Ingest documents into a wiki and typed knowledge graph with provenance.
npx skills add https://github.com/fizznix/mini-context-graph --skill mini-context-graph
Please help me install this Agent Skill. Skill: mini-context-graph Source: https://github.com/fizznix/mini-context-graph/tree/main Command: npx skills add https://github.com/fizznix/mini-context-graph --skill mini-context-graph
This Skill helps agents stop re-deriving knowledge from scratch by turning ingested documents into a persistent wiki plus a typed knowledge graph with evidence, so answers compound over time.
Ask an agent to ingest a new document by following ingestion.md, extract entities/relations, call ingest_with_content with supporting_text for each item, then query the knowledge graph with query_with_evidence to get an evidence-backed answer.
High-intent search queries and answers about installing and using this skill.
A knowledge graph with provenance links extracted entities and confidence-weighted relations back to source documents. It enables agents to traverse depth-capped BFS subgraphs and retrieve multi-hop answers with direct evidence chunk links.
You ingest unstructured text by following YAML-guided extraction rules to extract entities and relations, then call the ingestion process with supporting text for each item to store normalized nodes and edges in JSON.
Use a wiki-first retrieval system that builds persistent summaries and entity pages. This compounding wiki allows future queries to hit a fast path instead of re-processing unstructured text from scratch.
Yes, YAML-guided extraction rules and ontology-based normalization are required. They define how entities and relations are typed, normalized, and stored with confidence thresholds in the JSON knowledge graph.
Querying with evidence uses depth-capped BFS traversal to assemble relevant subgraphs from the typed knowledge graph. It returns answers alongside supporting documents and source text chunks for auditability.
Yes, the typed knowledge graph stores nodes and edges directly in JSON files with confidence thresholds and evidence chunk links. This persistent JSON structure enables compounding wiki-first retrieval without external databases.
vignesh2027Turn unstructured documents into queryable knowledge graphs with entity extraction and relationship mapping.
jamie8johnsonSeed semantic search results and expand them via call graph traversal.
BartunekMaintains a persistent, cross-referenced wiki from URLs, images, and text in a local workspace.
roy2392Retrieve long-term memory via vector search and recursive graph traversal on SQLite.
