production: harden catalog tenants analytics and widget install

widget keys: arbitrary N-day TTL + refuse minting for unknown db

polish for client go-live: contact-request identity fix, leads viewer, ingest-fail alert, nvidia provider

fix: stabilize tenant routing and active db persistence

admin: owner login reliably sets the logged-in DB Active (awaited, not fire-and-forget)

admin: surface collection coverage in db-stats + panel badge

admin: remove Admin Password field from Operations & Contact card

admin: trim Operations & Contact to email + WhatsApp + Gmail app password

admin: Preview Bot + Set Active follow the logged-in/active DB

eval panel: fix custom-tests 500, judge-key UX, product-question generation

perf(admin): idle-sleep — stop background polling when the panel is inactive

fix(admin): round re-crawl interval label to 1 decimal

admin: per-DB client password = DB name + owner view follows active DB

catalog_query: gate engine to product DBs only (is_catalog_db)

feat: eval code_pointer — map failure evidence to owning module/function

fix: no idle polling from open tabs - chat 3s config loop removed (focus/visibility checks remain), admin health poll 60s->5min + visibility-gated (scale-to-zero billing)

fix(widget): use stateless signed keys to avoid restart 401 mismatches

feat(chat): add per-DB 10/min limit and tighten widget-key reveal flow

feat(widget): add expiring widget keys with TTL options and expiry enforcement

admin: display last crawl error and cooldown

feat: eval history, price validator, Gemini timeout fix

fix(ui): improve eval detail view and crawl log polling

feat(eval): judge_api_key field in admin Ops card

evals: add context recall and likely-cause hints

evals: improve live scoring and admin diagnostics

evals: show retrieved chunk previews

evals: add owner RAG scoreboard UI + API

security: fix P1 issues - XSS, dockerignore, sessionStorage, tenant isolation

admin: convert all write fetch() calls to authFetch() for CSRF auto-refresh

fix: stop crawl button UX — small link below log window

ui: live stop crawl feedback + 10s refresh interval

ui: replace ugly Stop button with subtle 'stop crawl' link

fix: show last completed crawl time (not start time) in admin

fix: cancel button, 15s Playwright timeout cap, is_crawling live refresh

fix: concurrent auto-crawl + live admin refresh + restore store DB

Add action buttons to audit FAIL/WARN results

Add fix hints to audit FAIL/WARN results + collapsible UI

Fix audit 301s, crawl settings wrong-DB bug, live refresh after save

Fix Edit dropdown clipping by using position:fixed with viewport coords

Fix audit suite: remove watchdog, fix auth headers, SSE parse, live API skip, dynamic descs

Fix Edit dropdown clipping for last DB — flips upward when near bottom of viewport

Add per-DB health status: colored pill in DB list + detail box in expandable panel

Fix: updateHealthStatus was immediately reversing active DB switch for owners

Fix changeDbPass endpoint and use alert() for rename/pass feedback

Fix admin panel: password cursor, Edit dropdown, rename-db endpoint

fix: saveBranding try-catch — show real error msg instead of silent red line

fix: auto re-sync active DB when server restarts and drifts from admin's session

fix: always switch active DB on admin login (not just client mode)

fix: show Gmail SMTP to clients (they set their own app password)

feat: client mode hides crawl, DB hub, and SMTP sections