| name | llm-provider-tool |
| description | Use when checking InternAgent LLM provider abstraction, Gemma4/Ollama transport, provider health, generation reachability, model name, or timeout diagnostics. |
LLM Provider Tool
Tool Contract
Use this skill to inspect the configured LLMProvider. The project name may be claude, but the real deployment is Gemma4 through Ollama api/generate. Never print API keys or raw provider payloads.
Script Usage
Run inside the API container:
docker compose exec api python /app/skills/llm-provider-tool/scripts/check_provider.py --diagnose
Inputs: optional --diagnose. Without it, the script only returns provider name and model.
Output Contract
The script emits compact JSON: provider, model, and optional diagnostics with reachability fields. Errors return ok=false with a safe error string.
Answer Synthesis
Separate tag/model-list reachability from generation reachability. If generation fails or times out, say the app may still be up while model generation is unhealthy.
Validation
python skills/llm-provider-tool/scripts/check_provider.py --self-test
python skills/llm-provider-tool/scripts/check_provider.py --help
docker compose exec api python /app/skills/llm-provider-tool/scripts/check_provider.py
