Explain automatically detected filing values

[nonprofittechy]

Summary

• Replace extraction-based Recommended labels with * markers in cascading choices.
• Add an accessible help icon and modal explaining automatic detection and its possible mistakes.
• Show the same marker and explanation on the final review page for extracted case, document, and party values.

Matching accuracy (added after review)

Marking a value as "detected from your document" is a claim, so this PR also tightens the matching that decides which option gets the marker. This slightly broadens the original scope, but the old heuristics would have made the new marker say things that were not true.

• Courts are matched on whole words, not substrings. The extracted court guess was compared to court codes with an unanchored in, so a short county name inside a longer one won: a document reading "McHenry County" put Henry County at the top of the list and pre-selected it. Williamson matched Will, Whiteside matched White, Crawford matched Ford. Court codes drop the spaces inside a county name (stclair, rockisland), so the comparison is now against runs of whole words. That also lets a real caption match — "Circuit Court of Cook County, Illinois" now finds the Cook courts, where before it matched none.
• An ambiguous guess is prioritized but not chosen. When the guess names only the county, every division of that county matches and none of them is what the document said, so they sit at the top of the list unselected instead of the case being filed in whichever division sorts first. A guess that names the division still selects it.
• Levenshtein distance is gone from the other dropdowns. At a threshold loose enough to forgive a typo it also pairs unrelated options — "Motion" and "Notice" are three edits apart — and the marker turned that into a claim about the filer's document. Substring matching still covers the near misses that mattered ("Eviction" finds "Evictions"). efile/utils/str_dist.py had no other callers and is removed.
• The case lookup help text waits for a marker to exist. It was gated on there being a court guess, not on that guess matching a court, so it could promise an asterisk that appeared nowhere on the page. (Copilot's finding.)
• A docket number must match exactly to be marked. A filer who corrects 2026-CV-1234 to 2026-CV-123 has not left our guess in place, and the substring rule was still calling it automatically detected.
• The modal wording now reflects that several options in a list can carry the marker.
• The review page fetches its parties once instead of three times. (Copilot's finding.)

Validation

• Repository pre-commit hooks passed: Ruff, djLint, Stylelint, ESLint, type checking, Bandit, and migration checks.
• uv run pytest -q — 664 passed.
• npm run test:unit — 43 passed.
• Local Playwright browser smoke test passed against the Django server with LocalStack and the extraction worker running.

Screenshots

Screenshots are stored outside the repository in this public gist:

https://gist.github.com/nonprofittechy/68589d297e9072f1b27a9205b417de8d

🤖 Generated with Claude Code

https://claude.ai/code/session_01JCsBkAUTE8jxP4BrbPexAm

[Copilot]

🟡 Changes recommended

Court prioritization currently conflates extraction guesses with location matching, which can incorrectly apply “(Recommended)” labeling to extraction-derived suggestions instead of the intended * marker.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR updates the e-filing workflow UI to represent automatically detected (document-extracted) values with a compact * marker and provides an accessible, reusable help affordance (icon + modal) explaining what the marker means and that extraction can be wrong.

Changes:

• Replace extraction “(Recommended)” labeling with * markers in dropdown prioritization and downstream saved display strings.
• Add a shared help button and Bootstrap modal explaining the extraction marker, and surface it across relevant workflow pages.
• Mark extracted values on the final review page (case fields, parties, documents) and add test coverage for the new marker behavior.

File summaries

File	Description
efile_app/efile/views/review.py	Computes extracted_markers to flag extracted values for the final review page.
efile_app/efile/views/case_lookup.py	Adjusts guessed court source to use extraction guesses only.
efile_app/efile/api/dropdown_views.py	Switches extraction matches to * marker; refines court prioritization logic for extraction vs location.
efile_app/efile/tests/test_review_submit_flow.py	Adds an integration-style assertion that extracted markers and modal copy appear on final review.
efile_app/efile/tests/test_filing_types_amount_in_controversy.py	Adds a unit test verifying dropdown option text uses * instead of “Recommended”.
efile_app/efile/templates/efile/workflow_base.html	Includes the extraction marker modal globally in the workflow layout.
efile_app/efile/templates/efile/review.html	Displays * markers next to extracted case/doc/party values and shows page-level help when relevant.
efile_app/efile/templates/efile/organize_documents.html	Shows extraction marker help when a guessed filing type is present.
efile_app/efile/templates/efile/extraction_review.html	Adds extraction marker help icon in the extracted document section header.
efile_app/efile/templates/efile/components/extraction_marker_modal.html	New reusable modal explaining the asterisk marker.
efile_app/efile/templates/efile/components/extraction_marker_help.html	New reusable help-icon button that opens the modal.
efile_app/efile/templates/efile/case_lookup.html	Adds page-level help text/icon when an extracted court suggestion exists.
efile_app/efile/static/js/organize-documents.js	Normalizes saved option text by stripping “(Recommended)” and * suffixes.
efile_app/efile/static/js/extraction-review.js	Normalizes saved/displayed option text while still rendering the * marker when applicable.
efile_app/efile/static/js/case-lookup.js	Normalizes saved court name to avoid persisting UI markers into stored text.
efile_app/efile/static/css/reorganized-flow.css	Styles the page help row, help icon button, and extraction section header layout.

Review details

• Files reviewed: 16/16 changed files
• Comments generated: 2
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🟡 Changes recommended

There are a couple of user-facing/behavioral issues identified in the changed hunks (including a misleading help message condition and a small inefficiency in the review view) that should be addressed before approval.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Review details

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

efile_app/efile/views/review.py:70

• all_parties = list(parties) forces an immediate queryset evaluation but the list is only used once to build party_ids. You can iterate the existing parties queryset directly here and drop the extra list allocation.

• Files reviewed: 18/18 changed files
• Comments generated: 1
• Review effort level: Lite

[Copilot]

🔵 Needs a closer look

The final-review document marker logic can incorrectly mark supporting documents as “detected from your document” even though extraction data is only sourced from the current lead document.

Review details

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

efile_app/efile/views/review.py:83

• draft.extracted_guesses is only sourced from the current lead document’s extraction, but this logic marks every document whose filing_type_name matches the extracted filing type. If multiple documents share the same filing type, the review page will show * on supporting documents even though they were not the basis for the extraction, which can mislead filers.

• Files reviewed: 19/19 changed files
• Comments generated: 0 new
• Review effort level: Lite