amorfo77/paperless-ngx
1
0
Fork 0
mirror of https://github.com/paperless-ngx/paperless-ngx.git synced 2026-09-14 21:58:00 +00:00
Code Issues Projects Releases Wiki Activity

Enhancement (dev): Use OpenAI-like backend (#12668)

Browse Source
This commit is contained in:
shamoon
2026-04-28 10:06:59 -07:00
committed by GitHub
parent 2f8f126223
commit 69cb4d06c6
17 changed files with 136 additions and 83 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/paperless_ai/client.py
+10 -6
View File
@@ -1,10 +1,12 @@
import logging
from typing import TYPE_CHECKING
from paperless.models import LLMBackend
if TYPE_CHECKING:
from llama_index.core.llms import ChatMessage
from llama_index.llms.ollama import Ollama
from llama_index.llms.openai import OpenAI
from llama_index.llms.openai_like import OpenAILike
from paperless.config import AIConfig
from paperless.network import validate_outbound_http_url
@@ -22,8 +24,8 @@ class AIClient:
self.settings = AIConfig()
self.llm = self.get_llm()
def get_llm(self) -> "Ollama | OpenAI":
if self.settings.llm_backend == "ollama":
def get_llm(self) -> "Ollama | OpenAILike":
if self.settings.llm_backend == LLMBackend.OLLAMA:
from llama_index.llms.ollama import Ollama
endpoint = self.settings.llm_endpoint or "http://localhost:11434"
@@ -36,8 +38,8 @@ class AIClient:
base_url=endpoint,
request_timeout=120,
)
elif self.settings.llm_backend == "openai":
from llama_index.llms.openai import OpenAI
elif self.settings.llm_backend == LLMBackend.OPENAI_LIKE:
from llama_index.llms.openai_like import OpenAILike
endpoint = self.settings.llm_endpoint or None
if endpoint:
@@ -45,10 +47,12 @@ class AIClient:
endpoint,
allow_internal=self.settings.llm_allow_internal_endpoints,
)
return OpenAI(
return OpenAILike(
model=self.settings.llm_model or "gpt-3.5-turbo",
api_base=endpoint,
api_key=self.settings.llm_api_key,
is_chat_model=True,
is_function_calling_model=True,
)
else:
raise ValueError(f"Unsupported LLM backend: {self.settings.llm_backend}")
src/paperless_ai/embedding.py
+5 -5
View File
@@ -19,8 +19,8 @@ def get_embedding_model() -> "BaseEmbedding":
config = AIConfig()
match config.llm_embedding_backend:
case LLMEmbeddingBackend.OPENAI:
from llama_index.embeddings.openai import OpenAIEmbedding
case LLMEmbeddingBackend.OPENAI_LIKE:
from llama_index.embeddings.openai_like import OpenAILikeEmbedding
endpoint = config.llm_endpoint or None
if endpoint:
@@ -28,8 +28,8 @@ def get_embedding_model() -> "BaseEmbedding":
endpoint,
allow_internal=config.llm_allow_internal_endpoints,
)
return OpenAIEmbedding(
model=config.llm_embedding_model or "text-embedding-3-small",
return OpenAILikeEmbedding(
model_name=config.llm_embedding_model or "text-embedding-3-small",
api_key=config.llm_api_key,
api_base=endpoint,
)
@@ -54,7 +54,7 @@ def get_embedding_dim() -> int:
config = AIConfig()
model = config.llm_embedding_model or (
"text-embedding-3-small"
if config.llm_embedding_backend == "openai"
if config.llm_embedding_backend == LLMEmbeddingBackend.OPENAI_LIKE
else "sentence-transformers/all-MiniLM-L6-v2"
)
src/paperless_ai/tests/test_ai_indexing.py
+1 -1
View File
@@ -98,7 +98,7 @@ def test_update_llm_index_removes_meta(
config = AIConfig()
expected_model = config.llm_embedding_model or (
"text-embedding-3-small"
if config.llm_embedding_backend == "openai"
if config.llm_embedding_backend == "openai-like"
else "sentence-transformers/all-MiniLM-L6-v2"
)
assert meta == {"embedding_model": expected_model, "dim": 384}
src/paperless_ai/tests/test_client.py
+6 -4
View File
@@ -25,8 +25,8 @@ def mock_ollama_llm():
@pytest.fixture
def mock_openai_llm():
with patch("llama_index.llms.openai.OpenAI") as MockOpenAI:
yield MockOpenAI
