Skip to main content

Power tips

Improve reliability, security, and speed while using free NVIDIA-backed Claude Code workflows.

Workflow

  • 1Run /init at the start of any new project to create CLAUDE.md — this gives Claude permanent context about your codebase.
  • 2Use /model to switch to Opus for complex multi-file refactors, then switch back to Sonnet for speed.
  • 3Use /compact when long sessions slow down — it compresses history and frees context space.

Reliability

  • 1If a model goes DEGRADED, swap it in config.yaml and run docker restart litellm-nim. No need to restart Claude Code.
  • 2Use docker logs -f litellm-nim in a second terminal to watch which model handles each request live.
  • 3The --restart always flag means your proxy auto-starts with Docker Desktop — no manual docker start needed.

Security

  • 1Never share your nvapi-... key in screenshots, chat messages, or public repos. Regenerate immediately if exposed.
  • 2The sk-litellm-local key is just a local proxy password — it's safe to share in documentation.

Productivity

  • 1Be specific in prompts: 'fix the login bug in src/auth/login.js where empty password is accepted' beats 'fix bug'.
  • 2Use claude-nim --resume <session-id> to continue exactly where you left off after closing the terminal.

Quick Reference

Most used commands

claude-nim

Start Claude Code

/compact

Free up context

docker logs litellm-nim

Check proxy health

/model

Switch model slot