fix(python): resolve real interpreter for Process runner without pythonVersion

[jymaire]

What changes are being made and why?

Python Script/Commands tasks running on the Process task runner failed with /bin/sh: python: not found (exit code 127) on hosts that ship only python3 (Debian/Ubuntu, and the standard Kestra worker image).

Root cause: In PythonEnvironmentManager.setup(), the interpreter was only resolved to an actual installed binary when pythonVersion was explicitly set. When it was unset (the common case) on a Process runner, the interpreter stayed the literal "python", which doesn't exist on python3-only hosts. The Docker default runner was unaffected because python:3.13-slim symlinks python.

Fix: The Process runner now always resolves an installed interpreter. When pythonVersion is set, behavior is unchanged (resolved through the configured package manager, which may provision a managed Python via uv). When it's unset, we probe for an already-installed interpreter (python3 → python) via PackageManagerType.PIP.getPythonPath, avoiding a managed-Python download just to run a dependency-free script. The Docker/Kubernetes path is untouched.

Also:

• Updated the all_python sanity-check flow: python_commands_process now uses python3 main.py (user-authored command, not routed through the resolver).
• Added PythonEnvironmentManagerTest regression test asserting a Process-runner Script with no pythonVersion resolves to a runnable, non-"python" interpreter.

---

How the changes have been QAed?

• New PythonEnvironmentManagerTest passes.
• ScriptTest.task[PROCESS] — the exact case that reproduced the reported failure — now passes (previously exit 127).

id: repro
namespace: sanitychecks.plugin-scripts
tasks:
  - id: python_script_process
    type: io.kestra.plugin.scripts.python.Script
    taskRunner:
      type: io.kestra.plugin.core.runner.Process
    script: |
      print("Hello from a Process Task Runner")

On a python3-only host this failed with python: not found before the fix and now succeeds.

---

Contributor Checklist ✅

• I have read and followed the plugin contribution guidelines

[github-actions]

📦 Artifacts

Name	Size	Updated	Expiration
	56.71 MB	Jul 7, 26, 2:37:39 PM UTC	Jul 14, 26, 2:37:36 PM UTC

🧪 Java Unit Tests

	Tests	Passed ✅	Skipped	Failed	Time ⏱
Java Tests Report	239 ran	239 ✅	0 ⚠️	0 ❌	24m 44s 888ms

🔁 Unreleased Commits

✅ No unreleased commits found.

[github-actions]

Tests report quick summary:

success ✅ > tests: 239, success: 239, skipped: 0, failed: 0

unfold for details

Project	Status	Success	Skipped	Failed
plugin-script-bun	success ✅	3	0	0
plugin-script-deno	success ✅	3	0	0
plugin-script-dotnet	success ✅	8	0	0
plugin-script-go	success ✅	18	0	0
plugin-script-groovy	success ✅	8	0	0
plugin-script-jbang	success ✅	2	0	0
plugin-script-julia	success ✅	2	0	0
plugin-script-jython	success ✅	5	0	0
plugin-script-lua	success ✅	3	0	0
plugin-script-nashorn	success ✅	6	0	0
plugin-script-node	success ✅	13	0	0
plugin-script-perl	success ✅	3	0	0
plugin-script-php	success ✅	3	0	0
plugin-script-powershell	success ✅	4	0	0
plugin-script-python	success ✅	70	0	0
plugin-script-r	success ✅	2	0	0
plugin-script-ruby	success ✅	39	0	0
plugin-script-shell	success ✅	47	0	0

[jymaire]

QA

Flow

id: test_python_process_fix
namespace: sanitychecks.plugin_scripts

tasks:
    # ── THE FIX ──────────────────────────────────────────────────────────
    # Script on the Process runner, no pythonVersion.
    # BEFORE the fix: "/bin/sh: python: not found" (exit 127) on python3-only hosts.
    # AFTER: resolves python3 and prints its path.
    - id: script_process_default
      type: io.kestra.plugin.scripts.python.Script
      taskRunner:
        type: io.kestra.plugin.core.runner.Process
      script: |
        import sys
        print(f"Resolved interpreter: {sys.executable}")
        print(f"Version: {sys.version.split()[0]}")
        print("OK: ran on Process runner with no pythonVersion")

    # Same path, forcing PIP — this is the exact method the fix calls
    # (PackageManagerType.PIP.getPythonPath → probes python3 then python).
    - id: script_process_pip
      type: io.kestra.plugin.scripts.python.Script
      taskRunner:
        type: io.kestra.plugin.core.runner.Process
      packageManager: PIP
      script: |
        import sys
        print(f"Resolved interpreter: {sys.executable}")

    # Explicit pythonVersion on Process runner — unchanged behavior, must still work.
    - id: script_process_explicit_version
      type: io.kestra.plugin.scripts.python.Script
      taskRunner:
        type: io.kestra.plugin.core.runner.Process
      pythonVersion: "3.12"
      script: |
        import sys
        print(f"Resolved interpreter: {sys.executable}")

    # ── CONTROL ──────────────────────────────────────────────────────────
    # Docker default runner (python:3.13-slim) — always worked; must not regress.
    - id: script_docker_default
      type: io.kestra.plugin.scripts.python.Script
      script: |
        print("OK: ran on Docker runner")

Screen

[jymaire]

🧪 QA Report — ✅ PASS

Edition: OSS | Branch: develop (2.0.0-SNAPSHOT)

Verified end-to-end in the Kestra UI (Process runner executes on the worker host). This QA host has python3 but no python (which python → not found), i.e. it reproduces the exact "python3-only" condition from the bug report. Both tasks that were red in the original report now pass.

