fix(api): resolve model provider from config to prevent misrouting

[deboste]

When the model dropdown sends a prefixed ID like "anthropic/claude-xxx", AIAgent interprets the provider/model format as an OpenRouter path and routes through OpenRouter instead of the direct Anthropic API.

Fix: read the configured provider from config.yaml model section. If the model ID starts with the configured provider name followed by "/", strip that prefix and pass the provider explicitly to AIAgent. This ensures direct API providers (Anthropic, OpenAI, etc.) are used when configured, regardless of the model ID format from the dropdown.

[nesquena]

Review: PR #4 — model provider routing fix

Problem Statement

Valid concern — when the dropdown sends a prefixed ID like anthropic/claude-sonnet-4.6, it could be misinterpreted as an OpenRouter model path. This is a real issue worth fixing.

Issues Found

1. provider parameter may not exist on AIAgent (Critical)

Per ARCHITECTURE.md (lines 566-574), AIAgent's documented constructor parameters are: model, platform, quiet_mode, enabled_toolsets, session_id, stream_delta_callback, tool_progress_callback. There is no provider parameter. Without access to the hermes-agent source to verify, this could either silently pass (if AIAgent uses **kwargs) or raise a TypeError at runtime.

Can you confirm this parameter exists in your hermes-agent version?

2. Wrong provider for cross-provider model selection (Logic Bug)

The fix always passes the globally configured provider to AIAgent, even when the user selects a model from a different provider. Example:

• Config has provider: anthropic
• User selects openai/gpt-5.4-mini from the dropdown
• The prefix doesn't match, so the model stays as openai/gpt-5.4-mini
• But provider='anthropic' is still passed to AIAgent
• This would try to route an OpenAI model through the Anthropic API

The provider should be inferred from the model ID being selected, not from the global config.

3. Duplicated logic

The same provider-resolution block is copy-pasted in both routes.py (sync path) and streaming.py (async path). This should be a shared helper in api/config.py.

4. Variable naming style

Using underscore-prefixed names (_model, _prov, _mc, _hcfg) for local variables is unconventional — in Python, underscore prefix typically means module-private or 'unused'. Regular names like resolved_model and resolved_provider would be clearer.

Suggested Approach

A cleaner fix would be to add a helper in api/config.py:

def resolve_model_and_provider(model_id):
    """Strip provider prefix from model ID and return (model, provider)."""
    model_cfg = cfg.get('model', {})
    configured_provider = model_cfg.get('provider') if isinstance(model_cfg, dict) else None
    if configured_provider and '/' in model_id and model_id.startswith(configured_provider + '/'):
        return model_id.split('/', 1)[1], configured_provider
    return model_id, None

Then both routes.py and streaming.py can call it. But the cross-provider issue (point 2) still needs to be addressed — likely by parsing the provider from the model ID prefix itself, not from the global config.

Tests

Ran full test suite: 201 passed, 23 failed — all 23 failures are pre-existing. No regressions.

Verdict

The concept is right but the implementation has a logic bug with cross-provider model selection and a potential runtime error with the provider parameter. Requesting changes before merge.

[nesquena]

Update on point #1 from my review: I verified against the upstream hermes-agent source — AIAgent.__init__ does accept provider as a keyword argument (line 469 of run_agent.py). So that concern is resolved. Apologies for the incorrect call.

The remaining issues still stand:

• Cross-provider model selection bug (point #2): passing the global config provider even when the user selects a model from a different provider
• Duplicated logic (point #3): should be a shared helper

[nesquena]

Thanks @deboste — you identified a real and important routing issue here. I pushed a follow-up commit that:

1. Extracts the resolution logic into a shared resolve_model_provider() helper in config.py (eliminates the duplication between routes.py and streaming.py)
2. Extends the prefix matching to handle any known direct-API provider, not just the config provider — this covers edge cases like localStorage restoring a model ID from a different provider group

With those improvements in place, this is ready to merge. Thanks again for the solid bug find!