Build a reviewable filing evidence ledger without losing the original quote

A research spreadsheet often keeps a number but loses the sentence that defines it. Start this exercise with the quote and its location instead. The fixture contains three invented disclosure excerpts; it is not a real company filing.

1. Run the supplied fixture

Install Python 3.10 or newer, download or clone the repository, then run these commands from its root. There are no runtime dependencies.

python filing_evidence.py examples/sample.json --output examples/output
python -m unittest discover -p "test_*.py"

You can inspect the input JSON before running it. E01 says “Revenue was USD 12 million for FY2025.” Its stored value is the string 12.00, with USD, million and FY2025 kept as separate context fields. The tool does not convert the number or infer a different currency.

2. Read the output as a handoff, not a verdict

3 evidence rows; 2 need field review. Sources not verified.

The generated ledger.md groups the evidence by research question. ledger.json preserves the input rows and adds field checks. E02 has a value and unit but no period, so its checklist contains missing:period. E03 has a location but no quotation, producing missing:quote. E01 has no field-completeness issue; that does not mean its claim is true.

3. Complete the human review

For your own research, open the actual disclosure, confirm the wording and location, and fill missing context. Use a real source URL rather than the synthetic fixture's self-reference. Set status to reviewed only to record your own work; the tool keeps this separate from its checks.

When the question grows beyond one excerpt, Eqvero provides the related public-disclosure research workflow. This MIT utility remains only a local record organizer, not source verification or an investment recommendation.