[chopratejas]

Description

ProxyManager.buildLaunchSpecs() (plugins/openclaw/src/proxy-manager.ts) has a Python fallback: it probes with python -c "...find_spec('headroom')..." and starts the proxy with python -m headroom.cli proxy .... Both put the current directory first on sys.path, and the spawned process inherits OpenClaw's working directory, so a headroom/ folder there was found and run instead of the installed package.

Both the probe and the launch now pass -P (Python 3.11+: do not prepend the cwd to sys.path). Interpreters older than 3.11 reject -P, so their probe fails and the launcher moves on; headroom-ai requires 3.11+ so they could not hold a real install. -P rather than -I keeps PYTHONPATH and user-site installs working.

Companion to the same change in plugins/headroom-agent-hooks/bin/headroom-hook.sh (#4842); kept separate because this plugin has its own test suite.

Type of Change

• Bug fix (non-breaking change that fixes an issue)

Changes Made

• plugins/openclaw/src/proxy-manager.ts: -P on the configured-python and fallback-python launch and probe args.
• plugins/openclaw/test/proxy-manager.test.ts: updated expected args; new test that runs each Python spec's probe with a real Python 3.11+ (site-packages disabled) from a directory containing a headroom/ package and expects "not found".

Testing

• Unit tests pass (vitest)
• Linting passes (ruff check .) (no Python changed)
• Type checking passes (npm run typecheck)
• New tests added for new functionality
• Manual testing performed

Test Output

# new test on origin/main (fix stashed), python3 = 3.12
$ npx vitest run test/proxy-manager.test.ts -t "working directory"
AssertionError: expected +0 to be 1 // Object.is equality
      Tests  1 failed | 34 skipped (35)

# with the fix
$ npx vitest run test/proxy-manager.test.ts
      Tests  35 passed (35)
$ npx vitest run
      Tests  2 failed | 161 passed (163)
# the 2 failures are test/advancement-key-store.test.ts "...two separate OS processes..."; they fail identically on origin/main (2 failed | 160 passed) and are unrelated.
$ npm run typecheck
(no output, exit 0)

Real Behavior Proof

• Environment: macOS, Node + vitest from the plugin's lockfile, Python 3.12.
• Exact command / steps: in a temp dir containing headroom/__init__.py, run each Python launch spec's probe as python3 -S <checkArgs> (-S so no installed headroom can be seen).
• Observed result: main exits 0 (the folder in the cwd is accepted as headroom); this branch exits 1 (not found), so the launcher would move to the next spec.
• Not tested: Windows py launcher (it forwards -P to the interpreter).

Runtime Rollout Safety

• Rollout-managed feature(s): none
• Minimum rollout channel: n/a
• Stable/default behavior changed: the Python fallback only starts an installed headroom; a source checkout reachable only through the cwd is no longer picked up (set pythonPath/PYTHONPATH, or put headroom on PATH).
• Kill switch / disable path: a headroom executable on PATH or an explicit proxyUrl bypasses the fallback.
• Unsafe override required: no
• Qualification impact: none
• Rollback path: revert this PR

Review Readiness

• I have performed a self-review
• This PR is ready for human review

Checklist

• My code follows the project's style guidelines
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I did not edit CHANGELOG.md

Additional Notes

No docs change needed.

🤖 Generated with Claude Code

---

Reported by @JoaoMarcos44. Thank you for your contribution! 🩵

[github-actions]

PR governance

This PR follows the template and is marked ready for human review.

[JerrettDavis]

The Python fallback change looks correct. I independently ran all 35 proxy-manager tests and TypeScript checking successfully. Real Python probes from a directory containing headroom/init.py exit 0 with the target-branch implementation and exit 1 with this change, confirming that cwd discovery is excluded. The PYTHONPATH review finding is resolved by -E in the regression test. I found no additional code defect. Qualification remains held by the inherited shared workflow provenance audit. Separately, the current commitlint job failed; its diagnostic log is needed to determine the cause before readiness can be established. Windows py and the full plugin suite were not independently tested.

[JerrettDavis]

The Python launcher fix and its regression test remain supported by the prior independent 35-test run, TypeScript check, and real Python head/base probes at this same commit. The newly available CI log identifies the remaining blocker: commit e75532b (chore: merge main into fix/openclaw-python-launcher-path) has a 116-character Assay-Ask footer. Commitlint job 132208024108 in run 43518951621 fails with footer-max-line-length; it requires at most 100 characters per footer line. Please reword that commit to wrap the footer and obtain a passing commitlint result. For example:

Assay-Ask: The dependency audit failed because this branch predates
 the dependency updates on main.

No additional source defect found. Windows py and the full plugin suite were not independently tested. Inherited workflow provenance issues remain separately recorded and do not drive this change request under the supplied review-only baseline.

[JerrettDavis]

The previously reported commit-message blocker remains at the current head. CI run 43547215469, commitlint job 132301219755, fails with footer-max-line-length on commit e75532b. Its Assay-Ask footer is 116 characters; the limit is 100. Please reword that commit to wrap the footer and obtain a passing commitlint result. For example:

Assay-Ask: The dependency audit checks failed because this branch predates
 the dependency updates on main.

I rechecked both changed files against the current target and found no additional source defect. Earlier-head passing tests are context only; no new unit-test run is claimed. Inherited workflow provenance holds are recorded separately and are not the reason for this change request.

[chopratejas]

Jerrett, the commitlint blocker is cleared. The footer-length rule fix (#4868) is merged into main, and I merged main into this branch, so commitlint now passes on this head. No code changes in this update.