trustgraph

Why we included this project

TrustGraph is for teams that need agents to behave predictably in production. Instead of feeding an LLM flat text chunks, it organizes a domain into structured context graphs where entities, relationships, and evidence are first-class objects, and grounds every agent query against that graph. Answers carry traceable provenance and reasoning paths you can inspect, which tackles the common complaint that RAG-based agents are hard to trust or debug. The project bundles the full stack underneath your agents, from knowledge ingestion and structured storage to graph-grounded retrieval, orchestration, and an LLM inferencing layer, and runs without third-party services beyond optional cloud model APIs, so you can self-host models via vLLM or similar. Context is packaged into versioned, portable units that can be shared across teams and deployed in Docker or Kubernetes, so organizations get reusable, explainable knowledge infrastructure instead of one-off prompt pipelines.