gmakstutis/LightRAG
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
5950 commits 5 branches 72 tags 705 MiB
Commit graph

141 commits

Author SHA1 Message Date
yangdx
4ab4a7ac94 Allow embedding models to use provider defaults when unspecified 
- Set EMBEDDING_MODEL default to None
- Pass model param only when provided
- Let providers use their own defaults
- Fix lollms embed function params
- Add ollama embed_model default param
 2025-11-28 16:57:33 +08:00
yangdx
56e0365cf0 Add configurable model parameter to jina_embed function 
- Add model parameter to jina_embed
- Pass model from API server
- Default to jina-embeddings-v4
- Update function documentation
- Make model selection flexible
 2025-11-28 15:38:29 +08:00
yangdx
6e2946e78a Add max_token_size parameter to azure_openai_embed wrapper 2025-11-28 13:41:01 +08:00
EightyOliveira
8994c70f2f fix:exception handling order error 2025-11-25 16:36:41 +08:00
yangdx
7b76211066 Add fallback to AZURE_OPENAI_API_VERSION for embedding API version 2025-11-22 00:14:35 +08:00
yangdx
ffd8da512e Improve Azure OpenAI compatibility and error handling 
• Reduce log noise for Azure content filters
• Add default API version fallback
• Change warning to debug log level
• Handle empty choices in streaming
• Better Azure OpenAI integration
 2025-11-21 23:51:18 +08:00
yangdx
fafa1791f4 Fix Azure OpenAI model parameter to use deployment name consistently 
- Use deployment name for Azure API calls
- Fix model param in embed function
- Consistent api_model logic
- Prevent Azure model name conflicts
 2025-11-21 23:41:52 +08:00
yangdx
ac9f2574a5 Improve Azure OpenAI wrapper functions with full parameter support 
• Add missing parameters to wrappers
• Update docstrings for clarity
• Ensure API consistency
• Fix parameter forwarding
• Maintain backward compatibility
 2025-11-21 19:24:32 +08:00
yangdx
45f4f82392 Refactor Azure OpenAI client creation to support client_configs merging 
- Handle None client_configs case
- Merge configs with explicit params
- Override client_configs with params
- Use dict unpacking for client init
- Maintain parameter precedence
 2025-11-21 19:14:16 +08:00
yangdx
0c4cba3860 Fix double decoration in azure_openai_embed and document decorator usage 
• Remove redundant @retry decorator
• Call openai_embed.func directly
• Add detailed decorator documentation
• Prevent double parameter injection
• Fix EmbeddingFunc wrapping issues
 2025-11-21 18:03:53 +08:00
yangdx
b46c152306 Fix linting 2025-11-21 17:16:44 +08:00
yangdx
b709f8f869 Consolidate Azure OpenAI implementation into main OpenAI module 
• Unified OpenAI/Azure client creation
• Azure module now re-exports functions
• Backward compatibility maintained
• Reduced code duplication
 2025-11-21 17:12:33 +08:00
yangdx
02fdceb959 Update OpenAI client to use stable API and bump minimum version to 2.0.0 
- Remove beta prefix from completions.parse
- Update OpenAI dependency to >=2.0.0
- Fix whitespace formatting
- Update all requirement files
- Clean up pyproject.toml dependencies
 2025-11-21 12:55:44 +08:00
yangdx
9f69c5bf85 feat: Support structured output parsed from OpenAI 
Added support for structured output (JSON mode) from the OpenAI API in `openai.py` and `azure_openai.py`.

