            help="Number of timed repetitions for `run`/`profile` (default: 5).",
-        )
-        parser.add_argument(
-            "--label",
-            default="baseline",
-            help="Free-text tag for a `run`, printed and recorded in history only.",
-        )
-        parser.add_argument(
-            "--explain",
-            action="store_true",
-            default=False,
-            help="For `profile`: also capture and print the query plan.",
-        )
-
-    def handle(self, *args: Any, **options: Any) -> None:
-        action = options["action"]
-        if action == "seed":
-            self._handle_seed(options)
-        elif action == "run":
-            self._handle_run(options)
-        elif action == "profile":
-            self._handle_profile(options)
-        else:
-            self._handle_list_scenarios()
-
-    def _handle_seed(self, options: dict[str, Any]) -> None:
-        from paperless_benchmark.db import reset_benchmark_data
-        from paperless_benchmark.seeding import seed_benchmark_dataset
-
-        if options["reset"]:
-            if not options["yes_i_know_this_wipes_the_database"]:
-                raise CommandError(
-                    "--reset requires --yes-i-know-this-wipes-the-database. "
-                    "This deletes ALL users, ALL groups, and ALL documents, "
-                    "tags, correspondents, document types, and storage paths "
-                    "in this database -- not just benchmark-created ones. "
-                    "Only run this against a disposable benchmark database, "
-                    "never a real install. Re-run with "
-                    "--reset --yes-i-know-this-wipes-the-database to proceed.",
-                )
-            self.stdout.write("Resetting existing benchmark data...")
-            reset_benchmark_data()
-
-        data = seed_benchmark_dataset(options["tier"], seed=options["seed"])
-        self.stdout.write(
-            self.style.SUCCESS(
-                f"Seeded tier={options['tier']!r}: {data.documents} documents, "
-                f"{len(data.users)} users, {len(data.groups)} groups.",
-            ),
-        )
-
-    def _handle_run(self, options: dict[str, Any]) -> None:
-        from django.contrib.auth import get_user_model
-        from django.db import connection
-
-        from documents.models import Document
-        from paperless_benchmark.endpoints import run_endpoint_benchmarks
-        from paperless_benchmark.results import append_history
-
-        user_model = get_user_model()
-        try:
-            perf_target = user_model.objects.get(username="perf_target")
-            perf_admin = user_model.objects.get(username="perf_admin")
-        except user_model.DoesNotExist as e:
-            raise CommandError(
-                "No benchmark dataset found. Run `manage.py benchmark seed` first.",
-            ) from e
-
-        db_vendor = connection.vendor
-        document_count = Document.objects.count()
-
-        results = run_endpoint_benchmarks(
-            perf_target=perf_target,
-            perf_admin=perf_admin,
-            repeat=options["repeat"],
-        )
-
-        self.stdout.write(f"# label={options['label']} repeat={options['repeat']}")
-        self.stdout.write(
-            f"{'user':7s} {'endpoint':20s} {'queries':>8s} "
-            f"{'min_ms':>9s} {'median_ms':>10s} {'max_ms':>9s}",
-        )
-        for r in results:
-            self.stdout.write(
-                f"{r.user_label:7s} {r.endpoint_name:20s} {r.query_count:8d} "
-                f"{r.min_ms:9.1f} {r.median_ms:10.1f} {r.max_ms:9.1f}",
-            )
-            append_history(
-                {
-                    "mode": "run",
-                    "label": options["label"],
-                    "user": r.user_label,
-                    "endpoint": r.endpoint_name,
-                    "query_count": r.query_count,
-                    "min_ms": r.min_ms,
-                    "median_ms": r.median_ms,
-                    "max_ms": r.max_ms,
-                    "db_vendor": db_vendor,
-                    "document_count": document_count,
-                },
-            )
-
-    def _handle_profile(self, options: dict[str, Any]) -> None:
-        from django.contrib.auth import get_user_model
-        from django.db import connection
-
-        from documents.models import Document
-        from paperless_benchmark.db import capture_explain
-        from paperless_benchmark.harness import run_profile
-        from paperless_benchmark.results import append_history
-        from paperless_benchmark.scenarios import get as get_scenario
-
-        if not options["scenario"]:
-            raise CommandError(
-                "`profile` requires a scenario name; see `list-scenarios`.",
-            )
-
-        scenario = get_scenario(options["scenario"])
-
-        user_model = get_user_model()
-        try:
