Quick Start

Quick Start

Three commands take you from zero to natural-language, LLM-ready context.

columbus install                 # onboard: write .columbus.json, create db, first index + embed
columbus search "parse config"   # natural-language, ranked context with exact line ranges
columbus view                    # full-screen dashboard (index, memory, epics, tasks, graph)

What each step does

1. install

Installed columbus project proj_2fd171eebe922fad
  config:   /your/project/.columbus.json
  data dir: ~/Library/Application Support/columbus
  indexed:  4 files, 7 symbols, 7 embedded

Mints a stable project_id, writes .columbus.json (added to .git/info/exclude so it is never committed), creates the database in your OS data dir keyed by project_id — not in your repo — and runs the first index and embedding pass. See Configuration.

Embeddings need an onnxruntime shared library. Release archives bundle it; a go install build falls back to keyword search until one is present. Run columbus doctor to check. See Installation.

2. search

1. parseConfig  [function]  (score 0.74, semantic match)
   src/config.ts:4
   function parseConfig(env: Record<string, string>): Config {
2. Config  [type]  (score 0.68, semantic match)
   src/config.ts:1
   type Config = { port: number; host: string };
...

The query is embedded on-device and matched by vector similarity, then re-ranked by deterministic heuristics — so it finds things by meaning, not just exact words. Every hit carries a score, a why, and exact locations. Add --kind, --limit, --json, or --llm to shape it (see Searching Effectively).

3. view

A full-screen, read-mostly dashboard over the indexed project — index freshness, durable memory, epics & tasks, and dependency-graph hubs. It auto-refreshes. See Command: view.

Keeping it current

columbus reindex re-chunks and re-embeds only what changed. See Keeping the Index Fresh.

Then: point your agent at it

The whole point of Columbus is to be a tool your coding agent calls. The agent learns the commands through skills and consumes --json / --llm output. See Using Columbus with Your Agent.

Next steps

→ Your First Index & Search — a fuller guided tour with real output and outcomes.