Cara feedback loop - feedback + digest¶
| Stage | Building |
|---|---|
| Status | Partial |
| Design status | In review |
| Estimate | S |
| Confidence | Medium |
| Linear | PIN-291 ↗ |
| Linear status | Backlog |
| Design | Design ↗ |
| Linear epic | Cara AI & Automation |
| Module | Cara ↗ |
Priority: High · Route: /ai-hub/feedback
Scope¶
Cara-mediated feedback capture replacing the existing feedback text box: users submit feedback, feature requests, questions, or bug reports; Cara classifies each submission and stores it per-user with status. Conor gets a monthly digest of feedback grouped by classification and frequency, with an optional auto-create-Linear-issue path for high-signal items. (Source: PIN-291.)
Acceptance criteria¶
- Cara-mediated capture replaces the existing feedback text box (floating button / login screen)
- User-submitted feedback is auto-classified (feedback / feature request / question / bug)
- Per-user history shows the user's submitted items and their status
- Conor-only monthly digest: feedback grouped by classification and frequency for the prior month
- High-signal feedback (feature request, multiple users) can auto-create a Linear issue
Conor's comments¶
Is this backend work record system inputs and decisions? Don't want user access to it. How will we use this data to train or fine tune outputs?
Notes¶
Shipped leverage; auto-classify via Haiku
Open question for Conor¶
Verify Cara Feedback + digest with Marijana