fixed issue

app/services/validation.py

@@ -328,10 +328,24 @@ def is_plausible_phone(raw: str) -> bool:
     phone number at all: a character that isn't a digit or common phone punctuation, or too few digits.
     Empty is plausible (a lead is allowed to have no phone at all) -- only a GIVEN garbage value should
     block a save; see CreateLeadInput/UpdateLeadInput's own field_validators.
+
+    BUGFIX (2026-10-29): contact_phone has ALWAYS been allowed to carry several comma-joined numbers for
+    one lead -- see call_timezone()'s own `.split(",")[0]`, crud/contacts.py, lead_engine.py's
+    _all_phones(), and the frontend's PhoneNumberList, all of which already expect this. This function
+    didn't account for that at all (no comma in its character whitelist), so any lead with more than one
+    number on file -- a legitimate, pre-existing, everyday case -- failed validation and blocked Save the
+    moment entry (6)'s blocking field_validator shipped. Now checks each comma-separated piece on its own.
     """
     value = (raw or "").strip()
     if not value:
         return True
+    pieces = [p.strip() for p in value.split(",") if p.strip()]
+    if not pieces:
+        return False
+    return all(_is_plausible_single_phone(p) for p in pieces)
+
+
+def _is_plausible_single_phone(value: str) -> bool:
     if any(ch not in "0123456789+()-. " for ch in value):
         return False
     return sum(ch.isdigit() for ch in value) >= 7

docs.md

@@ -3,6 +3,89 @@
 Running record of features shipped, bugs fixed, and how each was verified.
 Newest entry first. See `CLAUDE.md` for the logging rule.
 
