fix(reasoning): detect bare READY markers and honor the last decision

[Kyou12138]

Summary

Fixes #6204.

Reasoning plan refinement only treated the full instructional phrase READY: I am ready to execute the task. as ready. Local/Ollama models often conclude with a short line such as READY / READY. / NOT READY, so the executor never left the refine loop even when the model had clearly finished planning.

Approach

Introduce response_indicates_ready() with:

1. Line-anchored markers — READY / NOT READY (optional trailing punctuation), case-insensitive
2. Last marker wins — after multi-step refinement, the final decision is used (NOT READY → READY, or the reverse)
3. Full instructional phrase as fallback — only when no line marker is present
4. No mid-sentence false positives — I am ready to begin researching… is not treated as ready

This is intentionally stricter than “substring contains READY”, which would mis-handle NOT READY.

Tests

lib/crewai/tests/utilities/test_response_indicates_ready.py covers:

• full instructional phrase
• bare READY / READY. / READY: …
• NOT READY must not match as ready
• last-marker-wins (both directions)
• mid-sentence “ready” ignored
• empty / unmarked plans

How to test

cd lib/crewai
pytest tests/utilities/test_response_indicates_ready.py -q

Notes / related PRs

Several community PRs also target #6204. Differentiation of this PR:

• explicit last-marker-wins semantics for refine loops
• NOT READY overrides a preceding full instructional READY phrase
• dedicated unit module with the cases above

Happy to close in favor of another PR if maintainers prefer a different approach.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 759a1fb9-e9f4-4c4e-957d-9d8cda4bbfd7

📥 Commits

Reviewing files that changed from the base of the PR and between ebe0082 and 324ac75.

📒 Files selected for processing (2)

• lib/crewai/src/crewai/utilities/reasoning_handler.py
• lib/crewai/tests/utilities/test_response_indicates_ready.py

---

📝 Walkthrough

Walkthrough

The reasoning handler adds shared detection for READY and NOT READY markers. Three response paths use this detector. Tests cover marker formats, precedence, false positives, empty responses, and planning parser integration.

Changes

Readiness detection

Layer / File(s)	Summary
Detector and response integration lib/crewai/src/crewai/utilities/reasoning_handler.py	Adds case-insensitive marker detection with last-marker precedence and updates function-calling, fallback, and static response parsing paths.
Readiness behavior validation lib/crewai/tests/utilities/test_response_indicates_ready.py	Tests full phrases, line markers, punctuation, case variations, marker precedence, empty responses, and planning parser readiness results.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main change: detecting bare READY markers and honoring the final readiness decision.
Description check	✅ Passed	The description directly explains the readiness detection bug, implementation approach, tests, and linked issue.
Linked Issues check	✅ Passed	The changes satisfy issue #6204 by detecting standalone READY output and correctly handling readiness markers during plan refinement.
Out of Scope Changes check	✅ Passed	The code and tests remain focused on readiness detection and planning response parsing described in issue #6204.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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.}

[Kyou12138]

Blocked on updating this branch to latest main: the GitHub OAuth token used for gh/git push lacks the workflow scope, and main now includes .github/workflows/* changes.

To unblock (one-time, on machine with Kyou12138 login):

gh auth refresh -h github.com -s workflow
cd crewAI
git fetch upstream main
git checkout <this-branch>
git rebase upstream/main
git push --force-with-lease origin <this-branch>

Or click Update branch in the GitHub UI (if the browser session has full permissions).

[Kyou12138]

Superseded by rebased PR #6882 (CountClaw fork; rebased onto latest main — original fork push blocked by missing workflow OAuth scope).

[Kyou12138]

Closing in favor of rebased #6882.