fix(managed_files): skip managed file rows with no file_object when listing files

[devin-ai-integration]

TLDR

Problem this solves:

• GET /v1/files 500s on one null file_object row
• Poller-registered batch output/error ids create those rows
• Whole listing is lost, not just the row

How it solves it:

• Parse each managed file row defensively, skip unparseable ones
• Handles the JSON-string form the writer stores too

Relevant issues

Fixes #35361

Linear ticket

Pre-Submission checklist

• I have added meaningful tests
• My PR passes all CI/CD checks (e.g., lint, format, unit tests)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

Screenshots / Proof of Fix

I could not stand up the DB-backed proxy needed for the end to end repro in this environment: managed files require Postgres, and both apt and the Docker registry are unreachable from here, so there is no way to get a database up. Reviewer runbook for a live repro is below; the regression test stands in as the automated proof (fails before the fix with the reported OpenAIFileObject() argument after ** must be a mapping, not NoneType, passes after)

Before the fix (71b825a7f0 plus the test only):

$ PYTHONPATH=./enterprise LITELLM_LOCAL_MODEL_COST_MAP=True uv run --no-sync pytest \
    tests/test_litellm/enterprise/proxy/test_managed_files_hook.py -q -k skips_rows
FAILED tests/test_litellm/enterprise/proxy/test_managed_files_hook.py::test_get_user_created_file_ids_skips_rows_without_file_object
E   TypeError: litellm.types.llms.openai.OpenAIFileObject() argument after ** must be a mapping, not NoneType
1 failed

After the fix (88aea5306f):

$ PYTHONPATH=./enterprise LITELLM_LOCAL_MODEL_COST_MAP=True uv run --no-sync pytest \
    tests/test_litellm/enterprise/proxy/test_managed_files_hook.py -q -k skips_rows
1 passed

The four other failures in that file (test_should_pass_credentials_to_afile_retrieve, test_should_fallback_when_no_router, test_should_not_double_wrap_already_unified_output_file_id, test_afile_content_bedrock_unified_id_end_to_end) fail identically on the unmodified base commit in this environment; they are missing optional deps here, not regressions from this change

Live repro for a reviewer with a database, on a proxy with enable_preview_features: true, database_url set and CheckBatchCost enabled:

1. curl -X POST http://localhost:4000/v1/files -H "Authorization: Bearer sk-1234" -F purpose=batch -F 'target_model_names=my-gpt' -F file=@batch.jsonl
2. curl -X POST http://localhost:4000/v1/batches -H "Authorization: Bearer sk-1234" -H 'Content-Type: application/json' -d '{"input_file_id": "<id from step 1>", "endpoint": "/v1/chat/completions", "completion_window": "24h"}'
3. Wait for the batch to complete and for one CheckBatchCost cycle, which registers the batch's output_file_id and error_file_id as managed ids with file_object null
4. curl "http://localhost:4000/v1/files?target_model_names=my-gpt" -H "Authorization: Bearer sk-1234"; before this change that returns a 500 naming NoneType, after it returns 200 with the parseable files and without the poller-registered rows

Type

🐛 Bug Fix

Changes

store_unified_file_id deliberately accepts file_object=None (LiteLLM_ManagedFileTable.file_object is Optional, and the writer only sets the column when the value is non-null), and the batch cost poller uses exactly that path when it re-registers a completed batch's output_file_id / error_file_id as managed ids. The reader disagreed: get_user_created_file_ids did OpenAIFileObject(**row.file_object) for every row, so one null column raised TypeError and the whole GET /v1/files response became a 500 for that caller

Rows are now parsed through a small helper that returns None for a null or unparseable column and logs a warning, and the listing skips those, matching how alist_batches already tolerates rows it cannot parse. The helper also accepts the JSON-string form, which is what the writer actually persists (file_object.model_dump_json() into a Json column), so a string column no longer depends on Prisma decoding it back into a dict

def _parse_managed_file_object(raw_file_object: object, unified_file_id: str) -> Optional[OpenAIFileObject]:
    if not raw_file_object:
        return None
    try:
        return (
            OpenAIFileObject.model_validate_json(raw_file_object)
            if isinstance(raw_file_object, str)
            else OpenAIFileObject.model_validate(raw_file_object)
        )
    except Exception as e:
        verbose_logger.warning(f"Failed to parse managed file object {unified_file_id}: {e}")
        return None

No backfill is needed; a poller-registered output row carries no provider metadata worth showing

One nearby inconsistency I did not touch, since it is a separate bug: delete_unified_file_id returns initial_value.file_object annotated as OpenAIFileObject while the DB gives back the raw column, so DELETE /v1/files/{id} can hand a JSON string or None straight to the response model

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[greptile-apps]

Greptile Summary

Fixes managed-file listing so one null or malformed stored file object does not fail the entire request.

• Adds defensive parsing for both JSON-string and object representations of OpenAIFileObject.
• Omits rows whose file objects are absent or cannot be parsed.
• Adds regression coverage for valid string/object rows alongside null and malformed rows.

Confidence Score: 5/5

The PR appears safe to merge, with the reported listing failure handled defensively and covered by a targeted regression test.

The parser supports the repository’s required Pydantic version and both persisted file-object shapes, while isolating invalid rows instead of allowing one row to fail the complete file listing.

Important Files Changed

Filename	Overview
enterprise/litellm_enterprise/proxy/hooks/managed_files.py	Adds per-row parsing and omission of unusable managed-file objects, consistently handling the persisted JSON-string representation without introducing an actionable defect.
tests/test_litellm/enterprise/proxy/test_managed_files_hook.py	Adds focused mock-based regression coverage confirming valid rows remain listed while null and malformed rows are skipped.

_{Reviews (1): Last reviewed commit: "fix(managed_files): skip managed file ro..." | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[mateo-berri]

Closing since #35365 merged the same null file_object skip in get_user_created_file_ids, so this is superseded