Skip to main content

Custom API server did not respond

Feature Request

Hi Vowen team,

I’m trying to connect a local OpenAI-compatible endpoint via “Custom API (OpenAI-compatible)”, but the Connect step fails with:

“Custom API server did not respond to either chat/completions or responses format. Check your Base URL and model name.”

My config in Vowen:

  • Base URL: http://127.0.0.1:20128 (also tested with /v1)

  • Default model: kr/claude-sonnet-4.5

  • API key: tested both empty and provided

The backend is reachable and responds correctly:

  1. GET /v1/models returns a valid model list including:

  • kr/claude-sonnet-4.5

  • kiro/claude-sonnet-4.5

  • kr/claude-haiku-4.5

  • kiro/claude-haiku-4.5

  1. POST /v1/chat/completions returns valid OpenAI-style responses.

  2. POST /v1/responses returns valid OpenAI Responses API output.

From testing, it looks like the connection validator may be rejecting valid OpenAI-compatible responses in this setup (possibly related to streaming/SSE handling or strict response parsing during Connect).

Could you please confirm what exact validation request Vowen sends during Connect and what response shape/content-type it expects?

I can provide full request/response samples if needed.

Posted by Дмитрий Салямон ·
Status: Completed2 comments

Log in to comment and vote

Comments2

  • Дмитрий Салямон

    •

    Apr 30

    Hey team, quick update: I managed to resolve it with help from Codex.

    Root cause was a response-format mismatch during Vowen’s Custom API validation:

    • Vowen expected JSON from chat/completions / responses

    • OmniRoute was often returning SSE unless stream:false was forced

    After validating endpoints and request format, I switched to a compatible local setup and the connection now works correctly in Vowen.

  • Vowen

    Team•

    Apr 30

    Hey @Дмитрий Салямон , can you provide which local LLM you are using and the curl requests for these endpoints, I will test them out and get back to you. Thanks