hermes - 💡(How to fix) Fix Configure inference provider for agent profiles [1 participants]

TL;DR

The Hermes agent requires a --provider flag or environment-level configuration to select the inference backend, and updating the bridge.py to pass the provider flag or set the config before invocation is likely necessary.

Guidance

• Update the bridge.py to pass the --provider flag when invoking the inference backend, ensuring the correct provider is selected.
• Set the provider configuration at the environment level or in agent profiles to avoid manual specification each time.
• Document the supported providers, including openai-codex and Anthropic, and outline the authentication requirements for each, such as the need for ANTHROPIC_API_KEY and claude CLI auth for Anthropic.
• Test the updated configuration with different providers to ensure compatibility and correct functionality.

Example

# Example of passing the provider flag in bridge.py
import subprocess

# ...

provider = 'openai-codex'  # or 'anthropic'
subprocess.run(['hermes', '--provider', provider, ...])

Notes

The solution assumes that the --provider flag is supported by the Hermes agent and that the environment-level configuration is properly set up. Additional testing may be required to ensure compatibility with different providers and authentication methods.

Recommendation

Apply workaround: Update the bridge.py to pass the provider flag or set the config before invocation, as this directly addresses the identified issue and allows for flexibility in selecting the inference backend.