+## 2026-10-29 (2)
+
+### Bug fix: meeting booked from Callbacks/Pipeline/Meetings didn't count in Today's "Meetings" stat
+
+User: "agar callbacks tab sy ya kahin sy meeting booked hoti hai woh meetings me count nhi hoti today me".
+Root cause: Today/This-Week's "Meetings" stat (`app/crud/call_events.py`'s `_count_outcomes(... (
+"meeting_booked",))`) only ever counts rows in the `call_events` table — written by
+`apply_call_outcome()`/`book_followup_meeting()`. Cold Call Desk's battlecard correctly books a meeting
+through `useCallOutcome()` (→ `apply_call_outcome`, which writes a call_events row), so those always
+counted. But two OTHER "book a meeting" buttons — found by grepping every frontend call site that sets
+`pipeline_stage: "meeting_booked"` — used a plain `useUpdateLead()` PATCH instead: a raw field write that
+never touches `call_events` at all. The lead still correctly moved to the Meetings calendar (that only
+needs `meeting_at`), so this was invisible unless you were specifically watching the Today counter.
+
+**Files touched** (both frontend-only — the backend's `/leads/{id}/outcome` endpoint already existed and
+already logs correctly, this was purely a wrong-hook wiring bug):
+- `outreach-frontend/components/pipeline/pipeline-view.tsx` — `PipelineActions`'s "Meeting Booked" button
+  (shared by `/pipeline` AND `/callbacks`, since Callbacks reuses this same component scoped differently)
+  now uses `useCallOutcome()` instead of `useUpdateLead()` for a first-time booking (the
+  already-correct `useBookFollowupMeeting()` path for `proposal_sent`-stage leads is untouched).
+- `outreach-frontend/components/meetings/needs-followup-list.tsx` — the Meetings page's "Mark as Meeting
+  Booked" button (on a lead sitting at Needs Follow-up), same swap.
+
+Both now go through the exact same `apply_call_outcome(outcome="meeting_booked")` path Cold Call Desk
+already used, which does everything the old raw PATCH did (same `pipeline_stage`/`meeting_at`) PLUS logs
+to `call_events` (counts in Today), appends a note, and clears `sent_to_desk_at` if the lead was still on
+a desk. Checked for other instances of the same bug pattern — two other `meeting_at:`-setting call sites
+remain (`add-lead-popover.tsx`, and the general multi-field save in `lead-edit-form.tsx`) but both are
+bulk/initial data entry, not a "disposition" action, so left as-is; only dedicated "book a meeting" action
+buttons were in scope.
+
+**How verified**: `npx tsc --noEmit` (0 errors) and `npx eslint` on both files (0 warnings/errors). Both
+mirrored to `E:\outreach-agent-frontend`, confirmed zero drift via `diff -q`.
+
+**Status**: ⚠️ Partial — code complete and statically verified, NOT yet live-tested. No backend restart
+needed (frontend-only change, no backend file touched) — picked up on next page load. Live test: book a
+meeting from the Callbacks tab (or Meetings page's "Mark as Meeting Booked"), then check Today's Meetings
+count went up by one.
+
+## 2026-10-29 (1)
+
+### Bug fix: phone validation (entry 2026-10-28 (6)) broke the existing multi-number format
+
+User screenshot: Contact Phone field with "+447786363281 ,+44 371 964 2113" (two numbers, the CRM's
+existing comma-joined-multiple-numbers format — see `components/leads/phone-number-list.tsx`,
+`lead_engine.py`'s `_all_phones()`, `call_timezone()`'s own `.split(",")[0]`) rejected as "Doesn't look
+like a valid phone number", Save stuck disabled. Root cause: entry (6)'s `is_plausible_phone()` (both
+backend and frontend) only ever validated a SINGLE number's characters — comma wasn't in the whitelist —
+so any lead that legitimately carries more than one number broke the moment that blocking validator
+shipped. This was a regression in a prior fix, not a pre-existing bug.
+
+Fix: `is_plausible_phone()` now splits the input on comma and validates each piece independently (all
+must pass), in both `outreach-backend/app/services/validation.py` and
+`outreach-frontend/lib/validation/lead.ts`.
+
+Also implemented the user's second ask in the same message: "agar koi add krna bhul jaye 2 3 numbers me
+sy ek k bech me toh system wahan auto add kr dy" — if someone types/pastes two numbers back-to-back with
+no separator at all, auto-insert one. New `autoSeparatePhoneNumbers()` in `lib/validation/lead.ts`,
+scoped to the one unambiguous boundary this data actually has (every real number here starts with
+"+<country code>"): a "+" immediately following another number's digits (no comma, only whitespace or
+nothing between) gets a ", " inserted before it. Wired to `onBlur` on the Contact Phone field in both
+`components/layout/add-lead-popover.tsx` and `components/leads/lead-edit-form.tsx` — reformats as soon as
+the rep leaves the field, before `phoneInvalid` is ever computed against it. An already-comma-separated
+input is left untouched (confirmed below).
+
+**Files touched**: `outreach-backend/app/services/validation.py`,
+`outreach-frontend/lib/validation/lead.ts`, `outreach-frontend/components/layout/add-lead-popover.tsx`,
+`outreach-frontend/components/leads/lead-edit-form.tsx`.
+
+**How verified**: `python3 -m py_compile` (backend, clean) plus a standalone interpreter check of the
+split-validation logic against the exact reported value and several others (all correct, including
+`""` → true and `"qw"` → false unchanged); `npx tsc --noEmit` (0 errors) and `npx eslint` (exit 0) on
+all touched frontend files; a standalone Node check of both `isPlausiblePhone()` and
+`autoSeparatePhoneNumbers()` together against 5 cases, including two numbers pasted with NO separator at
+all and the user's own reported value verbatim — all passed. All 4 files mirrored to `E:\mz_chatbot` /
+`E:\outreach-agent-frontend`, confirmed zero drift via `diff -q`.
+
+**Status**: ⚠️ Partial — code complete and verified statically + via standalone logic tests, NOT yet
+live-tested in the running app. Needs a backend restart (`validation.py` changed); frontend picked up on
+next page load. Live test: paste the user's exact reported value into Contact Phone on both Add Lead and
+Edit Lead, confirm Save is enabled; also type two numbers with no separator and tab out, confirm a comma
+gets inserted automatically.
+
 ## 2026-10-28 (9)
 
 ### Lint cleanup: unused `MailtoButton` import