[https://nvbugs/6060281][fix] Treat whitespace-only content in nano-v3 reasoning swap

[tijyojwad]

…(6060281)

When chat_template_kwargs enables force_nonempty_content, NemotronV3ReasoningParser promotes reasoning into content if the parsed content string is empty. Apply the same fallback when content is only whitespace (e.g. a lone newline after the closing think tag), matching the non-whitespace empty case.

Add unit coverage for newline- and space-only trailing content.

Made-with: Cursor

Summary by CodeRabbit

Bug Fixes

• Improved reasoning content parsing to properly handle cases where content contains only whitespace or newlines.
• Enhanced content validation logic to ensure consistent behavior when processing empty or whitespace-only text segments.

Description

Test Coverage

PR Checklist

Please review the following before submitting your PR:

• PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.

• PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.

• Test cases are provided for new code paths (see test instructions)

• Any new dependencies have been scanned for license and vulnerabilities

• CODEOWNERS updated if ownership changes

• Documentation updated as needed

• Update tava architecture diagram if there is a significant design change in PR.

• The reviewers assigned automatically/manually are appropriate for the PR.

• Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.

[tijyojwad]

/bot run --disable-fail-fast

[coderabbitai]

📝 Walkthrough

Walkthrough

Updated the reasoning parser to treat whitespace-only content after closing tags as empty, enabling proper reasoning-to-content swaps. Added corresponding test cases to verify this edge case handling.

Changes

Cohort / File(s)	Summary
Reasoning Parser Logic tensorrt_llm/llmapi/reasoning_parser.py	Modified _maybe_swap_content method to normalize result.content and check content.strip() instead of raw content, ensuring content with only whitespace/newlines triggers the reasoning-to-content swap.
Parser Tests tests/unittest/llmapi/test_reasoning_parser.py	Extended test_nano_v3_reasoning_parser parametrized test with two new cases covering whitespace-only content scenarios (newline and spaces/tabs).

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description check	✅ Passed	The description clearly explains the issue (whitespace-only content not triggering the swap) and solution, with test coverage mentioned; however, the Description and Test Coverage sections of the template are not fully filled out as standalone sections.
Title check	✅ Passed	The title accurately describes the main change: treating whitespace-only content in the nano-v3 reasoning parser's swap logic, which is the core purpose of the modifications to both the implementation and tests.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[tensorrt-cicd]

PR_Github #42623 [ run ] triggered by Bot. Commit: 59f7c6c Link to invocation

[tensorrt-cicd]

PR_Github #42623 [ run ] completed with state SUCCESS. Commit: 59f7c6c
/LLM/main/L0_MergeRequest_PR pipeline #33341 completed with status: 'SUCCESS'

CI Report

Link to invocation