-            perf_target = user_model.objects.get(username="perf_target")
-        except user_model.DoesNotExist as e:
-            raise CommandError(
-                "No benchmark dataset found. Run `manage.py benchmark seed` first.",
-            ) from e
-
-        profile = run_profile(
-            lambda: scenario.run(perf_target),
-            repeat=options["repeat"],
-        )
-        self.stdout.write(
-            f"{scenario.name}: best={profile.best_seconds:.4f}s "
-            f"queries={profile.query_count}",
-        )
-
-        if options["explain"]:
-            if scenario.queryset_for_explain is not None:
-                plan = capture_explain(scenario.queryset_for_explain(perf_target))
-                self.stdout.write(plan)
-            else:
-                self.stdout.write(
-                    self.style.WARNING(
-                        f"--explain was requested but scenario {scenario.name!r} "
-                        "does not support it (no queryset_for_explain); skipping.",
-                    ),
-                )
-
-        append_history(
-            {
-                "mode": "profile",
-                "scenario": scenario.name,
-                "best_seconds": profile.best_seconds,
-                "query_count": profile.query_count,
-                "db_vendor": connection.vendor,
-                "document_count": Document.objects.count(),
-            },
-        )
-
-    def _handle_list_scenarios(self) -> None:
-        from paperless_benchmark.scenarios import all_scenarios
-
-        for scenario in all_scenarios():
-            self.stdout.write(f"{scenario.name}: {scenario.describe}")
@@ -1,39 +0,0 @@
-# src/paperless_benchmark/results.py
-from __future__ import annotations
-
-import json
-import subprocess
-from datetime import UTC
-from datetime import datetime
-from pathlib import Path
-from typing import Any
-
-RESULTS_DIR = Path(__file__).resolve().parent.parent.parent / "benchmark_results"
-
-
-def _current_git_ref() -> str:
-    result = subprocess.run(
-        ["git", "rev-parse", "--short", "HEAD"],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-    return result.stdout.strip() or "unknown"
-
-
-def append_history(entry: dict[str, Any], *, code_ref: str | None = None) -> None:
-    """
-    Append one line to benchmark_results/history.jsonl -- a local-only,
-    append-only, cross-session record of every `benchmark run`/`profile`
-    invocation. Unlike a single overwritten snapshot file, this survives
-    across sessions so a benchmarking effort picked back up days later has
-    a full timeline instead of only the most recent result.
-    """
-    RESULTS_DIR.mkdir(exist_ok=True)
-    record = {
-        "timestamp": datetime.now(UTC).isoformat(),
-        "code_ref": code_ref or _current_git_ref(),
-        **entry,
-    }
-    with (RESULTS_DIR / "history.jsonl").open("a") as f:
-        f.write(json.dumps(record) + "\n")
@@ -1,110 +0,0 @@
-# src/paperless_benchmark/scenarios.py
-from __future__ import annotations
-
-from dataclasses import dataclass
-from typing import TYPE_CHECKING
-from typing import Any
-
-if TYPE_CHECKING:
-    from collections.abc import Callable
-
-    from django.contrib.auth.models import User
-    from django.db.models import QuerySet
-
-
-@dataclass(frozen=True, slots=True)
-class Scenario:
-    name: str
-    describe: str
-    run: Callable[[User], Any]
-    queryset_for_explain: Callable[[User], QuerySet] | None = None
-
-
-_SCENARIOS: dict[str, Scenario] = {}
-
-
-def register(scenario: Scenario) -> None:
-    _SCENARIOS[scenario.name] = scenario
-
-
-def get(name: str) -> Scenario:
-    from django.core.management.base import CommandError
-
-    try:
-        return _SCENARIOS[name]
-    except KeyError:
-        available = ", ".join(sorted(_SCENARIOS)) or "(none registered)"
-        raise CommandError(
-            f"Unknown benchmark scenario {name!r}. Available: {available}",
-        ) from None
-
-
-def all_scenarios() -> tuple[Scenario, ...]:
-    return tuple(_SCENARIOS.values())
-
-
-def _guardian_visibility_query_run(user: User) -> list[int]:
-    from documents.models import Document
-    from documents.permissions import get_objects_for_user_owner_aware
-
-    return list(
-        get_objects_for_user_owner_aware(
-            user,
-            "documents.view_document",
-            Document,
-        ).values_list("id", flat=True),
-    )
-
-
-def _guardian_visibility_query_queryset(user: User) -> QuerySet:
-    from documents.models import Document
-    from documents.permissions import get_objects_for_user_owner_aware
-
-    return get_objects_for_user_owner_aware(user, "documents.view_document", Document)
-
-
-register(
-    Scenario(
-        name="guardian_visibility_query",
-        describe=(
-            "Document-visibility queryset for a user with mixed owned/shared "
-            "documents -- exercises documents.permissions."
