Getting Started
New here? Start from what you’re trying to do — each goal points to the tool that calibrates it.
Run Nexus in five minutes
Docker is the only thing you need installed. Everything else runs inside the compose stack.
git clone https://github.com/LivingLikeKrillin/khala.gitcd khalacp nexus/.env.example nexus/.env # then set ANTHROPIC_API_KEY for narrated answerstask upNo Task? task up is three commands — run them from nexus/:
cd nexusdocker compose up -d --wait # containers + model auto-pulldocker compose exec -T nexus-app python -m scripts.migrate # ← skip this and the source console / document management breakThe first boot builds the image, which compiles mecab-ko from source — budget 10–20 minutes. Later runs start in seconds. The embedding model (nomic-embed-text, ~274 MB) is pulled automatically on first boot.
Now index something. An empty corpus answers every question with “not found”, so this step is not optional:
cd nexusdocker compose exec nexus-app nexus ingest ./docs # index the shipped design docsOpen http://localhost:8000/ and ask a question. Answers cite the chunks they came from; if the corpus doesn’t contain the answer, Nexus says so instead of guessing.
Without ANTHROPIC_API_KEY you still get the retrieved evidence — just not a narrated answer.
To stop: task down. To update after git pull: task update (rebuilds and runs DB migrations).
Pick your goal
| I want to… | → Tool |
|---|---|
| get grounded answers about my codebase/domain | Nexus / Archon |
| ground my PRs & troubleshooting in org context | Observer |
| stop rubber-stamping specs | Arbiter |
| make AI-generated tests actually verify behavior | Probe |
| know whether a human can still vouch for what the AI wrote | Adept |
5-minute tour
What is Khala? → Philosophy → Ecosystem
Prerequisites
- Common: git, a recent runtime.
- Nexus: Docker (Postgres etc. run inside the compose stack — no separate install). Observer: Node ≥20. Arbiter / Probe / Adept: Python (Probe needs
cosmic-ray).