Allow self-defined OpenAI API compatible endpoints in addition to the hardcoded AI providers

[markgifford]

Is your feature request related to a problem? Please describe.
I don't directly use any of the listed AI providers: OpenAI, Anthropic, DeepSeek, Gemini. I happen to use OpenRouter, which like many meta-providers, does provide an OpenAI API compatible endpoint. Through OpenRouter, I can access all the LLMs I choose. In the past I've also used self-hosted apps like LiteLLM to provide the same function - basically act as an API gateway, one place to manage access to the actual LLMs. I would like to be able to use those same LLMs in this way with the AI Assistant in Pulse.

Describe the solution you'd like
Another option under the AI provider configuration to allow self-defined OpenAI API compatible endpoints. User adds the endpoint (in my case it'd be "https://openrouter.ai/api/v1") and the API key, and from there it should function the same as if the user had selected the OpenAI provider directly.

Describe alternatives you've considered
I tried adding the OpenRouter endpoint to the Ollama field but couldn't get it to work. Not sure it's likely to anyway.

Additional context
It seems like allowing meta-providers like this would reduce the need for the team to keep adding providers to the app

[pulse-triage]

Triage Summary

Valid enhancement request for custom OpenAI-compatible API endpoints.

Use case: Users with meta-providers like OpenRouter, LiteLLM, or self-hosted proxies that provide an OpenAI-compatible API.

Current state: Pulse has hardcoded providers (OpenAI, Anthropic, DeepSeek, Gemini, Ollama). The Ollama option is the closest to a generic endpoint but uses a different API format.

Implementation approach:

• Add a "Custom OpenAI" provider option
• Allow user to specify base URL (e.g., https://openrouter.ai/api/v1)
• Use standard OpenAI SDK with custom base URL
• Already have precedent for custom base URLs (see NewAnthropicClientWithBaseURL)

This would benefit users of:

• OpenRouter
• LiteLLM
• vLLM
• LocalAI
• Any OpenAI-compatible gateway

---

*Triaged as enhancement. Good suggestion about reducing the need to add providers individually

[pulse-triage]

This already works with the current UI:

1. Go to Settings > AI
2. Select OpenAI as the provider
3. In the "OpenAI Base URL" field, enter your custom endpoint (e.g., https://openrouter.ai/api/v1)
4. Enter your OpenRouter (or other provider) API key in the API key field
5. Save

The OpenAI client uses the custom base URL you provide, so any OpenAI-compatible endpoint works. The UI says "OpenAI" but you're actually hitting your custom endpoint.

This works for OpenRouter, LiteLLM, vLLM, LocalAI, or any OpenAI-compatible API.

---

Closed by automated triage. Feel free to reopen if this needs further attention.