Skip to content

Support Chat Completions mode for OpenAI-compatible providers #33

@Hanjo92

Description

@Hanjo92

Problem\nThe OpenAI provider currently always sends requests to the Responses API path. Users can configure an OpenAI-compatible base URL such as a gateway or local model server, but many compatible servers primarily support /v1/chat/completions rather than /v1/responses. In that case connectivity checks can be fast while real authenticated review requests repeatedly time out.\n\nObserved local settings include provider openai, base URL https://llm.vibeprogramminglab.com/v1, and an Ollama-style model name gemma4:e4b, which strongly suggests an OpenAI-compatible chat-completions gateway rather than the official Responses API.\n\n## Acceptance Criteria\n- Users can choose OpenAI request format: Responses API or Chat Completions.\n- Chat Completions mode calls /chat/completions for base URLs that do not already include that path.\n- Response parsing supports the chat completions response shape.\n- Settings UI exposes the mode near the OpenAI base URL/model fields.\n- Timeout diagnostics remain clear and do not expose API keys or request bodies.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions