RAG・知識ベース
graphrag
Microsoft's graph-based RAG: index a corpus into a knowledge graph, then query it with an LLM
★ 36.3k スターMIT
ガイドを見る →llm-graph-builderは、neo4j-labsが公開しているオープンソースのRAG・知識ベースプロジェクトです（GitHub: neo4j-labs/llm-graph-builder）。GitHubスターは5,276、フォークは882で、Apache-2.0ライセンスの下で公開されています。GitHubでの最終更新は2026年10月6日です。
Build Neo4j knowledge graphs from unstructured data using LLMs — then chat with your graph.
Decide which unstructured sources should become a graph and what entities and relationships matter.
Create a Neo4j 5.23+ database (Aura free tier works) with APOC, and note its URI and credentials.
Run the backend and frontend (Docker Compose for most setups, or separately for Neo4j Desktop) and set the Neo4j and LLM environment variables.
Import a first source, generate the graph, view it in Neo4j Bloom, and ask questions in the chat interface.
Turns documents and other unstructured sources into nodes, relationships and properties using advanced LLMs.
Generate graphs under a custom schema you define, or reuse existing schemas configured in the settings.
Ingests local files, YouTube, Wikipedia, Amazon S3, Google Cloud Storage and web pages.
Chat with your data through the graph, with response metadata showing the sources behind answers; a standalone /chat-only route is included.
Works with multiple LLM providers (including OpenAI, Gemini, Anthropic and Diffbot options) and embedding models from OpenAI, Gemini, Amazon Titan or Sentence Transformers.
Optional per-user token usage tracking with daily/monthly consumption views and limit checks via API.
Clone LLM Graph Builder and enter the project directory.
git clone https://github.com/neo4j-labs/llm-graph-builder.git
cd llm-graph-builderSet your Neo4j connection and model keys, create the backend virtual environment and install requirements.
# Set NEO4J_URI, NEO4J_USERNAME, NEO4J_PASSWORD, NEO4J_DATABASE in backend/.env
cd backend
python3.12 -m venv venv && source venv/bin/activate
pip install -r requirements.txtRun the API server, then start the frontend dev server in a second terminal.
uvicorn score:app --reload
# frontend: cd frontend && yarn && yarn run devIt is a neo4j-labs project, maintained by Neo4j's labs team and contributors.
The app is Apache-2.0 licensed; you need a Neo4j database (Aura has a free tier) and an LLM provider account for extraction.
Extraction and chat use LLMs, so you need a key for a supported provider (or a local model setup such as Ollama, which the README covers).
Manual pipelines require custom extraction code per source; this tool provides the ingestion, LLM extraction, schema handling and chat UI as one ready application on Neo4j.
Create a Neo4j 5.23+ database, clone the repo, set the environment variables, and start the backend and frontend.
Microsoft's graph-based RAG: index a corpus into a knowledge graph, then query it with an LLM
Unstract — LLM-driven extraction of unstructured data, built for API deployments and ETL pipeline workflows.
Build a ChatGPT over your own data using natural language and LlamaIndex.
VideoRAG — chat with your videos, even hundreds of hours of them, with graph-driven retrieval.
An agentic AI data assistant - chat with your databases, run SQL and code, get reports
Private document chat for attorneys: local LLMs, verifiable citations.
LibreAIカタログに掲載されたRAG・知識ベースプロジェクトの中で最もスターの多いもの — llm-graph-builderが合わなかった場合に検討したい、現実的な候補リストです。
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or…
RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context…
Lightweight OCR toolkit turning PDFs and images into structured, LLM-ready data in 100+ languages
MinerU — turns PDFs and Office documents into LLM-ready Markdown and JSON for agents and RAG pipelines, Apache-2.0 plus terms
Stop renting your intelligence. Own it with AnythingLLM. Everything you need for a powerful local-first agent experience
カタログの中でllm-graph-builderとスター数が近いRAG・知識ベースプロジェクト — 前後の棚もそのまま見て回れます。
A low-code MCP framework for building complex, innovative RAG pipelines — visibly, not as a black box.
RAG for your monorepo: query, understand and edit multi-language codebases with a knowledge graph.
An OLTP graph database with native vector and full-text search, built in Rust on object storage.
An agent that finds anything on your computer — federated search over your files, in place.
Build a ChatGPT over your own data using natural language and LlamaIndex.
OpenKB compiles your documents into a persistent, interlinked wiki knowledge base — no vector DB