extract-knowledge-graph

Knowledge graph extraction from markdown documentation detects concepts, APIs, and entities, then emits navigable relationship edges in both JSON and Mermaid formats. It canonicalizes aliases into unified nodes with stable ULIDs to avoid dangling edges.

The best way to convert scraped docs into a Mermaid graph is by extracting semantic entities and relationship edges from the captured markdown corpus. The pipeline outputs a structured .mmd file alongside a knowledge.json file to build navigable graph facets.

Entity resolution and alias collapsing for a documentation corpus work by detecting equivalent entity mentions across pages and mapping them to canonical nodes. The pipeline reuses existing indexed IDs when available or allocates new ULIDs for stable identity.

Knowledge graph extraction works with Rust source repositories by using optional repo-backed enrichment. It leverages rustdoc JSON to enrich API extraction when a cloned source repository is available alongside the markdown documentation corpus.

When generating an API knowledge graph, the pipeline extracts relationships such as extracted-from, defines, exposes, and uses or part-of. These relation edges are validated against canonical nodes to prevent dangling connections in the graph output.

You should use a knowledge graph pipeline instead of a simple link graph when you need structured semantic extraction beyond page links. It transforms a topic's scraped markdown corpus into canonical nodes and navigable graph facets suitable for large corpora.