LLM Graph Builder is a tool for constructing knowledge graphs in Neo4j from unstructured data using large language models.
The tool addresses the challenge of converting unstructured documents—PDFs, Word files, plain text, YouTube videos, and web pages—into structured knowledge graphs. It uses LLMs and the LangChain framework to extract nodes, relationships, and their properties from source material, then stores the resulting graph in Neo4j. Users upload files from local storage, cloud buckets like GCS or S3, or web sources, select an LLM model, and the system generates a queryable graph structure. The application includes a chat interface for conversational interaction with the stored data and provides source attribution for query results.
Adoption makes sense for teams building knowledge graphs at scale or integrating graph-based retrieval into retrieval-augmented generation pipelines. The tool supports a wide range of LLM providers including OpenAI, Gemini, Anthropic, Groq, Ollama, and others, as well as multiple embedding model sources such as Sentence Transformers and Amazon Titan. It requires Neo4j 5.23 or later with APOC installed, and works with Neo4j Aura's free tier. Custom schemas can be applied during graph generation, and token usage tracking is available to monitor LLM consumption per user and database connection. The project provides both a full application with graph visualization in Neo4j Bloom and a dedicated chat-only interface.
Development activity shows consistent engagement with feature expansion across multiple LLM integrations and embedding model options. The codebase spans Python backend with FastAPI, React frontend, and Jupyter Notebook examples, indicating active maintenance of both core functionality and documentation. The project maintains support for diverse deployment scenarios including Docker, Neo4j Desktop, and cloud databases, suggesting responsiveness to different user environments.
