fix(providers): stop OpenRouter/HF stream timeout on reasoning models

[Ocean82]

One of the three provider bugs blocking launch. qwen/qwen3.6-27b streams reasoning tokens (delta.reasoning) before any content. The openaiCompatible stream parser only forwarded delta.content, so the caller's 30s first-byte timeout (providers.ts) fired during the reasoning phase — every OpenRouter/HuggingFace fallback timed out even though data was flowing.

Fix: emit one empty liveness ping on the first delta.reasoning chunk (bypassing the thinking-tag filter, which swallows empty strings) so the first-byte timeout disarms. No reasoning text is leaked; content streams normally.

Tests: +2 (reasoning-before-content pings exactly once; content-first path does not ping). Server suite 325/325.

Related: the other two provider bugs — Groq reasoning_format (fixed by #24's deploy, verified live 200 + clean output) and Ollama NUM_CTX=2048 overriding the Modelfile (fixed on the box, set to 8192) — are resolved. Groq is primary and healthy. Option 2 (actually suppress reasoning on these fallback providers, vs just not-timing-out) is logged in PRODUCTION-TODO as low-priority polish.

Summary by Sourcery

Keep OpenRouter and Hugging Face reasoning-model streams alive until content arrives without exposing reasoning output.

Bug Fixes:

• Prevent OpenRouter and Hugging Face streaming fallbacks from timing out while reasoning models emit reasoning before content.
• Keep reasoning tokens out of user-visible output while allowing normal content streaming.

Documentation:

• Document the remaining low-priority work to suppress reasoning generation on OpenRouter and Hugging Face fallbacks.

Tests:

• Add coverage for a single liveness ping during reasoning-before-content streams and no ping for content-first streams.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[sourcery-ai]

Reviewer's Guide

The PR updates the OpenAI-compatible SSE parser to treat the first reasoning-only delta as a one-time empty liveness signal, preventing first-byte timeouts while continuing to suppress reasoning content; tests cover both reasoning-first and content-first streams, and the remaining provider-specific reasoning suppression work is documented.

Sequence diagram for reasoning-stream liveness ping

sequenceDiagram
    participant Provider as OpenAI-compatible provider
    participant Parser as chatWithOpenAiCompatibleStream
    participant Caller as Stream caller

    Provider->>Parser: SSE delta.reasoning
    Parser->>Caller: onChunk('')
    Note over Parser,Caller: One-time ping disarms first-byte timeout
    Provider->>Parser: SSE delta.reasoning
    Note over Parser: Reasoning text is suppressed
    Provider->>Parser: SSE delta.content
    Parser->>Caller: onChunk(content)
    Parser-->>Caller: Return accumulated content

Loading

File-Level Changes

Change	Details	Files
Keep fallback streams alive during reasoning-only phases without exposing reasoning text.	Track whether a reasoning liveness ping has been sent. Detect delta.reasoning chunks when no content is present and invoke the callback once with an empty string, bypassing the thinking-tag filter. Continue accumulating and filtering only delta.content, preserving normal output behavior.	server/src/openaiCompatible.ts
Add coverage for reasoning-first and content-first streaming behavior.	Mock SSE streams with reasoning before content and verify exactly one empty ping, clean accumulated output, and correct callback chunks. Verify content-first streams produce no empty ping and preserve the complete response.	server/src/openaiCompatible.test.ts
Document remaining reasoning suppression as a separate fallback-provider polish task.	Record that timeout handling is fixed while reasoning generation remains unsuppressed on OpenRouter/HuggingFace. Note provider-specific request parameters and defer implementation pending live API verification.	docs/PRODUCTION-TODO.md

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Team

Run ID: 4bf8cecc-91d0-4419-b542-a3f2838abebe

---

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

[sourcery-ai]

Hey - I've reviewed your changes and they look great!

Sourcery assessment

Approved.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook