trustgraph-ai/trustgraph

What it solves

TrustGraph addresses the "common context understanding" problem in enterprise AI, where LLMs often hallucinate or fail when relying on fuzzy vector embeddings and semantic search. It replaces statistical probability with explicit, structured semantics to ensure that AI agents can communicate and reason through complex enterprise workflows and governance policies with deterministic outcomes.

How it works

TrustGraph implements a context orchestration layer using hypergraphs (based on RDF 1.2 and OWL standards). Unlike standard knowledge graphs that only support binary relationships, TrustGraph's hypergraph architecture allows for n-ary relationships, enabling it to group complex, multi-entity events into single addressable units.

The system converts raw enterprise data (PDFs, wikis, APIs) into AI-ready context by extracting entities and relationships via LLMs. It then manages this context through Workspaces (for data isolation), Collections (for partitioning knowledge bases), and Context Cores (modular units of knowledge). Agents interact with this data using SPARQL or GraphRAG to traverse explicit paths, eliminating the need for probabilistic guessing.

Who it’s for

It is designed for enterprise developers and organizations building autonomous AI agents that require high precision, strict governance, cryptographically verifiable traceability (provenance), and the ability to run open-source models on their own hardware (air-gapped or VPC).

Highlights

• Hyper-relational Context: Uses RDF 1.2 and Named Graphs to capture complex enterprise events beyond simple binary links.
• Hyperflows: Chained agentic workflows where LLM selection and graph access permissions are configured per step.
• Provenance: Real-time traceability of every decision an agent makes, solving the "black box" problem.
• Open LLM Stack: Built-in inference stack supporting open-source models on Nvidia, AMD, or Intel hardware.
• BYOO (Bring Your Own Ontology): Allows loading custom OWL ontologies to ensure semantic compliance for all ingested data.
• Integrated Tooling: Includes a 3D context explorer, ontology workbench, and GraphRAG visualizer.