Add the legacy /completions endpoint  #170

Activity

1. akushonkamen commented on Oct 10, 2026

   

   akushonkamen

   on Oct 10, 2026

   Contributor

   More actions

   Hi @codelion — I'd like to pick this up. The built-in inference server currently only exposes the chat protocol at optillm/server.py /v1/chat/completions (server.py:759); there is no legacy text-completions route.

   Plan: extract the request-parsing/dispatch core of proxy() into a shared internal helper (no behavior change for chat), then add POST /v1/completions that maps prompt → messages and returns the OpenAI legacy object: "text_completion" shape (choices[].text + usage, multiple choices for list prompts / n>1). Stream: first version returns 400 with a clear message (can follow up). Offline unit tests via server.app.test_client() following tests/test_approaches.py. PR incoming shortly — please confirm the first-version scope looks right.

   👍React with 👍1Reacted by Asankhaya Sharma
2. codelion

   closed this as completedin 086c678on Oct 10, 2026
3. codelion

   added a commit that references this issue on Oct 10, 2026

   feat(server): add legacy /v1/completions endpoint for the inbuilt inf…

   ...

   This issue will close once commit 086c678 is merged into the 'main' branch.

   086c678