Explainability at a Glance
See how TrustGraph traces answers back to their sources using the Workbench
Beginner
3 min
- TrustGraph deployed (Quick Start)
- Understanding of Core Concepts)
- A document already loaded and processed with Graph RAG (see Graph RAG guide)
Run an explainable GraphRAG query, review the reasoning trace, and trace selected edges back to source documents.
Explainability in TrustGraph is always on. Every query records a full reasoning trace in the context graph automatically. It’s down to the application to decide what to show the user and how to interpret the explainability data. The Workbench provides a built-in view, but any application built on TrustGraph can query the same data.
Step 1: Run a Query
In the Workbench, navigate to the Assistant page. Make sure Graph RAG is selected as the retrieval strategy.
Enter a question and submit the query. The answer will stream back, and TrustGraph records the full reasoning trace.
Step 2: View the Reasoning Trace
Once the answer appears, expand the How I found this answer section to see the explainability trace inline.
The trace shows key stages of the reasoning pipeline:
- Query — the original question
- Exploration — how many edges were traversed in the knowledge graph
- Selected Evidence — the edges chosen as most relevant, shown as subject → predicate → object triples, each with the LLM’s reasoning for why it was selected
All explainability traces are persistent. The Explainability using CLI guide shows how to list and review past explainability sessions using command-line tools.
Next Steps
- Explainability using CLI — the same workflow using command-line tools, with more control over output formats
- Explainability overview — deeper understanding of the architecture