8 views
-/https://github.com/berriai/litellm/issues/28458
GitHub · issue

#28458 [Bug]: websearch_interception not working for OpenAI provider when stream=true

  • State: open
  • Author: @novotl
  • Labels: bug, proxy, llm translation

### Check for existing issues

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

### What happened?

Web search interception doesn't work with OpenAI provider when `stream: true`. I would expect it to work the same way it does with `stream: false`

### Steps to Reproduce

1. Configure search tool & web_search interception as per https://docs.litellm.ai/docs/integrations/websearch_interception 2. Test interception works when `stream=false` ``` curl -X 'POST' \ 'https://{litellm-host}/v1/chat/completions' \ -H 'accept: application/json' \ -H 'x-litellm-api-key: sk-XXXXXX' \ -H 'Content-Type: application/json' \ -d '{ "model": "openai-gpt-5.1", "messages": [ { "role": "user", "content": "What is the weather in San Francisco today?" } ], "tools": [ { "type": "function", "function": { "name": "litellm_web_search", "description": "Search the web for information, always include URL in citations", "parameters": { "type": "object", "properties": { "query": { "type": "string" } }, "required": [ …

GitHub resolver

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

Refresh page
vote history (5 events)
#0 of 0 · 31d19h18m54s ago — entered · #import:https:::github.com:berriai:litellm post #1661
#31343 is harder because it spans cross-provider fallback behavior, tool-schema normalization, Agents SDK compatibility, and broader regression coverage; #28458 is more localized to the streaming interception path.
The left requires broader cross-cutting protocol, authentication, routing, configuration, compatibility, and test work, with greater standards-compliance and security risk. The right is comparatively localized debugging in an existing streaming translation path.
The right-hand task spans streaming control flow, provider translation, interception behavior, and regression coverage across multiple execution paths. The left-hand task is comparatively localized: add a guarded post-call persistence decision and validate rate-limit edge cases.
The right-side issue is harder because it likely requires tracing provider-specific streaming, interception, and tool-call handling across translation and response-generation paths, with broader regression coverage. The left-side issue is comparatively localized to UI state/diagnostic logic and targeted tests.
#0 of 0 · 31d18h3m41s ago — current · #import:https:::github.com:berriai:litellm post #2901
The left requires a cross-cutting authorization-model change with recursive resolution, persistence and update propagation, cycle/consistency safeguards, compatibility handling, and broad testing. The right is narrower debugging of an existing request-processing path, mainly involving streaming behavior and regression coverage.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search