# OpenClaw Workspace Boot Instructions

## Mission
Run a reliable local-first OpenClaw setup with Ollama for Telegram bot usage.

## Startup Checklist
1. Confirm Ollama API is reachable at `http://127.0.0.1:11434`.
2. Confirm local model availability with `ollama list`.
3. Prefer tool-capable local models in this order:
   - `ollama/devstral:24b`
   - `ollama/qwen3:14b`
   - `ollama/gpt-oss:20b`
4. Never auto-select `ollama/deepseek-coder-v2:16b` for Telegram routing (tool support issues).
5. If a model fails due to tool capability, switch to the next model in priority order.

## Telegram Behavior
- Keep replies practical and concise.
- If a command fails, return:
  - the exact failing command,
  - the actionable fix,
  - the next command to run.
- For setup errors, prioritize unblocking local runtime first before suggesting paid cloud providers.

## Safety and Secrets
- Never print full tokens, API keys, or secrets in chat output.
- If credentials are required, ask the user to set them via environment variables or onboarding prompts.

## Diagnostic Commands
Use these first when troubleshooting:

```bash
openclaw --version
openclaw models list
openclaw hooks list --eligible
openclaw logs --follow
ollama list
curl -s http://127.0.0.1:11434/api/tags
```