setup

Setup

Build, maintain, and monitor the code-index knowledge graph.

Index a Repository

code-index analyze              # current directory
code-index analyze /path/to/repo
code-index analyze --force      # rebuild from scratch
code-index analyze --no-embeddings  # skip semantic vectors (faster)

Check Health

Use the health_check tool to verify index quality:

• Coverage -- fraction of symbols with at least one edge (> 30% = healthy)
• Staleness -- hours since last index (< 168h = healthy)

Other Commands

code-index status     # check index status
code-index list       # list indexed repos
code-index clean      # remove index data
code-index search "query"  # CLI search
code-index mcp        # start MCP server

Pipeline Steps

1. File discovery -- walk repo, respect .gitignore
2. AST parsing -- tree-sitter extraction of symbols
3. Import resolution -- cross-file symbol table
4. Call tracing -- function/method call edges
5. Heritage -- inheritance and interface relationships
6. Community detection -- Louvain clustering of related code
7. Flow detection -- DFS-traced execution flows
8. Persistence -- SQLite + FAISS storage
9. Embeddings -- vector generation for semantic search

Storage

• .code-index/ in the repo root
• ~/.code-index/registry.json tracks all repos
• Safe to delete and re-index