[Setup]: Help to call model from 9router!

[metallicamu]

What's Going Wrong?

I got error when using model from 9router:"Could not reach the custom:9router API to validate cx/gpt-5.4. If the service isn't down, this model may not be valid."
And it returns API call failed with

I tried to config, via hermes wizard and manually in config.yaml but not works.

custom_providers:

• name: 9router
  base_url: http://localhost:20128/v1
  model: cx/gpt-5.4
  I did curl http://localhost:20128/v1/models and it returns full models.

Steps Taken

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Installation Method

Install script (curl | bash)

Operating System

Ubuntu 24.04

Python Version

No response

Hermes Version

v0.10.0 (2026.4.16

Debug Report

Full Error Output

API call failed (attempt 1/3): BadRequestError [HTTP 400]
   🔌 Provider: custom  Model: cx/gpt-5.4
   🌐 Endpoint: http://localhost:20128/v1
   📝 Error: HTTP 400: Error code: 400
⚠️ Non-retryable error (HTTP 400) — trying fallback...
❌ Non-retryable error (HTTP 400): HTTP 400: Error code: 400
❌ Non-retryable client error (HTTP 400). Aborting.
   🔌 Provider: custom  Model: cx/gpt-5.4
   🌐 Endpoint: http://localhost:20128/v1
   💡 This type of error won't be fixed by retrying.

What I've Already Tried

No response

[aniruddhaadak80]

To use OpenRouter (or any OpenAI-compatible provider) with Hermes, you should configure it in your config.yaml file. Here is a typical configuration:

providers:
  openrouter:
    base_url: "https://openrouter.ai/api/v1"
    api_key: "${OPENROUTER_API_KEY}"

Then, ensure your OPENROUTER_API_KEY is set in your .env file or environment. You can select this provider using hermes setup or by setting model: openrouter/openai/gpt-4o (or similar) in your config. If you are experiencing 'model not found' errors, double-check that the base_url includes the /v1 suffix if required by the provider.

[myugan]

@aniruddhaadak80 9router is different from OpenRouter

[manoocher]

I face the same problem. Anyone can help!

[ndriodna]

when i use 9router on pi agent, 9router need u send user agent header, but i can't set in config.yaml custom header user agent, anyone can help me

[Sasni]

Root cause found

The bug is in get_compatible_custom_providers() in hermes_cli/config.py (lines ~3983-3991):

custom_providers = config.get("custom_providers")
if custom_providers is not None:
    if not isinstance(custom_providers, list):
        return []  # ← BUG: early return, never reaches providers processing below
    for entry in custom_providers:
        _append_if_new(_normalize_custom_provider_entry(entry))

# This code is UNREACHABLE when custom_providers exists but is not a list:
for entry in providers_dict_to_custom_providers(config.get("providers")):
    _append_if_new(entry)

What happens: When custom_providers is set to a JSON string (e.g. '[{"name":"9router",...}]' — which is how hermes config set serializes it, and how many users manually paste it), the function sees it's not None but also not a list, so it returns [] immediately. The providers section processing below is never reached. This means any custom provider defined in the providers: dict is invisible.

Workaround for users:

1. Remove the custom_providers key from config.yaml entirely
2. Add your provider under the providers: section instead:

   providers:
     9router:
       base_url: http://127.0.0.1:20228/v1
       api_key: your-key
       api_mode: openai

3. Set model.provider: custom:9router

Proper fix in code: The function should not bail out with return [] when custom_providers is malformed — it should log a warning and continue to the providers dict processing below. Or alternatively, the config validator should catch string-typed custom_providers early and show a clear error message instead of the cryptic "Unknown provider" at runtime.

Reproduced and worked around on Hermes v0.16.0 (Windows).