-            "get_objects_for_user_owner_aware's guardian permission join."
-        ),
-        run=_guardian_visibility_query_run,
-        queryset_for_explain=_guardian_visibility_query_queryset,
-    ),
-)
-
-
-def _permitted_document_ids_run(user: User) -> list[int]:
-    from documents.models import Document
-    from documents.permissions import permitted_document_ids
-
-    return list(
-        Document.objects.filter(id__in=permitted_document_ids(user)).values_list(
-            "id",
-            flat=True,
-        ),
-    )
-
-
-def _permitted_document_ids_queryset(user: User) -> QuerySet:
-    from documents.models import Document
-    from documents.permissions import permitted_document_ids
-
-    return Document.objects.filter(id__in=permitted_document_ids(user))
-
-
-register(
-    Scenario(
-        name="permitted_document_ids",
-        describe=(
-            "Document-visibility query built from documents.permissions."
-            "permitted_document_ids -- the resolved-ID-set alternative to "
-            "guardian_visibility_query, for side-by-side comparison."
-        ),
-        run=_permitted_document_ids_run,
-        queryset_for_explain=_permitted_document_ids_queryset,
-    ),
-)
@@ -1,445 +0,0 @@
-# src/paperless_benchmark/seeding.py
-from __future__ import annotations
-
-import datetime
-import random
-import time
-from dataclasses import dataclass
-from typing import TYPE_CHECKING
-from typing import Literal
-
-if TYPE_CHECKING:
-    from django.contrib.auth.models import Group
-    from django.contrib.auth.models import User
-
-Tier = Literal["home", "medium", "large"]
-
-CHUNK_SIZE = 5_000
-
-MIME_TYPES = (
-    "application/pdf",
-    "image/png",
-    "image/jpeg",
-    "text/plain",
-)
-
-# Ownership split for documents, mirroring the shape used in the #11950
-# perf-benchmark dataset (owned-by-target / owned-by-other / unowned).
-OWNED_BY_TARGET_FRACTION = 0.60
-OWNED_BY_OTHER_FRACTION = 0.30
-# remainder (0.10) is unowned
-
-# Of documents owned by "other" users, the fraction explicitly shared
-# (view, or view+change) with perf_target via guardian permissions --
-# this is what exercises the get_user_can_change() per-row N+1 that the
-# `run` endpoint benchmarks measure.
-SHARED_WITH_TARGET_FRACTION = 0.5
-SHARED_WITH_CHANGE_FRACTION = 0.5
-
-# Guardian permission-row ratios measured from a real install (discussion
-# #13276): 1,414 user-perm rows / 27,232 group-perm rows over 12,000
-# documents. Layered across ALL owned documents for the general user/group
-# pool (not just perf_target's shares), so `profile` scenarios exercise a
-# realistic permission-join shape for arbitrary users, not only perf_target.
