fix(openai): handle missing 'id' field in streaming chunks for MiniMax

[chengyongru]

• Change chunk["id"] to chunk.get("id") for compatibility with MiniMax and other OpenAI-compatible providers
• ModelResponseStream auto-generates id when None is passed
• Add regression test test_chunk_parser_without_id_field

Relevant issues

Fixes KeyError: 'id' when using MiniMax m2.5 model with streaming responses

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Type

🐛 Bug Fix
✅ Test

Changes

Problem

When using MiniMax m2.5 model (or other OpenAI-compatible providers), streaming responses may include chunks without an id field. The code at gpt_transformation.py:809 used chunk["id"] which throws KeyError: 'id' when the field is missing.

Solution

• Changed chunk["id"] to chunk.get("id") in litellm/llms/openai/chat/gpt_transformation.py
• When id is None, ModelResponseStream automatically generates one via _generate_id()

Files Changed

• litellm/llms/openai/chat/gpt_transformation.py - Fix KeyError by using safe dict access
• tests/test_litellm/llms/openai/chat/test_openai_gpt_transformation.py - Add regression test

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Mar 18, 2026 2:50am

[CLAassistant]

All committers have signed the CLA.

[greptile-apps]

Greptile Summary

This PR fixes a KeyError: 'id' crash in the OpenAI streaming handler when an OpenAI-compatible provider (such as MiniMax m2.5) returns streaming chunks that omit the id field. The one-line change from chunk["id"] to chunk.get("id") is the correct minimal fix — ModelResponseStream already calls _generate_id() when id=None is supplied, so no additional changes are needed downstream.

Key changes:

• litellm/llms/openai/chat/gpt_transformation.py: chunk.get("id") replaces chunk["id"] at line 809, making chunk_parser tolerant of missing id fields from any OpenAI-compatible provider.
• tests/test_litellm/llms/openai/chat/test_openai_gpt_transformation.py: A new pure-mock regression test test_chunk_parser_without_id_field verifies the fix without making any real network calls, consistent with the test-folder policy.

Confidence Score: 5/5

• This PR is safe to merge — it is a minimal, well-tested, non-breaking bug fix.
• The change is a single-line defensive dict access that only affects previously-broken code paths (missing id field). Existing callers that supply id are completely unaffected. The fix is backed by ModelResponseStream's existing _generate_id() fallback, and a focused mock regression test validates the behavior end-to-end with no real network calls.
• No files require special attention.

Important Files Changed

Filename	Overview
litellm/llms/openai/chat/gpt_transformation.py	Single-line fix replacing chunk["id"] with chunk.get("id") in chunk_parser; ModelResponseStream already handles None by auto-generating an id via _generate_id().
tests/test_litellm/llms/openai/chat/test_openai_gpt_transformation.py	Adds a pure mock regression test test_chunk_parser_without_id_field that constructs a chunk dict without an id field and asserts no KeyError is raised and an id is auto-generated. No real network calls are made.

Sequence Diagram

sequenceDiagram
    participant Provider as MiniMax / OpenAI-compatible Provider
    participant Handler as OpenAIChatCompletionStreamingHandler
    participant MRS as ModelResponseStream

    Provider->>Handler: Streaming chunk (no 'id' field)
    Note over Handler: chunk.get("id") → None<br/>(was chunk["id"] → KeyError)
    Handler->>MRS: ModelResponseStream(id=None, ...)
    Note over MRS: id is None → _generate_id()
    MRS-->>Handler: ModelResponseStream(id="chatcmpl-<generated>", ...)
    Handler-->>Provider: Parsed chunk with auto-generated id

Loading

_{Last reviewed commit: "fix(openai): handle ..."}

[codspeed-hq]

Merging this PR will not alter performance

✅ 16 untouched benchmarks

---

_{Comparing chengyongru:fix/minimax-streaming-keyerror-id (ce3f24c) with main (cfeafbe)}