LightRAG
80 versions indexed · latest v1.5.8rc1 on Nov 15, 2026
202619
1. Configuration storage: LightRAG now remembers which embedding model built your vectors
New Feature: Add end-user-facing entry point to WebUI
Add an offline tool to migrate a graph from Apache AGE to PostgreSQL tables · feat: support graph-first ingestion with deferred vector indexing · feat: add explicit custom chunking selector
Introducing PGTableGraphStorage, a new graph storage backend built on native PostgreSQL tables. By eliminating the dependency on Apache AGE, it not only simplifies deployment but also delivers significantly higher graph…
Starting from v1.5.5, LightRAG's built-in DOCX parser introduces the Smart Heading feature -- enabling better section-based chunking even for poorly formatted DOCX documents.
feat(parser): native engine support for Markdown parsing, including embedded base64 images and .textpack bundle image integration · feat(parser): per-file engine parameters via hint/rule (Phase 2 — MinerU + Docling)
Due to PR #3228, upgrading an existing Milvus deployment triggers a one-time automatic migration on the next initialize_storages(): the legacy collection is copied into the new model-suffixed collection (schema-upgraded…
feat(parser): standardize the parser engine interface through a registration mechanism with support for third-party parsers
To prevent weaker LLMs from inadvertently incorporating few-shot examples from the prompt into the extraction results, the official entity-relation extraction prompt has been updated in the current version by removing…
🎉✨Feacture: LightRAG now supports multimodal document processing and can fully leverage images, tables, and formulas within documents to answer queries.
It's a breaking change for existing deployments on OpenSearch < 3.3.0. OpenSearch: Use version-aware sort tiebreaker for PIT search
ci: sign GHCR docker images with cosign · feat(embeddings): added embedding support for providers that don't support base64 encoding (e.g. Yandex Cloud)
feat(setup): support Atlas Local Docker for Mongo vector storage · perf: batch graph operations in ainsert_custom_kg for large-scale imports · examples: add AG2 multi-agent demo with LightRAG retrieval
Feat: add make dev bootstrap target · Feat: Add PostgreSQL performance timing instrumentation
Optimize Postgres Vector DB upsert performance by increasing the batch size to 200 records per operation. · Resolved issue with opening PDF files protected by permission-only restrictions
Integrated OpenSearch as a unified storage backend, providing comprehensive support for all four LightRAG storage types: KV, Vector, Graph, and DocStatus.
Add Makefile for quick deployment · Refactor(Makefile): split monolithic wizard into modular env-base/storage/server targets
feat: Add POSTGRES_ENABLE_VECTOR option to conditionally disable pgvector extension · Add i18n support for Vietnamese
Fix OpenAI LLM binding options not loaded from environment variables
202556
Hot Fix AttributeError in Neo4JStorage and MemgraphStorage when using storage specified workspace env var
Add Workspace Isolation for Pipeline Status and In-memory Storage: Multiple LightRAG instances with distinct workspaces can be created simultaneously, marking a significant advancement toward seamless workspace…
Feat: Add PDF Decryption Support for Password-Protected Files · Feat: Add optional Langfuse observability integration · Feat: Add RAGAS evaluation framework for RAG quality assessment
This update requires qdrant-client version 1.11.0 or later (due to the use of tenant indexing) when using Qdrant. Data migration may take a significant amount of time for large datasets.
Restore query generation example and fix README path reference · Refact: Graceful shutdown and signal handling in Gunicorn Mode · HotFix: Include swagger-docs static files in package distribution
🚀 PostgreSQL migration performance problem for large dataset in v1.4.9.4 · 🛑 Introduces a graceful pipeline cancellation mechanism for document processing operations
In production deployments, entity and relation metadata can grow unbounded as documents are continuously ingested. The source_id (chunk IDs) and file_path fields in entities and relations can accumulate thousands of…
Add temporary solution implemented to ensure compatibility with the newly introduced document status in Rayanything.
feat: Add endpoint and UI to retry failed documents · Refactor(webui): Improve document tooltip display with track ID and better formatting · feat: add options for Postgres connection
Feature(webui): Add KaTeX chemical formula rendering by supporting mhchem extension · Feat(webui): Prevent LaTeX Parsing Errors Show-up During Streaming · Fix dark mode graph labels for system theme and improve colors
v1.4.9 introduces key enhancements focused on refining the reference output format and incorporating structured references into query results.
Refactor: Add error handling with chunk ID prefixing in entity extraction · fix: resolve dark mode text visibility issue in knowledge graph view · Fix: Resolve DocumentManager UI freezing and enhance error handling
Introduced a raw data query API /query/data, enabling developers to retrieve complete raw data recalled by LightRAG for fine-grained processing.
Prompt Optimization: remove angle brackets from entity and relationship output formats · Refactor PostgreSQL Graph Query by Native SQL and Standardized Parameter Passing
Optimize Entity Extraction for Small Parameter LLMs with Enhanced Prompt Caching · Add LLM COT Rendering support for WebUI · Enhance document status display with metadata tooltips
Refactoring Prompt Template and enhancing robust mal-format handling for Knowledge Graph (KG) extraction using Small Parameter Language Models (LLMs). This will invalidate all cached LLM outputs.