-USER_PERM_ROWS_PER_DOC = 1_414 / 12_000
-GROUP_PERM_ROWS_PER_DOC = 27_232 / 12_000
-
-
-@dataclass(frozen=True, slots=True)
-class _TierCounts:
-    documents: int
-    tags: int
-    correspondents: int
-    document_types: int
-    storage_paths: int
-    other_users: int
-    groups: int
-    tags_per_doc: tuple[int, int]
-
-
-TIERS: dict[Tier, _TierCounts] = {
-    "home": _TierCounts(
-        documents=500,
-        tags=20,
-        correspondents=10,
-        document_types=8,
-        storage_paths=5,
-        other_users=3,
-        groups=2,
-        tags_per_doc=(1, 3),
-    ),
-    "medium": _TierCounts(
-        documents=20_000,
-        tags=100,
-        correspondents=300,
-        document_types=50,
-        storage_paths=20,
-        other_users=10,
-        groups=5,
-        tags_per_doc=(2, 6),
-    ),
-    "large": _TierCounts(
-        documents=360_000,
-        tags=1_000,
-        correspondents=5_000,
-        document_types=300,
-        storage_paths=50,
-        other_users=25,
-        groups=10,
-        tags_per_doc=(3, 7),
-    ),
-}
-
-
-def log(msg: str) -> None:
-    print(f"[{time.strftime('%H:%M:%S')}] {msg}", flush=True)  # noqa: T201
-
-
-@dataclass(frozen=True, slots=True)
-class SeededData:
-    """
-    Summary of a completed seed run. `documents`/`tags`/`correspondents`/
-    `document_types`/`storage_paths` are counts, not the seeded ORM
-    instances: at the `large` tier (360,000 documents) holding every
-    instance in memory simultaneously is a real risk for zero benefit, since
-    no caller consumes anything but the counts. `perf_target`/`perf_admin`/
-    `users`/`groups` stay as real objects -- at most ~26 users/12 groups even
-    at `large` tier, and small enough to be useful to a future caller.
-    """
-
-    perf_target: User
-    perf_admin: User
-    users: tuple[User, ...]
-    groups: tuple[Group, ...]
-    documents: int
-    tags: int
-    correspondents: int
-    document_types: int
-    storage_paths: int
-
-
-def _grant_model_level_permissions(user: User) -> None:
-    """
-    Grant perf_target Django model-level view/add/change permissions on
-    Document and Tag, on top of the per-object guardian grants seeding
-    creates elsewhere. DRF's PaperlessObjectPermissions checks model-level
-    permissions before guardian's object-level ones are ever consulted, so
-    without this perf_target gets a blanket 403 on /api/documents/ and
-    /api/tags/ regardless of which documents guardian says it can see.
-    """
-    from django.contrib.auth.models import Permission
-    from django.contrib.contenttypes.models import ContentType
-
-    from documents.models import Document
-    from documents.models import Tag
-
-    for model in (Document, Tag):
-        content_type = ContentType.objects.get_for_model(model)
-        codenames = [
-            f"{action}_{model._meta.model_name}" for action in ("view", "add", "change")
-        ]
-        perms = Permission.objects.filter(
-            content_type=content_type,
-            codename__in=codenames,
-        )
-        user.user_permissions.add(*perms)
-
-
-def _create_users_and_groups(
-    counts: _TierCounts,
-) -> tuple[User, User, tuple[User, ...], tuple[Group, ...]]:
-    from django.contrib.auth.models import Group
-
-    from documents.tests.factories import UserFactory
-
-    perf_target = UserFactory.create(username="perf_target")
-    _grant_model_level_permissions(perf_target)
-    perf_admin = UserFactory.create(username="perf_admin", superuser=True)
-    other_users = tuple(UserFactory.create_batch(counts.other_users))
-    groups = tuple(
-        Group.objects.create(name=f"benchmark_group_{i}") for i in range(counts.groups)
-    )
-    log(
-        f"Created users: 1 target, 1 superuser, {len(other_users)} other, "
-        f"{len(groups)} groups.",
-    )
-    return perf_target, perf_admin, other_users, groups
-
-
-def _create_lookup_tables(counts: _TierCounts):
-    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 CorrespondentFactory
-    from documents.tests.factories import DocumentTypeFactory
-    from documents.tests.factories import StoragePathFactory
-    from documents.tests.factories import TagFactory
-
-    tags = tuple(Tag.objects.bulk_create(TagFactory.build_batch(counts.tags)))
-    correspondents = tuple(
-        Correspondent.objects.bulk_create(
-            CorrespondentFactory.build_batch(counts.correspondents),
-        ),
-    )
-    document_types = tuple(
-        DocumentType.objects.bulk_create(
-            DocumentTypeFactory.build_batch(counts.document_types),
-        ),
-    )
-    storage_paths = tuple(
-        StoragePath.objects.bulk_create(
-            StoragePathFactory.build_batch(counts.storage_paths),
-        ),
-    )
-    log(
-        f"Created {len(tags)} tags, {len(correspondents)} correspondents, "
-        f"{len(document_types)} document types, {len(storage_paths)} storage paths.",
-    )
-    return tags, correspondents, document_types, storage_paths
-
-
-def _assign_owner(rng: random.Random, perf_target: User, other_users: tuple[User, ...]):
-    roll = rng.random()
-    if roll < OWNED_BY_TARGET_FRACTION:
-        return perf_target, "target"
-    if roll < OWNED_BY_TARGET_FRACTION + OWNED_BY_OTHER_FRACTION:
-        return rng.choice(other_users), "other"
-    return None, "unowned"
-
-
-def _seed_documents(
-    rng: random.Random,
-    counts: _TierCounts,
-    tags,
-    correspondents,
-    document_types,
-    storage_paths,
-    perf_target,
-    other_users,
-) -> tuple[int, list[int]]:
-    """
-    Bulk-create `counts.documents` documents in chunks. Returns the total
-    document count plus a lightweight list of pks for documents that ended
-    up with an owner (target or other) -- that's all
-    `_grant_general_permissions` needs to sample from, so full `Document`
-    instances aren't accumulated across chunks (a real memory concern at the
-    `large` tier's 360,000 documents).
