Adapters

Adapters represent LoRA or related fine-tuning artifacts that can be attached to self-hosted runtimes.

Endpoints

POST /v1/adapters

Register an adapter.

Key request fields:

• task_id
• base_model_id
• name
• storage_path
• optional format
• optional size_bytes
• optional rank
• optional alpha
• optional target_modules
• optional metadata

GET /v1/adapters

List adapters, optionally filtered by task_id.

GET /v1/adapters/:id

Get one adapter.

PUT /v1/adapters/:id

Update adapter metadata, path, name, or active state.

DELETE /v1/adapters/:id

Soft-archive an adapter.

Notes

• Adapters are currently for vllm and llama_cpp runtimes.
• External API providers like OpenAI and Anthropic do not support user-supplied adapters through this path.