Doc-id based chunk filtering feature is remove from PostgreSQL vector storage. · Prompt template has been updated, invalidating all LLM caches
feat(performance): Optimize Document Deletion Performance with Entity/Relation Indexing · refactor: improve JSON parsing reliability with json-repair library
Feat(webui): add document list pagination for webui · Feat: add Document Processing Track ID Support for Frontend · Better prompt for entity description extraction to avoid hallucinations
Refac: optimizing context builder to improve query performance · Improve query for find_text_unit · feat: Integrate Jina embeddings API support · feat: Add SSL support for PostgreSQL database connections
Feat: change default query mode from hybrid to mix · Add Azure deployment · Remove separate requirements.txt and update Dockerfile to use pip install
Fix Milvus DataNotMatchException · fix: change default edge weight from 0.0 to 1.0 in entity extraction and graph storage
Optimize PostgreSQL database migrations speed for LLM cache · Fix LLM cache key translation error for PostgreSQL · Fix PostgreSQL AGE graph storage workspace error after migration
After upgrading, please clear documents and cache, then re-index your documents for document deletion feature functions correctly. Remember to select the 'clear cache simultaneously' option when clearing documents.
docs: Add clear initialization requirements and troubleshooting section · Add doc for explaining LightRAG's multi-document concurrent processing mechanism
Fix: Correct GraphML to JSON mapping in xml_to_json function · Fix js files MIME type error for Windows environment · Reflect core library parameter update in API server
Feat: Add user prompt to QueryParam · Deduplicate merged relation keywords · Allow max_connections to be configured in postgres · Persistent LLM cache on error
Hotfix: Resolve send button not working for some old version browser for crypto.randomUUID · Improve parallel processing logic for document extraction and merging
Resolved missing dependency issues during PyPI package installation via pip
Feat: Add Markdown & Mermaid support for WebUI · Feat: add custom tokenizer · Feat: Add query mode prefix support in retrieval test of WebUI · Improve PostgreSQL AGE query speed by write a customized BFS implementation
Resolved special character handling issue in PostgreSQL graph storage · Properly handling of backslashes and quotation marks in entity and relation names
Key improvements: Enhanced document indexing speed. Significantly improved the query performance of Neo4j and PostgreSQL AGE graph storage.
Refactoring PostgreSQL AGE graph storage implementation, which introducing breaking changes that are incompatible with existing graph data
refactor(SiteHeader): Hide logout button in guest mode · Refactor auth and version checks for improved reliability · Refactor LightRAG Server documentation for clarity and structure. · Fix mongo doc status
Optimize parallel file processing in folder scan · Fix dead lock problem in Gunicorn multi workers mode · Fix empty graph not display correctly after database dropped
Added camera rotation to graph panel and merge login feature · Added loginPage · Fixed get_by_id type error in PSQL impl · Hot fix: Server can not startup with legacy data (json_doc_status_impl.py)
Resolve the issue with making API calls to Azure OpenAI service · force keywords_extraction output as JSON · Improve Entity Extraction Robustness for Truncated LLM Responses · Filter by ID during Query for Postgres VDB
Add summary language setting by env · Completed the missing methods in various databases to support modifications for the Graph. · Made .env file loaded from the current folder instead of the script folder
Add multi workers support for API Server · adding full_doc_id to insert · Improved caching check · Add max nodes limit for graph retrieval
Use consistent format as in examples · feat: fix delete by document id
fix AttributeError: 'NoneType' object has no attribute 'dim' · Add retrieval message code highlighting and copy button · Enhance Graph Visualization with Configurable Depth and Layout Iterations
Added more file types support to the webui · Simplify dockerfile and docker-compose files · add support of providing ids for documents insert · fix insert_custom_chunks skipping every new doc
Revert 'removed get_knowledge_graph' · Updated version to 1.1.11
Template github · updated template · Updated templates · Removed all useless try except around imports as pipmaster already handles that
Fix ImportError by Removing Redundant Try-Except Block
Cleaned typing · Lightrag Server Launch fix. · fixed missing install · Refactor Database Connection Management and Improve Storage Lifecycle Handling
chroma_impl supports local persistent client: PersistentClient · Add MongoDB VectorDB Support and Optimize Existing MongoDB Implementations · Enhance Neo4j graph storage with error handling and label validation
Improve LLM Error Handling for API Server · Split ollama api to separated file · fix graph viewer crash on macos when show file dialog · Fixing .env file loading problem for Ollama API module
Fix broken main · feat: Added webui management, including file upload, text upload, Q&… · Add history support for ollama api · fix README
Fixed API docs link in Readme.md · Addind a simple tool to view the lightrag database in 3d · Replacing ParisNeo with this repo owner, HKUDS
Major Updates: Docker Support, Enhanced Configuration, and Documentation Restructuring #554 · add readme_zh · Add Ollama compatible API server · Fixes #596 - Hardcoded model deployment name in azure_openai_complete
Fix for mongoDB document 16MB limit and add os env NEO4J_MAX_CONNECTION_POOL_SIZE for neo4j · Enhancement: Multi-format document support and dynamic dependency management #565
Add character segmentation functionality for chunks · Fix:Optimized logic for automatic switching modes when keywords do not exist · Implement custom chunking feature
Add Optional API Key Authentication to FastAPI Services · Added PostgreSQL implementation · Add support for selecting a specific database in Neo4j
Streamlined API Installation and Documentation Enhancement · Fixed keywords_extraction_examples format error · Fixed the async issue while running on Windows
20245
Add LoLLMs Integration to LightRAG · GremlinStorage: support more Gremlin compatible DBs · Added Azure OpenAI api sample with streaming · API Installation Options and Documentation Improvements
Add Apache AGE graph storage · refactor(lightrag): Optimize the initialization logic of · fix(llm): Fix the initialization issue of · support TiDBGraphStorage
chore: update llm.py · fix(lightrag): Fix the logic for handling chunks with entities but no relationships · Fix/lazy import · fix: rare embedding issue.
Added the ability to use embedding cache during queries · Fixed a bug where args_hash was not calculated during embedding caching due to only being computed in regular caching scenarios · Update oracle_impl.py
Add MongoDB as KV storage · Fix JSON parsing error