-    """
-    from django.contrib.auth.models import Permission
-    from django.contrib.contenttypes.models import ContentType
-    from guardian.models import UserObjectPermission
-
-    from documents.models import Document
-    from documents.tests.factories import DocumentFactory
-
-    tag_ids = [t.pk for t in tags]
-    correspondent_ids = [c.pk for c in correspondents]
-    document_type_ids = [d.pk for d in document_types]
-    storage_path_ids = [s.pk for s in storage_paths]
-
-    doc_content_type = ContentType.objects.get_for_model(Document)
-    view_perm = Permission.objects.get(
-        codename="view_document",
-        content_type=doc_content_type,
-    )
-    change_perm = Permission.objects.get(
-        codename="change_document",
-        content_type=doc_content_type,
-    )
-    through_model = Document.tags.through
-
-    document_count = 0
-    owned_document_pks: list[int] = []
-    remaining = counts.documents
-    while remaining > 0:
-        chunk_n = min(CHUNK_SIZE, remaining)
-        remaining -= chunk_n
-
-        batch = []
-        owner_buckets = []
-        for _ in range(chunk_n):
-            doc = DocumentFactory.build(
-                mime_type=rng.choice(MIME_TYPES),
-                page_count=rng.randint(1, 30),
-                correspondent_id=(
-                    rng.choice(correspondent_ids)
-                    if correspondent_ids and rng.random() < 0.8
-                    else None
-                ),
-                document_type_id=(
-                    rng.choice(document_type_ids)
-                    if document_type_ids and rng.random() < 0.6
-                    else None
-                ),
-                storage_path_id=(
-                    rng.choice(storage_path_ids)
-                    if storage_path_ids and rng.random() < 0.8
-                    else None
-                ),
-                # Document.created is a plain DateField (default: today).
-                # Leaving it unset would give every seeded document the same
-                # date, collapsing Document's ("-created",) ordering index
-                # into a single-valued sort key -- spread it over a
-                # realistic multi-year window instead.
