hermes - 💡(How to fix) Fix bug, configuration [1 pull requests]

Environment:
- OS: Android (Termux)
- Python: 3.13.13
- Hermes version: latest (install via curl script)
- Provider: nvidia
- Model ID: qwen/qwen3.5-122b-a10b (with vendor prefix)

Issue Description:

hermes doctor incorrectly reports a configuration error when using NVIDIA provider with a vendor-prefixed model ID:


⚠ model.default 'qwen/qwen3.5-122b-a10b' uses a vendor/model slug but provider is 'nvidia' (vendor-prefixed slugs belong to aggregators like openrouter)


Reproduction Steps:

1. Set provider to nvidia in config.yaml:
   bash
   hermes config set model.provider nvidia


2. Set model to a NVIDIA-supported model with vendor prefix:
   bash
   hermes config set model.default qwen/qwen3.5-122b-a10b


3. Run:
   bash
   hermes doctor


4. Observe false positive warning about vendor prefix being incompatible with NVIDIA provider.

Actual Behavior:
- hermes doctor warns: "vendor-prefixed slugs belong to aggregators like openrouter"
- hermes doctor --fix cannot resolve the issue
- Removing the prefix (qwen3.5-122b) causes model lookup to fail (NVIDIA NIM requires the full name with prefix)

Expected Behavior:
- hermes doctor should recognize that NVIDIA NIM API supports vendor-prefixed model IDs
- No warning should be raised, OR a more accurate warning should be shown if there's actual incompatibility

Verification:
- Current configuration works correctly (agent can make API calls successfully)
- Removing vendor prefix breaks model identification
- NVIDIA NIM officially documents support for models like qwen/qwen-2.5-72b-instruct

Impact:
- False positive confuses users into thinking their config is broken
- No automatic fix available via --fix
- Forces users to choose between a broken config or a non-functional model

Probable Root Cause:

The doctor check logic in hermes_cli/commands.py (or related config validation code) likely assumes that only aggregators like OpenRouter support vendor-prefixed model slugs. However, NVIDIA NIM also uses this format for their model catalog.

Proposed Fix:

Update the configuration validation logic to:
1. Recognize that nvidia provider supports vendor-prefixed model IDs, OR
2. Remove the blanket warning and instead validate against a list of known working model names per provider, OR
3. Show a more nuanced warning that doesn't imply incompatibility when the model actually works

Suggested Code Location:
- hermes_cli/commands.py - doctor command implementation
- agent/prompt_builder.py - model name validation logic
- hermes_cli/config.py - configuration validation rules

Additional Context:
- This is a Termux/Android environment, but the issue likely affects all platforms using NVIDIA provider
- User confirmed via experimentation that removing prefix breaks functionality
- Current session runs successfully despite the warning, proving the config is valid



Labels: bug, doctor, configuration, nvidia, false-positive