fix(runner): restore caller stdout instead of raw sys.__stdout__

[aadya940]

Problem

When log_file_path is set on a verb, runner.run() installs its own _Tee over sys.stdout. The finally block then restores with:

sys.stdout = sys.__stdout__

sys.__stdout__ is the raw original stdout — not whatever the caller had installed. If the caller had replaced sys.stdout with their own tee (e.g. a workflow runner that logs all output to a file), this line silently destroys it. Every print() call after the verb returns goes to the void.

Symptom

In a workflow that tees sys.stdout to a log file and runs multiple orbit verbs in sequence, the log file is truncated after the first verb's completion banner. All subsequent node labels, status checks, and error tracebacks are swallowed silently. The run shows status=error in the DB with no visible cause.

Fix

Save sys.stdout before touching it, restore to that:

_prev_stdout = sys.stdout  # save caller's stdout (may be a workflow-level tee)
if self.log_file_path:
    ...
    sys.stdout = _Tee(sys.__stdout__, log_f)
...
finally:
    if log_f:
        sys.stdout = _prev_stdout  # restore caller's stdout, not raw __stdout__
        log_f.close()

Two-line change, no behaviour change when log_file_path is not set.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Fixes Agent.run()’s stdout restoration when log_file_path is enabled so Orbit doesn’t clobber a caller-provided sys.stdout (e.g., a workflow-level tee/capture) after the run completes.

Changes:

• Capture the caller’s current sys.stdout before installing Orbit’s _Tee.
• Restore sys.stdout back to the captured value in the finally block (instead of sys.__stdout__) when closing the log file.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.