-                created=datetime.date.today()
-                - datetime.timedelta(days=rng.randint(0, 365 * 3)),
-            )
-            owner, bucket = _assign_owner(rng, perf_target, other_users)
-            doc.owner_id = owner.pk if owner else None
-            batch.append(doc)
-            owner_buckets.append(bucket)
-
-        created = Document.objects.bulk_create(batch, batch_size=CHUNK_SIZE)
-
-        through_rows = []
-        for doc in created:
-            k = rng.randint(*counts.tags_per_doc)
-            for tag_id in rng.sample(tag_ids, min(k, len(tag_ids))):
-                through_rows.append(through_model(document_id=doc.pk, tag_id=tag_id))
-        if through_rows:
-            through_model.objects.bulk_create(through_rows, batch_size=CHUNK_SIZE)
-
-        perm_rows = []
-        for doc, bucket in zip(created, owner_buckets, strict=True):
-            if bucket == "unowned":
-                continue
-            owned_document_pks.append(doc.pk)
-
-            if bucket != "other":
-                continue
-            if rng.random() >= SHARED_WITH_TARGET_FRACTION:
-                continue
-            perm_rows.append(
-                UserObjectPermission(
-                    permission=view_perm,
-                    content_type=doc_content_type,
-                    object_pk=str(doc.pk),
-                    user=perf_target,
-                ),
-            )
-            if rng.random() < SHARED_WITH_CHANGE_FRACTION:
-                perm_rows.append(
-                    UserObjectPermission(
-                        permission=change_perm,
-                        content_type=doc_content_type,
-                        object_pk=str(doc.pk),
-                        user=perf_target,
-                    ),
-                )
-        if perm_rows:
-            UserObjectPermission.objects.bulk_create(perm_rows, batch_size=CHUNK_SIZE)
-
-        document_count += len(created)
-        log(f"  {document_count}/{counts.documents} documents seeded")
-
-    return document_count, owned_document_pks
-
-
-def _grant_general_permissions(
-    rng: random.Random,
-    owned_document_pks: list[int],
-    users,
-    groups,
-) -> None:
-    """
-    Layer realistic (issue #13276-derived) guardian permission-row ratios
-    across owned documents for the general user/group pool, so `profile`
-    scenarios exercise the same permission-join shape regardless of which
-    user they check visibility for (not just perf_target).
-    """
-    from django.contrib.auth.models import Permission
-    from django.contrib.contenttypes.models import ContentType
-    from guardian.models import GroupObjectPermission
-    from guardian.models import UserObjectPermission
-
-    from documents.models import Document
-
-    if not owned_document_pks or not users:
-        return
-
-    doc_content_type = ContentType.objects.get_for_model(Document)
-    view_perm = Permission.objects.get(
-        codename="view_document",
-        content_type=doc_content_type,
-    )
-
-    n_user_perms = round(len(owned_document_pks) * USER_PERM_ROWS_PER_DOC)
-    n_group_perms = (
-        round(len(owned_document_pks) * GROUP_PERM_ROWS_PER_DOC) if groups else 0
-    )
-
-    user_rows = [
-        UserObjectPermission(
-            permission=view_perm,
-            content_type=doc_content_type,
-            object_pk=str(rng.choice(owned_document_pks)),
-            user=rng.choice(users),
-        )
-        for _ in range(n_user_perms)
-    ]
-    if user_rows:
-        UserObjectPermission.objects.bulk_create(
-            user_rows,
-            batch_size=CHUNK_SIZE,
-            ignore_conflicts=True,
-        )
-
-    group_rows = [
-        GroupObjectPermission(
-            permission=view_perm,
-            content_type=doc_content_type,
-            object_pk=str(rng.choice(owned_document_pks)),
-            group=rng.choice(groups),
-        )
-        for _ in range(n_group_perms)
-    ]
-    if group_rows:
-        GroupObjectPermission.objects.bulk_create(
-            group_rows,
-            batch_size=CHUNK_SIZE,
-            ignore_conflicts=True,
-        )
-
-    log(f"  Granted {len(user_rows)} user perms, {len(group_rows)} group perms.")
-
-
-def seed_benchmark_dataset(tier: Tier, *, seed: int = 42) -> SeededData:
-    """
-    Build a benchmark dataset at the given scale tier: a named perf_target
-    (mixed owned/shared documents) and perf_admin (superuser) for endpoint
-    benchmarking, plus a general user/group pool with realistic guardian
-    permission-row ratios for profile scenarios.
-    """
-    counts = TIERS[tier]
-    rng = random.Random(seed)
-
-    log(f"Seeding tier={tier!r}")
-    perf_target, perf_admin, other_users, groups = _create_users_and_groups(counts)
-    tags, correspondents, document_types, storage_paths = _create_lookup_tables(counts)
-    document_count, owned_document_pks = _seed_documents(
-        rng,
-        counts,
-        tags,
-        correspondents,
-        document_types,
-        storage_paths,
-        perf_target,
-        other_users,
-    )
-    all_users = (perf_target, *other_users)
-    _grant_general_permissions(rng, owned_document_pks, all_users, groups)
-
-    log(f"Done. {document_count} documents seeded for tier={tier!r}.")