with patch("llama_index.llms.openai_like.OpenAILike") as MockOpenAILike:
yield MockOpenAILike
def test_get_llm_ollama(mock_ai_config, mock_ollama_llm):
@@ -45,7 +45,7 @@ def test_get_llm_ollama(mock_ai_config, mock_ollama_llm):
def test_get_llm_openai(mock_ai_config, mock_openai_llm):
mock_ai_config.llm_backend = "openai"
mock_ai_config.llm_backend = "openai-like"
mock_ai_config.llm_model = "test_model"
mock_ai_config.llm_api_key = "test_api_key"
mock_ai_config.llm_endpoint = "http://test-url"
@@ -56,12 +56,14 @@ def test_get_llm_openai(mock_ai_config, mock_openai_llm):
model="test_model",
api_base="http://test-url",
api_key="test_api_key",
is_chat_model=True,
is_function_calling_model=True,
)
assert client.llm == mock_openai_llm.return_value
def test_get_llm_openai_blocks_internal_endpoint_when_disallowed(mock_ai_config):
mock_ai_config.llm_backend = "openai"
mock_ai_config.llm_backend = "openai-like"
mock_ai_config.llm_model = "test_model"
mock_ai_config.llm_api_key = "test_api_key"
mock_ai_config.llm_endpoint = "http://127.0.0.1:1234"
src/paperless_ai/tests/test_embedding.py
+9 -7
View File
@@ -54,15 +54,17 @@ def mock_document():
def test_get_embedding_model_openai(mock_ai_config):
mock_ai_config.return_value.llm_embedding_backend = LLMEmbeddingBackend.OPENAI
mock_ai_config.return_value.llm_embedding_backend = LLMEmbeddingBackend.OPENAI_LIKE
mock_ai_config.return_value.llm_embedding_model = "text-embedding-3-small"
mock_ai_config.return_value.llm_api_key = "test_api_key"
mock_ai_config.return_value.llm_endpoint = "http://test-url"
with patch("llama_index.embeddings.openai.OpenAIEmbedding") as MockOpenAIEmbedding:
with patch(
"llama_index.embeddings.openai_like.OpenAILikeEmbedding",
) as MockOpenAIEmbedding:
model = get_embedding_model()
MockOpenAIEmbedding.assert_called_once_with(
model="text-embedding-3-small",
model_name="text-embedding-3-small",
api_key="test_api_key",
api_base="http://test-url",
)
@@ -72,7 +74,7 @@ def test_get_embedding_model_openai(mock_ai_config):
def test_get_embedding_model_openai_blocks_internal_endpoint_when_disallowed(
mock_ai_config,
):
mock_ai_config.return_value.llm_embedding_backend = LLMEmbeddingBackend.OPENAI
mock_ai_config.return_value.llm_embedding_backend = LLMEmbeddingBackend.OPENAI_LIKE
mock_ai_config.return_value.llm_embedding_model = "text-embedding-3-small"
mock_ai_config.return_value.llm_api_key = "test_api_key"
mock_ai_config.return_value.llm_endpoint = "http://127.0.0.1:11434"
@@ -109,7 +111,7 @@ def test_get_embedding_model_invalid_backend(mock_ai_config):
def test_get_embedding_dim_infers_and_saves(temp_llm_index_dir, mock_ai_config):
mock_ai_config.return_value.llm_embedding_backend = "openai"
mock_ai_config.return_value.llm_embedding_backend = "openai-like"
mock_ai_config.return_value.llm_embedding_model = None
class DummyEmbedding:
@@ -129,7 +131,7 @@ def test_get_embedding_dim_infers_and_saves(temp_llm_index_dir, mock_ai_config):
def test_get_embedding_dim_reads_existing_meta(temp_llm_index_dir, mock_ai_config):
mock_ai_config.return_value.llm_embedding_backend = "openai"
mock_ai_config.return_value.llm_embedding_backend = "openai-like"
mock_ai_config.return_value.llm_embedding_model = None
(temp_llm_index_dir / "meta.json").write_text(
@@ -142,7 +144,7 @@ def test_get_embedding_dim_reads_existing_meta(temp_llm_index_dir, mock_ai_confi
def test_get_embedding_dim_raises_on_model_change(temp_llm_index_dir, mock_ai_config):
mock_ai_config.return_value.llm_embedding_backend = "openai"
mock_ai_config.return_value.llm_embedding_backend = "openai-like"
mock_ai_config.return_value.llm_embedding_model = None
(temp_llm_index_dir / "meta.json").write_text(