feat(livekit): make reasoning middleware configurable per model - #1902
Draft
icycodes wants to merge 4 commits into
Draft
feat(livekit): make reasoning middleware configurable per model#1902icycodes wants to merge 4 commits into
icycodes wants to merge 4 commits into
Conversation
The reasoning middleware was only enabled by a hardcoded well-known model list, so users could not turn it on for their own <think>-emitting models. Expose it as a `useReasoningMiddleware` model option (like `useToolCallMiddleware`), keeping the well-known list as fallback. 🤖 Generated with [Pochi](https://getpochi.com) | [Task](https://app.getpochi.com/share/p-c1444bad2f084203a178cac40505c024) Co-Authored-By: Pochi <noreply@getpochi.com>
Models such as GLM emit signed thinking blocks like <think signature="...">, which the reasoning middleware previously failed to detect because it only matched the bare <think> tag. Parse attributes on the opening tag and keep them in the reasoning part provider metadata so they are not lost. 🤖 Generated with [Pochi](https://getpochi.com) | [Task](https://app.getpochi.com/share/p-c1444bad2f084203a178cac40505c024) Co-Authored-By: Pochi <noreply@getpochi.com>
A multi-turn round-trip test showed the extracted reasoning was replayed as an SDK reasoning part, so providers either dropped it or mapped it to a field the model never wrote. Re-serialize assistant reasoning into the original tag (attributes included) before the request leaves the client. 🤖 Generated with [Pochi](https://getpochi.com) | [Task](https://app.getpochi.com/share/p-c1444bad2f084203a178cac40505c024) Co-Authored-By: Pochi <noreply@getpochi.com>
Codecov Report❌ Patch coverage is ❌ Your patch status has failed because the patch coverage (40.00%) is below the target coverage (75.00%). You can increase the patch coverage or adjust the target coverage. Additional details and impacted files@@ Coverage Diff @@
## main #1902 +/- ##
=======================================
Coverage 37.02% 37.02%
=======================================
Files 770 770
Lines 81194 81209 +15
Branches 3646 3646
=======================================
+ Hits 30065 30071 +6
- Misses 51014 51023 +9
Partials 115 115 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
useReasoningMiddlewaremodel flag so reasoning-tag extraction can be enabled/disabled per model, wired through the config schema, display-model mapping, and LLM request data.<think>-style reasoning tags that carry attributes (not just a bare<think>tag), preserving them for later use.assets/config.schema.jsonto include the newuseReasoningMiddlewarefield.Test plan
bun run testinpackages/livekit(reasoning-middleware tests cover tag-with-attributes parsing and the multi-turn echo-back round trip)bun check/bun tscat repo root🤖 Generated with Pochi | Task