Custom / OpenAI-compatible with Empryo

Empryo has no separate "custom" key to set, you define one or more providers in config, then store each one's key by its id. There is no plugin and no code.

1. Add the provider to configLink to this section

Edit ~/.empryo/config.json (global, applies everywhere) or .empryo/config.json in a repo (project scope wins on matching id):

{
  "providers": [
    {
      "id": "together",
      "name": "Together AI",
      "baseURL": "https://api.together.xyz/v1",
      "envVar": "TOGETHER_API_KEY",
      "models": ["meta-llama/Llama-3.3-70B-Instruct-Turbo"]
    }
  ]
}

2. Store the key and select a modelLink to this section

empryo --set-key together <your-key>
empryo --model together/meta-llama/Llama-3.3-70B-Instruct-Turbo

--set-key together writes to your OS keychain (Login Keychain on macOS, libsecret on Linux, DPAPI on Windows), keyed off the provider id. You can also pass it per-launch with the env var and persist nothing: TOGETHER_API_KEY=... empryo.

3. Local server (no key)Link to this section

Drop envVar entirely for a server with no auth:

{
  "providers": [{
    "id": "local",
    "name": "Local LLM",
    "baseURL": "http://localhost:8080/v1",
    "models": ["llama-3.3-70b"]
  }]
}

(Ollama at localhost:11434 and LM Studio at localhost:1234 are already auto-detected as built-ins, only use a custom block to point at vLLM, SGLang, llama.cpp, or a non-default port.)

4. Dynamic model lists and gatewaysLink to this section

For a corporate gateway, let Empryo pull the catalog instead of hardcoding it:

{
  "providers": [{
    "id": "corp",
    "name": "Corp Gateway",
    "baseURL": "https://llm.internal.corp.com/v1",
    "envVar": "CORP_LLM_KEY",
    "modelsAPI": "https://llm.internal.corp.com/v1/models"
  }]
}

5. Reasoning modelsLink to this section

Add a reasoning block in the style your endpoint accepts. Fields are injected into every request body:

{
  "id": "dashscope",
  "baseURL": "https://dashscope.aliyuncs.com/compatible-mode/v1",
  "envVar": "DASHSCOPE_API_KEY",
  "models": ["qwen-plus"],
  "reasoning": { "enabled": true, "budget": 8192 }
}

• OpenAI-style: effort: "low" | "medium" | "high" | "xhigh" | "none"
• DashScope-style: enabled: true, budget: 8192 (Qwen/Alibaba)
• Raw: extraParams: { ... } forwarded verbatim, overriding other keys on collision

NotesLink to this section

• If your id collides with a built-in (e.g. anthropic), Empryo renames it {id}-custom rather than replacing the built-in. Custom providers always show [custom] in /keys and --list-providers.
• Verify with empryo --list-providers and empryo --list-models together.
• Inside the TUI, press Ctrl+L to switch models and /router to assign your custom model to specific task slots.