Configure AIFast in Hermes Agent

Hermes Agent currently supports custom OpenAI-compatible endpoints. Recommended to pass hermes model Interactive wizard configuration; from the old tutorial LLM_MODEL It is no longer the source of the current configuration and should not be copied.

Check before you start

Prepare the following information:

Base URL: https://www.aifast.club/v1
API Key: you are here AIQuick station created independently Key
Model: /v1/models The actual model returned ID

First use curl to verify the Key and model list:

export AIFAST_API_KEY="yours_API_Key"

curl https://www.aifast.club/v1/models \
  -H "Authorization: Bearer $AIFAST_API_KEY"

Configuration steps

Run:

hermes model

Select the custom endpoint in the wizard and fill in the following:

1. API base URL：https://www.aifast.club/v1
2. API key: AIFast independent Key
3. Model name: real model ID
4. API mode: Select the corresponding mode when the target model uses Chat Completions

Hermes will save the current model and endpoints to ~/.hermes/config.yaml. After the configuration is completed, check whether the core fields in the file are consistent:

model:
  default: YOUR_MODEL_ID
  provider: custom
  base_url: https://www.aifast.club/v1

Do not submit configuration files containing real keys to the Git repository. If using a named custom provider, priority is given to key_env Reference environment variables.

Why not use it first? OPENAI_BASE_URL

Hermes will still be openai-api provider reads OPENAI_BASE_URL, but custom endpoints are recommended hermes model or model.base_url. This prevents providers, old environment variables, and current model configuration from overwriting each other.

Minimal verification

After starting Hermes, send the short task first:

Reply only：AIQuick station connection successful

Then check:

1. The actual provider used is custom.
2. Model ID vs. /v1/models consistent.
3. The request address is not duplicated /v1。
4. After the normal conversation is successful, test tool calls and long tasks.

FAQ

Still requesting old provider after configuration

Check ~/.hermes/config.yaml in model.provider and model.base_url, and clean up old environment variables that conflict with the current settings. Restart Hermes and verify again.

/model custom Unable to automatically select model

Automatically select dependent endpoints /models Return. If multiple models are returned, Hermes cannot tell which one you want to use, and the real model ID should be filled in explicitly.

Text generation succeeds but Agent tool fails

Custom endpoints must correctly handle the request and response fields required for tool calls. Confirm the link with a normal conversation first, and then test the tool call individually; do not attribute all failures to model capabilities.

Next step

Set API Key, usage limit and log retention policy individually for Hermes. Test tool calls, long contexts, 429 backoffs, and model switches before going live, and make sure profile permissions are not open to other local users.