fix: post token tracking nits (follow up on #392)

[seayang-nv]

Summary

Addresses the four nit suggestions @binaryaaron left on #392 after the main token-tracking refactor was merged. No behavior change — all four are cleanups / readability improvements around code introduced in that PR.
Closes #426.

Changes

1. results.py — collapse repetitive token-field wiring

Replaced the 13-line block of num_valid_records = None, if isinstance(results, GenerateJobResults): num_valid_records = results.num_valid_records, ..., SafeSynthesizerSummary(num_valid_records=num_valid_records, ...) with:

• _GENERATE_RESULT_FIELDS tuple — names that pass through verbatim from GenerateJobResults to SafeSynthesizerSummary.
• _REPORT_SCORE_FIELDS dict — {summary_field: report_score_name} for report-derived scores.
  make_nss_summary now does a single getattr loop for the gen fields, computes valid_record_token_fraction inline, projects report scores (or None) via a dict comprehension, and **-unpacks both into one SafeSynthesizerSummary(...) call. ~45 lines gone.

2. batch.py — single-pass token aggregation

Added a private _record_token_totals() -> (valid, invalid) that walks self._responses exactly once. The three token properties all share it:

• total_valid_record_tokens / total_invalid_record_tokens destructure one side.
• total_non_record_tokens gets both and reuses them for the clamped-to-zero warning context (so the extra={} payload no longer re-triggers the scan).

3. processors.py — extract _valid_parsed helper

The comprehension [r.parsed for r in group_records if r.is_valid and r.parsed is not None] appeared twice inside GroupedDataProcessor._process_text_generation (non-unique-groupby check and order_by check). Hoisted into a module-level helper with a docstring explaining the is not None narrows the type for static analysis (valid records always carry parsed).

4. record_utils.py — reframe ParsedResponse docstring

Dropped the "backward compatibility with callers that want the old parallel-list shape" framing. The doc now describes valid_records / invalid_records / errors as convenience views that project records into the shapes expected by downstream aggregation code (parsed dicts / original text / (msg, validator) tuples).

Pre-Review Checklist

Ensure that the following pass:

• make format && make check or via prek validation.
• make test passes locally
• make test-e2e passes locally
• make test-ci-container passes locally (recommended)
• GPU CI status check passes -- comment /sync on this PR to trigger a run (auto-triggers on ready-for-review)

Pre-Merge Checklist

• New or updated tests for any fix or new behavior
• Updated documentation for new features and behaviors, including docstrings for API docs.

Other Notes

• Closes nits: readability / perf / doc cleanups after generation-token tracking (PR #392) #426

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Copilot]

Pull request overview

Follow-up cleanup PR that refactors recently introduced token-tracking and parsed-record codepaths for readability and minor efficiency improvements, without changing intended behavior (per PR description and review of the diffs).

Changes:

• Refactors make_nss_summary to populate SafeSynthesizerSummary via field lists/dicts instead of repetitive local wiring.
• Avoids redundant scanning of batch responses when computing non-record token totals and warning context.
• Extracts a shared helper for “valid parsed dicts” in grouped processing and updates ParsedResponse docstring to reflect current usage.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
src/nemo_safe_synthesizer/results.py	Collapses repetitive summary-field forwarding into _GENERATE_RESULT_FIELDS + _REPORT_SCORE_FIELDS driven dict construction.
src/nemo_safe_synthesizer/generation/batch.py	Introduces _record_token_totals() to reuse a single scan of responses for token aggregations (notably for total_non_record_tokens).
src/nemo_safe_synthesizer/generation/processors.py	Extracts _valid_parsed() helper to remove duplicated comprehensions in grouped record validation logic.
src/nemo_safe_synthesizer/data_processing/record_utils.py	Updates ParsedResponse docstring to describe valid_records/invalid_records/errors as primary convenience views.

---

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