6 views
-/https://github.com/berriai/litellm/issues/25848
GitHub · issue

#25848 [Bug]: Litellm v1.83.8, v1.83.3, v1.82.2 Claud Code v2.1.110, v2.1.92, v2.1.52 OpenAI-compatible Translation Empty

  • State: open
  • Author: @lkaupp
  • Labels: bug, proxy, llm translation, claude code

### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

OpenAI compatible endpoint returned output is empty, but the call of the api is successfull. In the repsonse json the message is returned, however the output in "pretty" is empty. For that reason, claude code tested in versions v2.1.110, v2.1.92, v2.1.52 and littellm versions: v1.83.8, v1.83.3, v1.82.2 do not deliver the answer to "hi".

Here is the response from the server: ``` { "id": "resp_bGl0ZWxsbTpjdXN0b21fbGxtX3Byb3ZpZGVyOm9wZW5haTttb2RlbF9pZDoxNDc1NjNlNGZmNDRkNjkyMzliNGY4MDJkZGVhOWE3NDYzMzYzMzg2ZWQyNzI5ZWU5ZmU4NDYwMjAxN2JjNTFkO3Jlc3BvbnNlX2lkOnJlc3BfYmQ0OTg4NzRiNTJkZTg0Zg==", "text": null, "user": "user_a4e122d37d0f4540af7d7944c9ebb8a2ba530f6ae907ac7047429bd570c", "error": null, "model": "qwen3.5-397b-a17b", "store": null, "tools": [ { "name": "Task", "type": "function", "strict": null, "parameters": { "type": "object", "$schema": "https://json-schema.org/draft/2020-12/schema", "required": [ "description", "prompt", "subagent_type" …

GitHub resolver

Import GitHub neighbors on demand. Results are saved as system ingests.

Refresh page
vote history (2 events)
#0 of 0 · 31d18h14m35s ago — entered · #import:https:::github.com:berriai:litellm post #2611
The left issue requires tracing response normalization across multiple compatibility layers and client versions, with higher uncertainty around streaming, schema, and provider-specific behavior. The right issue is comparatively contained in logging configuration and level propagation, likely involving a focused proxy/runtime change and regression tests.
#0 of 0 · 31d18h12m48s ago — current · #import:https:::github.com:berriai:litellm post #2641
The left issue is substantially harder: it requires cross-cutting design across persistence, resource relationships, serialization, secret handling, portability, compatibility, and likely both API and CLI workflows, with broad integration and regression testing. The right issue is more likely a focused translation-path regression requiring diagnosis and a localized fix.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search