When `response_format` is used to request structured data, the new logic checks for the `message.parsed` attribute. If it exists, it's serialized into a JSON string as the final content. If not, the code falls back to the existing `message.content` handling, ensuring backward compatibility.
 2025-11-21 12:46:31 +08:00
yangdx
c9e1c86e81 Refactor keyword extraction handling to centralize response format logic 
• Move response format to core function
• Remove duplicate format assignments
• Standardize keyword extraction flow
• Clean up redundant parameter handling
• Improve Azure OpenAI compatibility
 2025-11-21 12:10:04 +08:00
yangdx
46ce6d9a13 Fix Azure OpenAI embedding model parameter fallback 
- Use model param if provided
- Fall back to deployment name
- Fix embedding API call
- Improve parameter handling
 2025-11-20 18:20:22 +08:00
Amritpal Singh
30e86fa331 use deployment variable which extracted value from .env file or have default value 2025-11-20 09:00:27 +00:00
yangdx
680e36c6eb Improve Bedrock error handling with retry logic and custom exceptions 
• Add specific exception types
• Implement proper retry mechanism
• Better error classification
• Enhanced logging and validation
• Enable embedding retry decorator
 2025-11-14 18:51:41 +08:00
yangdx
05852e1ab2 Add max_token_size parameter to embedding function decorators 
- Add max_token_size=8192 to all embed funcs
- Move siliconcloud to deprecated folder
- Import wrap_embedding_func_with_attrs
- Update EmbeddingFunc docstring
- Fix langfuse import type annotation
 2025-11-14 18:41:43 +08:00
yangdx
77ad906d3a Improve error handling and logging in cloud model detection 2025-11-13 20:41:44 +08:00
LacombeLouis
844537e378 Add a better regex 2025-11-13 12:17:51 +01:00
Louis Lacombe
f7432a260e Add support for environment variable fallback for API key and default host for cloud models 2025-11-12 16:11:05 +00:00
yangdx
2f16065256 Refactor keyword_extraction from kwargs to explicit parameter 
• Add keyword_extraction param to functions
• Remove kwargs.pop() calls
• Update function signatures
• Improve parameter documentation
• Make parameter handling consistent
 2025-11-09 12:02:17 +08:00
yangdx
88ab73f6ae HotFix: Restore streaming response in OpenAI LLM 
The stream and timeout parameters were moved from **kwargs to explicit
parameters in a previous commit, but were not being passed to the OpenAI
API, causing streaming responses to fail and fall back to non-streaming
mode.Fixes the issue where stream=True was being silently ignored, resulting
in unexpected non-streaming behavior.
 2025-11-09 11:52:26 +08:00
yangdx
3d9de5ed03 feat: improve Gemini client error handling and retry logic 
• Add google-api-core dependency
• Add specific exception handling
• Create InvalidResponseError class
• Update retry decorators
• Fix empty response handling
 2025-11-08 22:10:09 +08:00
yangdx
de4ed73652 Add Gemini embedding support 
- Implement gemini_embed function
- Add gemini to embedding binding choices
- Add L2 normalization for dims < 3072
 2025-11-08 03:34:30 +08:00
yangdx
f83ea3394e Add section header comment for Gemini binding options 2025-11-08 02:07:31 +08:00
yangdx
ffeeae4208 refactor: simplify jina embedding dimension handling 2025-11-07 22:09:57 +08:00
yangdx
01b07b2be5 Refactor Jina embedding dimension by changing param to optional with default 2025-11-07 22:04:34 +08:00
yangdx
d8a6355e41 Merge branch 'main' into apply-dim-to-embedding-call 2025-11-07 20:48:22 +08:00
yangdx
33a1482f7f Add optional embedding dimension parameter control via env var 
* Add EMBEDDING_SEND_DIM environment variable
* Update Jina/OpenAI embed functions
* Add send_dimensions to EmbeddingFunc
* Auto-inject embedding_dim when enabled
* Add parameter validation warnings
 2025-11-07 20:46:40 +08:00
yangdx
fc40a36968 Add timeout support to Gemini LLM and improve parameter handling 
• Add timeout parameter to Gemini client
• Convert timeout seconds to milliseconds
• Update function signatures consistently
• Add Gemini thinking config example
• Clean up parameter documentation
 2025-11-07 15:50:14 +08:00
yangdx
3cb4eae492 Add Chain of Thought support to Gemini LLM integration 
- Extract thoughts from response parts
- Add COT enable/disable parameter
 2025-11-07 15:22:14 +08:00
yangdx
6686edfd35 Update Gemini LLM options: add seed and thinking config, remove MIME type 2025-11-07 14:32:42 +08:00
Yasiru Rangana
d94aae9c5e Add dimensions parameter support to openai_embed() 2025-11-07 09:55:06 +11:00
yangdx
8c27555358 Fix Gemini response parsing to avoid warnings from non-text parts 2025-11-07 04:00:37 +08:00
yangdx
6e36ff41e1 Fix linting 2025-11-06 16:01:24 +08:00
yangdx
5f49cee20f Merge branch 'main' into VOXWAVE-FOUNDRY/main 2025-11-06 15:37:35 +08:00
yangdx
10f6e6955f Improve Langfuse integration and stream response cleanup handling 
• Check env vars before enabling Langfuse
• Move imports after env check logic
• Handle wrapper client aclose() issues
• Add debug logs for cleanup failures
 2025-11-03 13:09:45 +08:00
anouarbm
9495778c2d refactor: reorder Langfuse import logic for improved clarity 
Moved logger import before Langfuse block to fix NameError.
 2025-11-03 05:27:41 +01:00
anouarbm
626b42bc40 feat: add optional Langfuse observability integration 
This contribution adds optional Langfuse support for LLM observability and tracing.
Langfuse provides a drop-in replacement for the OpenAI client that automatically
tracks all LLM interactions without requiring code changes.

