fix(research): keep decimals intact when splitting citation claims

[vishnujayvel]

What & why

buildCitationGraph splits synthesis text into claims with a regex that treats every . as a sentence boundary. Decimal amounts in the report ($62.3 billion, 27.7 percent) were therefore cut mid-number, so citation_graph[].claim started after the decimal ("3 billion...", "7 percent...") while the report body itself stayed correct.

Fixes #274.

Credit: @oQAQo1 for the repro and root-cause analysis (sentence split at digit.digit in src/research/citation-graph.ts).

Changes

• A digit-flanked . is no longer treated as a sentence boundary, so decimal numbers stay inside the claim.
• Regression tests cover a single decimal, multiple decimals in one sentence, and correct multi-sentence splitting.
• Abbreviations (e.g., U.S.) remain a known limitation of this splitter (out of scope for this fix).

Testing

• npx vitest run tests/unit/research/citation-graph.test.ts passes (14 tests)
• npx tsc --noEmit passes
• Added/updated tests for the new behavior
• Revert-proof: with the src change stashed, the new tests fail with claim text starting mid-number ("3 billion in Q4 FY2026..."), matching the issue symptom

Checklist

• Follows Conventional Commits
• Focused change (no unrelated refactoring)
• I have read CONTRIBUTING.md and agree to its contribution terms

Summary by CodeRabbit

• Bug Fixes

  • Improved sentence splitting to preserve decimal values, preventing periods in numbers from being treated as sentence breaks.
  • Maintained accurate citation mapping and confidence information across sentences containing multiple decimal values.

• Tests

  • Added regression coverage for decimal numbers in citation claims and multi-sentence content.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 88db79c2-0bf7-4ee6-9204-4c2da2408814

📥 Commits

Reviewing files that changed from the base of the PR and between b3ccf92 and db9df57.

📒 Files selected for processing (2)

• src/research/citation-graph.ts
• tests/unit/research/citation-graph.test.ts

---

📝 Walkthrough

Walkthrough

The citation graph sentence splitter now preserves decimal values. Regression tests verify complete claims, correct sentence boundaries, source-index mappings, and confidence levels for decimal-containing research synthesis text.

Changes

Citation decimal handling

Layer / File(s)	Summary
Decimal-safe sentence splitting and regression coverage src/research/citation-graph.ts, tests/unit/research/citation-graph.test.ts	The sentence matcher no longer splits on periods between digits. Tests verify single and multiple decimal values, sentence separation, source indices, and confidence levels.

Estimated code review effort: 2 (Simple) | ~10 minutes

🚥 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 and concisely describes preserving decimal values during citation-claim sentence splitting.
Linked Issues check	✅ Passed	The implementation and regression tests address issue #274 by preserving decimals and maintaining sentence splitting, source mapping, and confidence behavior.
Out of Scope Changes check	✅ Passed	All changes support the linked issue and PR objective; the abbreviation note and focused regression tests are directly relevant.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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.}

[vishnujayvel]

Thanks for the review, and for independently probing $62.3 / 27.7 / $5.8 and
Version 1.2.3 shipped yesterday [1]. rather than taking the test file's word for it.

To restate the limitation plainly so it isn't lost if this gets revisited: abbreviations
(e.g., U.S.) still fragment. That is pre-existing behaviour, unchanged by this PR, and fixing
it needs an abbreviation list rather than a lookaround — a different change with a different risk
profile. Happy to open a follow-up issue for it if you'd find that useful.

[vishnujayvel]

Thanks @Frankie-Xu for the careful reviews and re-verification — and for confirming the e.g./U.S. fragmenting is a pre-existing limitation rather than a regression from this change.