fix(streaming): don't attach finish_reason to content-bearing chunk

[EnjoyBacon7]

Problem

stream_with_source_filtering (openrag/components/utils.py) emits the final
content of a streamed chat completion in a chunk that also carries
finish_reason: "stop".

At end-of-stream the function builds the tail chunk via
copy.deepcopy(chunk_template), and chunk_template at that point is the
upstream finish chunk (the one whose choices[0].finish_reason is set).
The resulting tail chunk therefore contains both the remaining delta.content
and finish_reason: "stop":

[0] finish_reason='stop'  content="Bonjour ! Comment puis-je vous aider ..."   <-- both
[1] finish_reason='stop'  content=None
[2] [DONE]

Spec-compliant OpenAI streaming clients treat a chunk with a non-null
finish_reason as terminal and ignore its delta. Such clients (e.g. the
Twake Workplace RAG integration) therefore drop the tail content:

• short answers render blank (the whole answer fits in the held-back
  look-ahead window and is flushed only in the finish-tagged tail chunk);
• long answers lose their final ~look-ahead window of text (silent
  truncation).

The HTTP response is still 200, so the regression is invisible server-side.
Behaviour was correct in 1.1.7, which emitted content chunks with
finish_reason: null and a separate empty finish chunk.

Fix

Clear finish_reason on the content-bearing tail chunk. The separate finish
chunk that immediately follows still emits finish_reason with an empty
delta, matching the OpenAI streaming contract.

[0] finish_reason=None    content="Bonjour ! Comment puis-je vous aider ..."
[1] finish_reason='stop'  content=None
[2] [DONE]

Verification

• Short answer: full content delivered in a non-terminal chunk.
• Long answer (202 chunks): 0 content-chunks-carrying-finish_reason, single
  finish chunk, complete text (no truncation), sources still delivered in
  extra.
• Non-streaming path unaffected.

Summary by CodeRabbit

• Bug Fixes
  • Fixed streamed response termination so trailing content is no longer mistaken for the final signal. Streams now emit the content tail separately from the final termination notification, ensuring clients do not prematurely close and the final finish message carries the correct filtered metadata.

[coderabbitai]

Warning

Review limit reached

@EnjoyBacon7, we couldn't start this review because you've reached your PR review rate limit.

More reviews will be available in 3 minutes and 23 seconds. Learn how PR review limits work.

Your organization has run out of usage credits. Purchase more credits in the billing tab to continue.

⌛ How to resolve this issue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans include higher PR review limits than trial, open-source, and free plans. In all cases, reviews become available again over time. During sustained high-volume PR review activity, CodeRabbit may temporarily slow when the next review becomes available.

Please see our Fair Usage Limits Policy for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: d95ec5c8-f6ab-4bcb-a025-06e1b00e1794

📥 Commits

Reviewing files that changed from the base of the PR and between 7d45bd2 and 714f2a8.

📒 Files selected for processing (1)

• openrag/components/utils.py

📝 Walkthrough

Walkthrough

The PR nullifies choices[0]["finish_reason"] on the final content tail chunk in stream_with_source_filtering during the [DONE] path so the tail chunk is non-terminal; a separate finish chunk still carries the terminal finish_reason.

Changes

Stream finish reason handling fix

Layer / File(s)	Summary
Tail chunk finish reason separation openrag/components/utils.py	When emitting the final tail chunk during [DONE] handling, choices[0].finish_reason is now explicitly set to None so the chunk carries only delta content while a separate finish chunk later provides the terminal finish_reason.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• linagora/openrag#257: Both PRs modify stream_with_source_filtering end-of-stream chunk emission logic, specifically how the buffered tail content and finish chunk terminal signaling are handled.

Suggested labels

fix

Suggested reviewers

• paultranvan

Poem

🐰 I nudged the tail to speak, then hush —
The finish waits, no final rush,
Content hops out, calm and clear,
The terminal note comes after, near,
A rabbit's nod to proper flow ✨

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and specifically describes the main fix: preventing finish_reason from being attached to content-bearing chunks in streaming responses.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/streaming-finish-reason-on-content-chunk

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}