Features:
- Optional Langfuse integration with graceful fallback
- Automatic LLM request/response tracing
- Token usage tracking
- Latency metrics
- Error tracking
- Zero code changes required for existing functionality

Implementation:
- Modified lightrag/llm/openai.py to conditionally use Langfuse's AsyncOpenAI
- Falls back to standard OpenAI client if Langfuse is not installed
- Logs observability status on import

Configuration:
To enable Langfuse tracing, install the observability extras and set environment variables:

```bash
pip install lightrag-hku[observability]

export LANGFUSE_PUBLIC_KEY="your_public_key"
export LANGFUSE_SECRET_KEY="your_secret_key"
export LANGFUSE_HOST="https://cloud.langfuse.com"  # or your self-hosted instance
```

If Langfuse is not installed or environment variables are not set, LightRAG
will use the standard OpenAI client without any functionality changes.

Changes:
- Modified lightrag/llm/openai.py (added optional Langfuse import)
- Updated pyproject.toml with optional 'observability' dependencies

Dependencies (optional):
- langfuse>=3.8.1
 2025-11-01 21:40:22 +01:00
Humphry
0b3d31507e extended to use gemini, sswitched to use gemini-flash-latest 2025-10-20 13:17:16 +03:00
yangdx
a5c05f1b92 Add offline deployment support with cache management and layered deps 
• Add tiktoken cache downloader CLI
• Add layered offline dependencies
• Add offline requirements files
• Add offline deployment guide
 2025-10-11 10:28:14 +08:00
Yasiru Rangana
ae9f4ae73f fix: Remove trailing whitespace for pre-commit linting 2025-10-09 15:01:53 +11:00
Yasiru Rangana
ec40b17eea feat: Add token tracking support to openai_embed function 
- Add optional token_tracker parameter to openai_embed()
- Track prompt_tokens and total_tokens for embedding API calls
- Enables monitoring of embedding token usage alongside LLM calls
- Maintains backward compatibility with existing code
 2025-10-08 14:36:08 +11:00
yangdx
b9c37bd937 Fix linting 2025-10-03 02:10:02 +08:00
yangdx
112349ed5b Modernize type hints and remove Python 3.8 compatibility code 
• Use collections.abc.AsyncIterator only
• Remove sys.version_info checks
• Use union syntax for None types
• Simplify string emptiness checks
 2025-10-02 23:15:42 +08:00
yangdx
42d1d04147 Fix boolean parser problem for for LLM environment variable 
• Add custom boolean parser for argparse in BindingOptions
 2025-09-28 19:23:57 +08:00
yangdx
cff6029508 Ensure COT tags are properly closed in all stream termination scenarios 
- Add COT closure after stream completion
- Handle COT in exception scenarios
- Add final safety check in finally block
- Prevent unclosed thinking tags
- Log COT closure failures
 2025-09-22 00:09:27 +08:00
yangdx
077d9be5d7 Add Deepseek Style Chain of Thought (CoT) Support for OpenAI Compatible LLM providers 
- Add enable_cot parameter to all LLM APIs
- Implement CoT for OpenAI with <think> tags
- Log warnings for unsupported providers
- Enable CoT in query operations
- Handle streaming and non-streaming CoT
 2025-09-09 22:34:36 +08:00