| name | cadis-model-provider |
| description | Use when implementing or reviewing CADIS model provider traits, provider capabilities, streaming responses, tool-call support, OpenAI, Ollama, Anthropic, Gemini, OpenRouter, LM Studio, or custom HTTP providers. |
CADIS Model Provider
Read First
docs/01_PRD.md
docs/04_TRD.md
docs/06_IMPLEMENTATION_PLAN.md
Rules
- Keep model providers behind a shared trait.
- Expose capabilities.
- Support streaming and cancellation.
- Map provider errors into CADIS errors.
- Do not leak API keys into events or logs.
- Do not hardcode one provider into core orchestration.
Capabilities
tool_calling
vision
json_schema
reasoning_effort
prompt_cache
max_context_tokens
local_model
streaming
Provider Order
Prove one cloud provider and one local provider early. Good defaults are OpenAI and Ollama.
For OpenAI-specific work, also use the openai-docs system skill.