Core Capabilities
graphchat unifies static analysis, semantic extraction and graph clustering into a single skill that any AI coding assistant can invoke. Click Read more on any card to see the underlying algorithms and design decisions.
Multi-Modal Extraction
Parses code (.py, .js, .go, .java, …), Markdown, PDFs and images. Tree-sitter (via WASM) runs locally in the gph CLI to extract ASTs and call graphs; LLMs extract concepts from prose; vision models read diagrams.
Local-First AST Extraction
Source code never leaves your machine. The gph CLI bundles WASM grammars for ten languages and runs Tree-sitter directly on disk, sending only structured graph metadata to the API.
Knowledge Graph Build
Merges all extracted nodes and edges into a NetworkX graph and applies the Leiden algorithm for semantic community detection — no vector embeddings required.
God Nodes & Surprises
Identifies the highest-degree "god nodes" at the heart of the system and flags unexpected cross-file or cross-domain connections worth investigating.
Interactive Outputs
Exports an interactive graph.html, a queryable graph.json, and a human-readable GRAPH_REPORT.md audit report.
gph CLI — 12-Command Graph Tool
A purpose-built CLI for humans and AI agents: semantic search with token budgets, shortest-path traversal, AI explain, live watch mode, incremental index, and structured JSON export.
API Key + JWT Auth for CLI
Scoped API keys (sk-graphchat-…) are minted in the dashboard and exchanged for short-lived JWT access + refresh tokens. The CLI auto-renews tokens invisibly — no repeated logins.
Live Watch & Incremental Index
gph watch keeps the graph in sync automatically: file-system watcher mode for continuous polling, or git hook mode that re-indexes only on commit — zero manual sync steps.
Secure by Design
Strict input validation: only http/https URLs, size and timeout limits, path containment, HTML-escaped node labels — defending against SSRF, injection and XSS.
Edge Confidence Tiers
Every edge carries a confidence tier — extracted, inferred, or speculative — so agents know how much to trust a relationship before reasoning over it.
Community Cache (Redis)
Graph communities and their serialised context bundles are cached in Redis, keyed by graph version — ensuring fast repeated queries with zero stale-data risk.