fix(workflow): dedupe progress rows, close leaked coroutines, cap runaway loops

[elkaix]

Summary

Three bugs found in a post-merge review of #186 (Workflow tool):

• render_progress() re-listed agents that had been truncated by the per-phase
  max_agents cap as duplicate "unphased" rows at the bottom of the progress
  block — a phase with 8 agents and max_agents=6 showed the two oldest
  agents twice. Fixed by filtering the trailing "unphased" section on phase
  membership instead of a rendered-id set.
• parallel() raised on invalid arguments (non-list, or a bare function mixed
  into the list) without closing the agent() coroutines already created for
  the other list items, leaking "coroutine was never awaited" warnings.
• The Python port of the reference implementation dropped its 1000-agent
  lifetime backstop. A workflow script with an unbounded loop (e.g.
  while True: await agent(...)) and no token_budget set could spawn
  subagents indefinitely. Restored the cap as MAX_TOTAL_AGENTS = 1000,
  checked inside the concurrency semaphore alongside the existing budget
  check.

No behavior changes outside these three fixes; no new dependencies.

Test plan

• Added regression tests: truncated-phase dedup, phaseless-agent
  rendering, lifetime-cap enforcement, and coroutine-leak-on-validation-
  failure (via warnings.catch_warnings)
• make check-pythinker-code (ruff + format + pyright + ty) — clean
• make test-pythinker-code — 6505 passed, 7 skipped, 1 xfailed (unit);
  65 passed, 4 skipped (e2e)
• CHANGELOG.md updated under ## Unreleased

Summary by CodeRabbit

• Bug Fixes
  • Fixed workflow progress rendering so truncated agents no longer reappear later in the output.
  • Improved progress display for agents without a phase, ensuring they still appear when appropriate.
  • Added a safeguard to stop runaway workflow loops after a high number of agent runs.
  • Prevented invalid workflow input from leaving behind unclosed async tasks, reducing warning noise and resource leaks.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: a61d754f-03bc-483f-81e4-cd2785740e1f

📥 Commits

Reviewing files that changed from the base of the PR and between 3c44c7e and f149543.

📒 Files selected for processing (5)

• CHANGELOG.md
• src/pythinker_code/tools/workflow/display.py
• src/pythinker_code/tools/workflow/engine.py
• tests/tools/test_workflow_display.py
• tests/tools/test_workflow_engine.py

---

📝 Walkthrough

Walkthrough

This PR fixes duplicate agent rows in workflow progress rendering by switching truncation logic from ID-based tracking to phase-order membership, closes leaked coroutines when parallel() rejects invalid arguments, and adds a 1000-agent lifetime cap to stop runaway workflow loops. Includes regression tests and a changelog entry.

Changes

Workflow progress and engine fixes

Layer / File(s)	Summary
Phase-based truncation fix in render_progress src/pythinker_code/tools/workflow/display.py, tests/tools/test_workflow_display.py	Replaces the rendered ID-set with phase_order membership checks so truncated phase agents no longer reappear as unphased rows; adds tests for truncated and phaseless agent rendering.
Agent lifetime cap in engine src/pythinker_code/tools/workflow/engine.py, tests/tools/test_workflow_engine.py	Adds MAX_TOTAL_AGENTS = 1000 and enforces it in agent(), raising WorkflowRuntimeError once the cap is hit; adds a monkeypatched regression test.
Coroutine cleanup on parallel() validation failure src/pythinker_code/tools/workflow/engine.py, tests/tools/test_workflow_engine.py	Calls _close_coroutines before raising in parallel() for invalid arguments or non-awaitable callables; adds a test confirming no "never awaited" warnings.
Changelog entry CHANGELOG.md	Documents the three fixes under Unreleased.

Estimated code review effort: 2 (Simple) | ~12 minutes

Sequence Diagram(s)

sequenceDiagram
  participant Caller
  participant Parallel
  participant CloseCoroutines
  Caller->>Parallel: parallel(items)
  Parallel->>Parallel: validate items
  alt invalid argument or item
    Parallel->>CloseCoroutines: _close_coroutines(items)
    Parallel-->>Caller: raise WorkflowRuntimeError
  else valid
    Parallel-->>Caller: proceed with execution
  end

Loading

Suggested labels: bug

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title follows conventional commits and accurately summarizes the workflow fixes.
Description check	✅ Passed	The description is detailed and covers summary, test plan, and checklist items, with only the related issue section omitted.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
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.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/workflow-progress-dup-and-runaway-cap

---

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

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/pythinker_code/tools/workflow/engine.py	80.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!