[sliu382]

After every conversation, the responder files a short form: topics, an outcome, and how confident they felt (TDD 4.3.2). It has no free text, so it can feed analytics without holding anything a student wrote. Coordinators manage the topic and outcome choices, and choices get retired instead of deleted so old reports stay valid. This ticket adds all three tables and the routes for the options. Submitting a report comes later.

Changes:

• Added ReportOption, ConversationReport, and ConversationReportTopic models plus ReportOptionKind enum in models/report.py
• ConversationReport has a check that confidence is between 1 and 5. ConversationReport.conversation_id is also unique
• Made a migration for new models
• Added a GET report-options in portal for staff. Returns only active options and is ordered by display_order
• Added admin routes for GET, POST, and PATCH. Did not add DELETE.
• Added schema for report options
• Added a Dev seed with 8 topics and 4 outcomes
• Tested that retired options are hidden from the portal but all options are listed for admins.
• Tested that the ordering was correct based on display_order.
• Tested that the database will only accept one report per conversation and also tested that the database rejects confidence ratings outside of 1-5.

Follow-up: In the report submission POST, need to verify that the selected outcome has kind "OUTCOME". The current foreign key only verifies that the option exists. Codex says application validation would be the simplest fit based on the current schema

Closes #96

[github-actions]

Backend Test Results

331 tests  +18   331 ✅ +18   9s ⏱️ +2s
  1 suites ± 0     0 💤 ± 0 
  1 files   ± 0     0 ❌ ± 0 

Results for commit 7a3a6c3. ± Comparison against base commit 4b88c3a.

♻️ This comment has been updated with latest results.

[calebyhan]

Everything in #96 is covered, and tests, alembic check, mypy and ruff all pass. One blocking issue in the migration's downgrade, noted inline.

[calebyhan]

LGTM