fix(runtime): handle input=None in generic handler

[deanq]

Summary

• Fix AttributeError crash when malformed job sends {"input": null} to generic handler
• Changed job.get("input", {}) to job.get("input") or {} in both create_handler and create_deployed_handler — the default {} only applies when the key is missing, not when it's explicitly null
• Added tests for input=None and missing input key scenarios

Test plan

• test_create_handler_input_none — verifies graceful error response instead of crash
• test_create_handler_input_missing — verifies missing input key handled
• All 2491 existing tests pass
• Coverage at 85.79%

Fixes AE-2317

[Copilot]

Pull request overview

Fixes a runtime crash in the generic handler when a malformed job payload provides {"input": null} by ensuring handler code can safely proceed without AttributeError.

Changes:

• Update create_handler and create_deployed_handler to treat job["input"] == None as an empty dict.
• Add unit tests covering input=None and missing input key for both handler factories.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
src/runpod_flash/runtime/generic_handler.py	Normalizes job["input"] to {} when falsy to avoid .get() crashes.
tests/unit/runtime/test_generic_handler.py	Adds regression tests for input=None and missing input scenarios for both handler types.

Comments suppressed due to low confidence (1)

src/runpod_flash/runtime/generic_handler.py:176

• With input=None now normalized to {}, function_name becomes None and the handler returns "Function 'None' not found...". Consider explicitly detecting a missing/empty function_name and returning a clearer error (e.g., that job['input']['function_name'] is required) to make malformed-job debugging easier.

        job_input = job.get("input") or {}
        function_name = job_input.get("function_name")
        execution_type = job_input.get("execution_type", "function")

        if function_name not in function_registry:
            return {
                "success": False,
                "error": f"Function '{function_name}' not found in registry. "

---

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

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

---

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

[runpod-Henrik]

PR #286 — fix(runtime): handle input=None in generic handler

Repo: runpod/flash
Severity: LOW
Key findings: Crash on null input is fixed. Two remaining gaps: missing test for one handler variant, and inconsistent error schemas between the two handler types.

---

1. Missing test: deployed handler with no input key

create_deployed_handler now handles the case where the input key is absent from the job entirely (job = {}). There is no test for this path. The create_handler variant has an equivalent test, but the deployed handler variant only tests {"input": null}. If the missing-key path is broken, nothing will catch it.

User scenario: a caller sends a malformed job payload to a deployed endpoint with no input field — this path fires and is untested.

2. Inconsistent error schema between the two handler types

When input is invalid, the two handlers return different error shapes. One includes a success: false field; the other returns only {"error": ...} with no success key. If any client code checks response["success"] on the deployed handler's error response, it gets a KeyError instead of the expected boolean.

This inconsistency pre-existed for the crash path and is now extended to the new type-guard branches. This PR is a good opportunity to align them.

Nits

• test_create_handler_input_none passes because null input routes through a "function not found" path rather than an explicit null-input rejection path. A comment in the test explaining why "not found" is the expected message would prevent future confusion about what this test is actually verifying.

Verdict

PASS WITH NITS — the crash fix is correct. Add a test for create_deployed_handler with a missing input key, and consider aligning the error schema between the two handlers before the mismatch causes a client-side bug.

🤖 Reviewed by Henrik's AI-Powered Bug Finder