Skip to content

CLAUDE.md

Latest commit

 

History

History
28 lines (19 loc) · 1.33 KB

CLAUDE.md

File metadata and controls

28 lines (19 loc) · 1.33 KB
Raw

Model Names

Always use full model IDs (e.g. "gemini-2.5-flash", "claude-haiku-4-5-20251001") in tutorial code. Never use short aliases ("flash", "haiku", "sonnet", etc.) — except in tutorial 06 which demonstrates ResolveModelAlias.

Environment & API Keys

  • A local .env file in the project root contains all API keys (Anthropic, Gemini, OpenAI, xAI, Mistral).
  • Always source .env before running any tutorial or test that hits a cloud provider.
  • Ollama runs locally and needs no API key.

Testing

  • Start small: test one tutorial with one local model (ollama) + one cloud model first, then expand.
  • Short tests: use go test -short ./... — the full capability/IQ test (20_capability_test) is very long-running.
  • Each tutorial is its own Go module with its own go.mod — update deps per-module.
  • 19_stress_tests uses mocks and is safe to run without API keys or -short.
  • 20_capability_test is the full IQ test — never run it casually; it's expensive and slow.

Git Workflow

  • Always git pull first before starting any work — ensure you have the latest remote changes.

Releases

  • Always update CHANGELOG.md before committing a version bump or tagged release.
  • Follow the existing format: ## vX.Y.Z — YYYY-MM-DD with ### Added, ### Changed, ### Fixed sections.