docs: Cherry-pick Fern release documentation build.

[FrankD412]

Cherry-picks #1031 that fixes the release documentation publishing for Fern docs.

Summary by CodeRabbit

• New Features

  • Added dry-run capability for documentation releases via the fern-release-dryrun make target.

• Bug Fixes

  • Improved error handling for malformed API responses—now gracefully degrades and returns null instead of raising exceptions.
  • Dashboard plots now skip gracefully when required data is unavailable instead of failing.

• Tests

  • Added comprehensive test coverage for error handling with malformed/missing data scenarios.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

Try out this PR

Quick install:

pip install --upgrade --force-reinstall git+https://github.com/ai-dynamo/aiperf.git@0f8791e850a235865be53b707e6c32ba72dfc9d2

Recommended with virtual environment (using uv):

uv venv --python 3.12 && source .venv/bin/activate
uv pip install --upgrade --force-reinstall git+https://github.com/ai-dynamo/aiperf.git@0f8791e850a235865be53b707e6c32ba72dfc9d2

---

Last updated for commit: 0f8791e • Browse code

[coderabbitai]

Caution

Review failed

Failed to post review comments

Walkthrough

This PR enhances resilience across three domains: OpenAI response parsing now degrades gracefully for malformed bodies, plot generation skips unavailable metrics without crashing, and the Fern documentation release workflow gains strict validation and local dry-run tooling for pre-publication snapshots.

Changes

Documentation Release Workflow and Tooling

Layer / File(s)	Summary
Workflow dispatch input and release setup .github/workflows/fern-docs.yml	Added force_rebuild boolean input to workflow_dispatch. Reworked release-version job to check out tagged source and initialize isolated docs-checkout worktree.
Content transformation, validation, and publish .github/workflows/fern-docs.yml	Updated release to convert markdown from tagged commit using md_to_mdx.py, generate version config from tag's docs/index.yml, insert strict validation (broken-link checks) before push, and adjust publish step working directory.
Dry-run tool for local validation Makefile, tools/fern_release_dryrun.sh	New fern-release-dryrun make target and standalone shell script for local, offline snapshot validation: rebuilds versioned pages, converts markdown, generates config, validates with Fern CLI, and reports success without authentication.

Graceful Degradation in OpenAI Response Parsing

Layer / File(s)	Summary
Endpoint parsing behavior changes src/aiperf/endpoints/openai_chat.py, src/aiperf/endpoints/openai_completions.py	Modified chat and completions endpoints to return None instead of raising ValueError for unrecognized OpenAI object field values, enabling soft degradation for arbitrary/error/proxy response bodies.
Endpoint unit test coverage tests/unit/endpoints/test_chat_endpoint_parse_response.py, tests/unit/endpoints/test_completions_endpoint_parse_response.py, tests/unit/endpoints/test_openai_chat_attack.py	Added parametrized tests verifying endpoints return None for malformed bodies with missing or error-shaped object fields. Updated chat attack test to expect None instead of ValueError. Tests construct specific malformed payloads using orjson and pytest.param.
Parser integration tests for resilience tests/unit/records/test_inference_result_parser.py	Added end-to-end test feeding malformed response payloads through real ChatEndpoint (not mocked). Verifies parser records failures as InvalidInferenceResultError without crashing and confirms no "Unsupported OpenAI object type" error messages appear.

Plot Generation Error Handling for Missing Data

Layer / File(s)	Summary
Error handling in dashboard and exporter src/aiperf/plot/dashboard/callbacks.py, src/aiperf/plot/exporters/png/single_run.py	Dashboard callbacks and PNG exporter now catch DataUnavailableError and KeyError during plot creation. Missing-data scenarios log at debug level and return None (dashboard) or skip export (exporter) instead of propagating errors.
Plot generation test coverage tests/unit/plot/test_dashboard_callbacks.py, tests/unit/plot/test_png_exporter.py	Added fixtures for non-streaming runs and helper for plot specs. Dashboard tests verify missing-column requests return None with debug logging. PNG exporter test confirms missing metric columns are skipped without errors while other plots render successfully.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Poem

🐰 A rabbit hops through docs with care,
Graceful parsing floats on air,
Plots skip stones in data streams,
Release pipelines chase their dreams!
Soft degradation reigns supreme. 🚀

🚥 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 accurately describes the main objective: cherry-picking changes related to Fern release documentation build, which is confirmed by the commit messages and file changes focusing on release documentation workflow and related fixes.
Docstring Coverage	✅ Passed	Docstring coverage is 87.50% which is sufficient. The required threshold is 80.00%.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!