admin: feature store browser — features sent to ml/serving per scoring call #64

New Issue

alvis · 2026-04-16T03:29:31Z

alvis commented

2026-04-16 03:29:31 +00:00

Per-user view of the feature vector logged at each /score call: hour_sin, hour_cos, is_overdue, task_age_days, priority. Show diff across time for a user to spot drift.\n\nRequires ml/serving to log feature vectors (currently scores but doesn't persist them). Add a /features/{user_id} endpoint to ml/serving, store last N vectors on disk alongside model state.\n\nPart of the M1 admin console build sequence (step 7 of 16).

Per-user view of the feature vector logged at each `/score` call: `hour_sin`, `hour_cos`, `is_overdue`, `task_age_days`, `priority`. Show diff across time for a user to spot drift.\n\nRequires `ml/serving` to log feature vectors (currently scores but doesn't persist them). Add a `/features/{user_id}` endpoint to ml/serving, store last N vectors on disk alongside model state.\n\nPart of the M1 admin console build sequence (step 7 of 16).

alvis added this to the M1 — Real signal milestone 2026-04-16 03:29:31 +00:00

alvis added the admin label 2026-04-16 03:29:43 +00:00

alvis closed this issue

2026-04-16 03:56:23 +00:00

alvis referenced this issue from a commit

2026-04-24 15:10:18 +00:00

feat: M1 admin console — all 10 remaining pages + signal/quality/ops infrastructure

Sign in to join this conversation.