How-to guides

Task-oriented recipes. Each guide starts with a concrete goal, lists prerequisites, and walks through copy-paste steps with a verification section at the end.

For a learning-oriented introduction, see the Tutorial. For the why behind decisions, see Design & rationale.

Available now

• Migrate SQLite → Postgres — swap the backend without changing app code.
• Team deployment with Docker — Postgres + UI in one docker compose up.
• Configure MCP with Claude Code — let an agent query your runs.
• Wire up Slack alerts — webhook on metric breach.
• Track sklearn models — libpython-clj2
  • train-and-evaluate!.
• Log from Python or curl — HTTP write API end-to-end.

Planned

• Use S3 for artifacts (MinIO local, AWS S3 prod)
• Clean up old runs in CI / cron
• Build a multi-step pipeline
• Run a hyperparameter grid search
• Share runs across a team (user attribution)
• Back up the database
• Compare runs and model versions

If a recipe you need isn't here, file an issue with the use-case. The library has all the primitives; the gaps are in the writing.