Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

[Bug]: Muse: continuation turns drop the thread's reasoning effort

Open Beginner friendly
  • Is something wrong?
#17,502 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
2/5
Estimated time
Half a day
Newbie friendliness
72/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Active
Tech stack
typescript
Domain
backend

Research direction

The bug is in the Muse adapter's start function under apps/server, where effort is set to undefined for continuation turns, so reasoningEffort is dropped from the turn/start request. Read how selectionEffort and targetModel are handled, then check whether skipping effort on continuations was intentional. Done when a continuation turn sends the thread's selected effort, and a test or logged request shows reasoningEffort present.

Written by the indexing model from the issue text.

Description

Before submitting
  • I searched existing issues and did not find a duplicate.
  • I included enough detail to reproduce or investigate the problem.
Area

apps/server

Steps to reproduce
  1. Select a Muse model with reasoningEffort set to max on a thread.
  2. Send a first turn that makes Muse launch a background workflow tool call. That run ends while the workflow keeps running.
  3. Wait for the workflow to finish. T3 starts a continuation run ("Background task completed.").
  4. Inspect the turn/start request of the continuation run, or use a PreLLMCall hook that logs the effort.
Expected behavior

The continuation turn reuses the thread's selected reasoning effort, the same as a user turn does.

Actual behavior

The continuation request carries no reasoningEffort, although the thread's modelSelection has one (for example max). The continuation silently runs at the provider default.

Impact

Major degradation or frequent failure

A PreLLMCall hook that requires an explicit effort blocks the continuation run. The run ends failed and the lead never posts its final reply.

Version or commit

T3 Code 0.0.46-nightly.20261009.2861

Environment

Linux, Muse Code CLI 1.4.4

Logs or stack traces

Code pointer, from the compiled server binary (Muse adapter, start):

const continuation = turnInput.message.createdBy === "agent" && turnInput.message.creationSource === "provider";
const adopted = continuation ? wake : void 0;
const effort = continuation ? void 0 : yield* selectionEffort(turnInput.modelSelection);
...
...effort ? { reasoningEffort: effort } : {}

For continuation turns effort is always void 0, so reasoningEffort is left out of the request. selectionEffort(turnInput.modelSelection) is only evaluated for non-continuation turns. targetModel is skipped the same way, so the model is kept from the session, but the effort is not.

Workaround

None found on the T3 side. Sending a manual user message after the failed continuation starts a normal turn with the selected effort.

Dominant language
TypeScript
Stars
24.8k
Forks
6.4k
Avg merge
10h 41m
Merged PRs (30d)
352

Getting set up

Open in Codespaces

Starts the project's dev container in your browser, under your own GitHub account.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from pingdotgg/t3code

All issues in pingdotgg/t3code

Similar issues

More TypeScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.