Skip to main content

Overview

Lore is a CLI tool that builds persistent LLM knowledge bases from any content. It ingests documents, compiles them into an interlinked markdown wiki, and provides full-text search, Q&A, and export capabilities.

What Lore Optimizes For​

  • durable knowledge artifacts over transient chat context
  • markdown-native outputs that are easy to review and version
  • practical operations loops for ingest, compile, lint, query, and export
  • agent interoperability through MCP and structured command outputs

Key Features​

  • Multi-format ingestion -- markdown, PDF, DOCX, HTML, JSON, images, URLs, videos
  • LLM compilation -- raw documents compiled into structured wiki articles with backlinks, hash-based incremental compile, and compile lock safety
  • FTS5/BM25 search -- fast full-text search with ranking and snippets
  • BFS/DFS traversal -- navigate the knowledge graph via backlinks
  • Watch automation -- debounced raw change detection with queued auto-compile coordination
  • Health diagnostics -- lint summary + line-aware diagnostics for broken links and weak pages
  • Concept metadata index -- generated .lore/wiki/concepts.json with canonical names, aliases, tags, and confidence
  • Obsidian compatible -- [[wiki-links]], YAML frontmatter, .canvas files
  • MCP server -- agent-accessible search and query
  • Multiple exports -- bundle, slides, PDF, DOCX, web, canvas, GraphML

Core Workflow​

flowchart LR
	A[Ingest sources] --> B[Compile wiki articles]
	B --> C[Index + graph refresh]
	C --> D[Search, query, explain]
	D --> E[Lint and improve]
	E --> F[Export or serve]

Who This Is For​

PersonaTypical use
Engineering teamsPreserve architecture decisions and implementation context
Product/ops teamsBuild searchable operational runbooks from mixed docs
AI-agent workflowsProvide a persistent knowledge surface for MCP tools
Individual researchersCurate and query evolving topic maps

Use Cases​

Engineering decision memory​

  • ingest RFCs, PR notes, and design docs
  • compile and query for historical decision rationale
  • use Angela to capture why behind code changes

Team onboarding acceleration​

  • compile internal docs into concept-linked pages
  • run lore explain on unfamiliar components
  • export to web/pdf for broader distribution

Agent maintenance loop​

  • run MCP tools to discover gaps/orphans/ambiguity
  • repair index, recompile, and re-lint
  • ask targeted questions over refreshed graph state

Quick Example​

lore init
lore ingest ./docs
lore compile
lore query "How does compile lock recovery work?"
lore export web