Bug

When using custom_providers in config.yaml, the model name from model.name is not included in the API request body. The request sends model= (empty string), causing 400 errors from OpenAI-compatible backends.

Repro

# ~/.hermes/config.yaml
model:
  name: claude-sonnet-4-20250514
  provider: my-litellm

custom_providers:
  - name: my-litellm
    base_url: http://litellm-proxy:4000/v1
    api_key: sk-...
    models:
      - claude-sonnet-4-20250514

hermes chat -q "Hello"
# Error: HTTP 400: Invalid model name passed in model=

Expected

The model name claude-sonnet-4-20250514 should be sent in the request body.

Root cause

provider_model_ids() in models.py doesn't handle custom providers — it only has handlers for openrouter, nous, xai, copilot, etc. Custom provider names fall through to the default path which returns an empty model list, causing the model field to be empty in the API call.

Version

Hermes Agent v0.15.1 (2026.5.29)