Build and explore AI-powered knowledge graphs
Create interactive 3D knowledge graphs from any data source. Detect communities, filter by signal, and ask graph‑grounded questions with a minimal, production‑ready interface.
Why GraphRAG Workbench
High signal. No noise.
A focused toolkit to ingest documents, build 3D knowledge graphs, and explore insights fast.
From PDF to Insight
Upload PDFs, run indexing, and explore a 3D knowledge graph in minutes.
Graph‑Grounded Answers
Ask questions with responses grounded on the active subgraph for precise, explainable results.
Community Intelligence
Detect clusters, surface key entities and relationships, and isolate hierarchies to focus analysis.
Precision Filtering
Slice by source, confidence, and tags to zero in on signal without recomputing layout.
Configurable Pipeline
Tune models, prompts, chunking, and workflows via settings.yaml; extend the pipeline as needed.
Simple Ops & APIs
File‑based inputs/outputs and small Next.js API surface for data, corpus, and chat.
See it in action
From document to explorable graph
Import a PDF, extract entities and relations, then navigate the graph with semantic search and filters.
Graph view
Communities, bloom, filters
Node inspector
Entity details, relations, sources
Terminal logs
Pipeline progress and tasks
Graph view
Communities, bloom, filters
Node inspector
Entity details, relations, sources
Terminal logs
Pipeline progress and tasks
On the horizon
Roadmap
Focused additions that deepen analysis, improve privacy, and sharpen insight extraction.
Deep research
Multi-hop retrieval with iterative queries and evidence synthesis.
Local LLM usage
Run extraction and chat via local models for privacy and control.
Direct API access
Programmatically interact with graphs and chat via REST endpoints.
Community pruning
Cull low-signal nodes and edges to keep graphs crisp.