Flow	Result
test_python_process_fix (focused)	✅ SUCCESS — 3/3 tasks
all_python (repo sanity flow)	✅ SUCCESS — 9/9 tasks incl. assert

Before → after (root cause): previously the plugin built the command as python /…/script.py, which on a python3-only host fails with /bin/sh: python: not found (exit 127). After the fix, the Process runner resolves a real interpreter — the executed command is now python3.14 /…/script.py.

---

Flow 1: test_python_process_fix (✅ SUCCESS)

Flow YAML

id: test_python_process_fix
namespace: sanitychecks.plugin_scripts
tasks:
  - id: script_process_default        # Script + Process, no pythonVersion — THE fix
    type: io.kestra.plugin.scripts.python.Script
    taskRunner:
      type: io.kestra.plugin.core.runner.Process
    script: |
      import sys
      print(f"Resolved interpreter: {sys.executable}")
      print(f"Version: {sys.version.split()[0]}")
      print("OK: ran on Process runner with no pythonVersion")
  - id: script_process_pip             # Process + packageManager PIP (exact method the fix calls)
    type: io.kestra.plugin.scripts.python.Script
    taskRunner:
      type: io.kestra.plugin.core.runner.Process
    packageManager: PIP
    script: |
      import sys
      print(f"Resolved interpreter (PIP): {sys.executable}")
  - id: script_docker_default          # Docker runner — control / non-regression
    type: io.kestra.plugin.scripts.python.Script
    script: |
      print("OK: ran on Docker runner (control / non-regression)")

Gantt

Task	Runner	Status	Duration
script_process_default	Process (no pythonVersion)	SUCCESS	11.04s
script_process_pip	Process (PIP)	SUCCESS	0.07s
script_docker_default	Docker	SUCCESS	0.72s
Total		SUCCESS	12.01s

Logs synthesis

• script_process_default → executed python3.14 /…/script.py; logged Resolved interpreter: /usr/bin/python3.14, Version: 3.14.4. No python: not found.
• script_process_pip → executed python3.14 …; Resolved interpreter (PIP): /usr/bin/python3.14.
• script_docker_default → executed python /…/script.py inside the container (unchanged Docker path — image symlinks python), confirming no regression.

---

Flow 2: all_python (✅ SUCCESS) — literal reproduction of the reported failure

Flow YAML (repo sanity-check flow, with the python3 main.py fix)

# plugin-script-python/src/test/resources/sanity-checks/all_python.yaml
# Key tasks (the two that failed in the original report):
  - id: python_script_process
    type: io.kestra.plugin.scripts.python.Script
    taskRunner:
      type: io.kestra.plugin.core.runner.Process
    script: |
      print("Hello, World from a Process Task Runner")
  - id: python_commands_process
    type: io.kestra.plugin.scripts.python.Commands
    taskRunner:
      type: io.kestra.plugin.core.runner.Process
    inputFiles:
      main.py: |
        print("Hello, World from the Commands Task and Process Task Runner")
    commands:
      - python3 main.py     # <-- changed from `python main.py`

Gantt

Task	Status	Duration
python_script	SUCCESS	0.61s
python_script_process ⭐	SUCCESS	0.10s
python_commands	SUCCESS	0.49s
python_commands_process ⭐	SUCCESS	0.09s
python_script_logs	SUCCESS	25.40s
python_script_metrics	SUCCESS	21.44s
python_script_outputs	SUCCESS	13.85s
python_commands_etl_outputs	SUCCESS	1.34s
assert	SUCCESS	0.04s
Total	SUCCESS	~65s

⭐ = tasks that failed with python: not found (exit 127) in the original report.

Logs synthesis

• python_script_process → executed python3.14 /…/script.py → Hello, World from a Process Task Runner.
• python_commands_process → executed python3 main.py → Hello, World from the Commands Task and Process Task Runner.
• assert passed all exitCode == 0 conditions; python_commands_etl_outputs produced processed_orders.csv.

Outputs synthesis — python_script_outputs and python_commands_etl_outputs returned the total output and the processed_orders.csv output file, as asserted.

---

Timeouts / notes: none. The _logs/_metrics/_outputs tasks take 13–25s (network: pip install, HuggingFace dataset download, pydata image) but all completed and are unrelated to this fix. Topology/artifact testing not applicable (this PR does not add plugin artifacts).