fix(jobs): forward transformed plugin job specs in JSON mode

[arpitsardhana]

Summary

• Fix HTTP 500 on evaluator durable-job submit when metrics use cloudpickle bundles (non-UTF-8 binary payloads in EvaluateSpec).
• Dump transformed plugin job specs with model_dump(mode="json") before forwarding to core jobs so bytes fields serialize as base64 instead of raw bytes.
• Add unit test asserting the create-job route forwards JSON-safe transformed specs.

Fixes: https://nvbugspro.nvidia.com/bug/6466868

Root cause

The generated plugin create_job route re-dumped transformed specs in Python mode. Cloudpickle metric bundle blob fields stayed as raw bytes, which failed JSON serialization in the typed jobs client (PydanticSerializationError: invalid utf-8 sequence).

Test plan

• uv run --frozen pytest packages/nemo_platform_plugin/tests/test_jobs_create_spec_json.py -v
• uv run python repro_cloudpickle_eval.py against a live local platform (optional manual check)

NVBugs #6466868

Summary by CodeRabbit

• Bug Fixes
  • Fixed job creation so Pydantic job specifications are serialized using JSON-mode, ensuring binary-like fields are sent as base64-encoded strings.
  • Improved compatibility with typed clients by guaranteeing the submitted job payload is fully JSON-serializable.
• Tests
  • Added a regression test covering JSON-mode serialization for transformed job specs containing binary data.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 21bbf9f7-878e-4c30-b688-5b1e29ffdbb1

📥 Commits

Reviewing files that changed from the base of the PR and between b5877d3 and 1f5561c.

📒 Files selected for processing (1)

• packages/nemo_platform_plugin/tests/test_jobs_create_spec_json.py

🚧 Files skipped from review as they are similar to previous changes (1)

• packages/nemo_platform_plugin/tests/test_jobs_create_spec_json.py

---

📝 Walkthrough

Walkthrough

Changes

Job specification JSON serialization

Layer / File(s)	Summary
JSON-mode job specification serialization packages/nemo_platform_plugin/src/nemo_platform_plugin/jobs/api_factory.py	create_job now uses model_dump(mode="json") so binary-like fields are serialized for JSON transport.
Binary payload regression coverage packages/nemo_platform_plugin/tests/test_jobs_create_spec_json.py	Adds a FastAPI job-creation test verifying binary output is base64-encoded and the forwarded payload is JSON-serializable.

Possibly related PRs

• NVIDIA-NeMo/nemo-platform#585: Also changes job_route_factory job-spec serialization in api_factory.py.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main fix: forwarding transformed job specs in JSON mode.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix-nvbugs-july-16-6466868

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@packages/nemo_platform_plugin/src/nemo_platform_plugin/jobs/api_factory.py`:
- Around line 876-877: Update the transformed-spec handling around job_spec and
spec_dict to enforce JSON-safe output for both BaseModel and raw-dict results
before constructing the SDK body. Configure or apply base64 normalization for
binary fields, including non-UTF-8 bytes, and validate raw transformer
dictionaries rather than passing them through unchanged.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 708c7baf-71bd-433a-8a3a-09c9772b6ce3

📥 Commits

Reviewing files that changed from the base of the PR and between 2412ab7 and b5877d3.

📒 Files selected for processing (2)

• packages/nemo_platform_plugin/src/nemo_platform_plugin/jobs/api_factory.py
• packages/nemo_platform_plugin/tests/test_jobs_create_spec_json.py

[github-actions]

Suite	Lines Covered	Line Rate	Branch Rate
Unit Tests	25533/32734	78.0%	62.7%
Integration Tests	14731/31383	46.9%	19.2%