-
-    return SeededData(
-        perf_target=perf_target,
-        perf_admin=perf_admin,
-        users=all_users,
-        groups=groups,
-        documents=document_count,
-        tags=len(tags),
-        correspondents=len(correspondents),
-        document_types=len(document_types),
-        storage_paths=len(storage_paths),
-    )
Author	SHA1	Message	Date
stumpylogandClaude Sonnet 5	631c521827	Fix: address code review feedback on table gateways - Document sqlite3.Row row_factory precondition in module docstring - Strengthen test_create_is_idempotent to verify populated table survives - Collapse three roundtrip tests into one @pytest.mark.parametrize Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>	2026-07-30 14:29:34 -07:00
stumpylogandClaude Sonnet 5	22b4ce7938	Feature: add table-gateway module for document_chunks/document_meta/index_meta Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>	2026-07-30 14:29:34 -07:00
Trenton H	424a148482	Refactor: extract migration infrastructure, add has_pending_migration() (#13410 )	2026-07-30 14:28:21 -07:00
Trenton HandGitHub	e79f0d4106	Performance: Use server side iterators during LLM index updating (#13430 )	2026-07-30 08:53:01 -07:00
GitHub Actions	c29c9178f0	Auto translate strings	2026-07-30 14:39:47 +00:00
shamoonandGitHub	baae99eb49	Fix: normalize monetary decimal symbol by locale (#13427 )	2026-07-30 07:37:52 -07:00
Trenton HandGitHub	f50a021ef6	Fix: fold overflowing path text in sanity checker (#13426 )	2026-07-30 14:35:19 +00:00
shamoonandGitHub	c8e00dca04	Fix: correct missing add field button for custom fields (#13424 )	2026-07-30 07:16:25 -07:00
shamoonandGitHub	0a56018d6b	Fix: hide attributes collapse / show button without UI settings (#13425 )	2026-07-30 07:08:50 -07:00
Trenton H GitHub Claude Sonnet 5	350684cd6b	Fix: consolidate born-digital PDF detection between archive decision and OCR (#13409 ) * Fix: unify born-digital PDF detection between archive decision and OCR should_produce_archive() and RasterisedDocumentParser.parse() each reimplemented the "does this PDF have real text" check independently, using different normalization of pdftotext output. Raw pdftotext output can be non-empty (whitespace/form-feed layout padding) even when there is no real content, so the two checks could disagree: consumer.py treated a tagged-but-textless PDF as born-digital and skipped the archive, while the parser's own (stricter, normalized) check found no text and ran OCR anyway, leaving the document with no archive despite real OCR text (GH #13387). Both call sites now share one predicate, pdf_born_digital_text() in paperless/parsers/utils.py, so they can no longer drift apart. * Fix: restore extract_text seam for born-digital detection in parse() parse() had switched to calling pdf_born_digital_text() directly for its initial text/born-digital check, bypassing the parser's own extract_text instance method. That broke test mockability (tests patch tesseract_parser.extract_text to control the born-digital decision) and caused CI failures with mismatched OCR call counts and text. Split pdf_born_digital_text() into is_born_digital_text(text, path, log) - a pure decision function - and a thin pdf_born_digital_text() wrapper for callers without text in hand (consumer.should_produce_archive). parse() now extracts via self.extract_text(None, document_path) and passes the result to is_born_digital_text(), restoring the seam with no change to production behavior. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * Cleanup: simplify born-digital detection, close #13387 test gap Simplification pass over the born-digital detection consolidation: - is_born_digital_text(): drop the has_text temp for an early return. - consumer.py: standardize the archive-decision log lines on plain hyphens (was a mix of em-dash and hyphen) and hoist the duplicated text_length computation. - Parametrize TestPdfBornDigitalText instead of four near-identical tests. Code review follow-up: the existing tests only ever exercised pdf_born_digital_text() through mocks, so the actual #13387 scenario (a tagged PDF whose only "text" is layout padding) was never checked against real pdftotext/pikepdf output - a regression in the normalize-before-decide logic itself would have gone undetected. Moved tagged_no_text_pdf_file from parsers/conftest.py up to the shared paperless/tests/conftest.py (it was previously only visible to tests under parsers/) and added a non-mocked regression test against the real sample file. Also fixed two stale comments in test_consumer.py referencing a _extract_text_for_archive_check helper that no longer exists. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>	2026-07-29 12:41:09 -07:00
				`@@ -0,0 +1 @@`
				`/etc/s6-overlay/s6-rc.d/init-llmindex-migrate/run`