16 views
-/https://github.com/berriai/litellm/issues/20975
GitHub · issue

#20975 [Bug]: LiteLLM Proxy - Responses API streaming omits necessary SSE event types

  • State: open
  • Author: @jackcooney44
  • 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?

When streaming Azure Responses API requests through the LiteLLM proxy, it sends `response.output_text.delta` events but **omits the required setup events** (`response.created`, `response.in_progress`, `response.output_item.added`, `response.content_part.added`), causing clients to reject the deltas.

Endpoints: `/openai/v1/responses` `/v1/responses` `/responses`

This was discovered when using an Azure model via the Lite LLM Proxy with Codex. Error: `codex_core::util: OutputTextDelta without active item`

**Expected SSE sequence (sourced from Azure OpenAI response directly):** ``` {"type":"response.created",...} {"type":"response.in_progress",...} {"type":"response.output_item.added",...} {"type":"response.content_part.added",...} {"type":"response.output_text.delta","delta":"Hello",...} {"type":"response.output_text.delta","delta":" there",...} {"type":"response.output_text.done",...} {"type":"response.content_part.done",...} {"type":"response.output_item.done",...} {"type":"response.completed",...} ``` **Ac…

GitHub resolver

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

Refresh page
vote history (4 events)
#0 of 0 · 31d18h46m50s ago — entered · #import:https:::github.com:berriai:litellm post #2107
The left issue requires coordinated changes to streaming translation, event ordering, provider compatibility, and regression coverage across the Responses API, with greater protocol and integration risk. The right issue is a localized routing fix with comparatively straightforward handler and test updates.
The right issue is harder because it requires tracing and correcting streaming event generation across the Responses API translation layer, preserving protocol ordering and compatibility across endpoints and providers, with comprehensive stream-level tests. The left issue is comparatively localized to customer deletion lifecycle cleanup and a Redis key operation.
The right issue is harder because it requires tracing and preserving a protocol-level streaming lifecycle across proxy and provider translation paths, with ordering, compatibility, and end-to-end streaming tests. The left issue is more localized to deployment replacement and router-state cleanup, though it still carries consistency and regression risk.
#0 of 0 · 31d18h1m27s ago — current · #import:https:::github.com:berriai:litellm post #2849
The Responses API issue is harder because it requires reconstructing and validating a complete streaming event lifecycle across proxy translation paths, with ordering, state, and client-compatibility concerns. The MCP issue is comparatively localized to making progress-token handling type-safe and adding focused regression coverage.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search