Build a CRM System in Lovable

Add an activity timeline to the ContactDetail page and to a new DealDetail page.

On ContactDetail:
- At the top of the page, show the contact form (already built).
- Below it, add an ActivityTimeline component. It fetches activities WHERE contact_id = {contactId} ordered by occurred_at DESC.
- Above the timeline, add an activity composer: a type dropdown (note / call / email / meeting) + textarea for body + Submit button. On submit, INSERT into activities with actor_id = auth.uid(), contact_id = contactId, occurred_at = now().

On DealDetail (new page at /deals/:id):
- Show a summary card (title, stage badge, value, expected close date, contact chip linking to ContactDetail).
- Below it, same ActivityTimeline + composer pattern but filtered by deal_id.

Group timeline entries by calendar day using a date section header (e.g., 'Today', 'Yesterday', 'Jun 23'). Each entry shows: type icon (Phone, Mail, MessageSquare, Calendar from lucide-react), actor full name (from profiles), body text, and relative time ('2 hours ago').

Route: add /deals/:id to App.tsx.

Add outbound email capability so reps can send emails directly from the CRM and have them logged in the activity timeline.

1. Add RESEND_API_KEY to Cloud tab → Secrets (I've already done this).

2. Create Edge Function at supabase/functions/send-email-to-contact/index.ts. The function should:
   - Accept POST with JSON body: { contact_id, subject, body_html, sender_name }
   - Verify the caller is authenticated (check Authorization header via supabase.auth.getUser)
   - POST to https://api.resend.com/emails with the RESEND_API_KEY from Deno.env.get('RESEND_API_KEY')
   - On success, INSERT one row into activities: { actor_id: caller uid, contact_id, type: 'email', body: `Sent: ${subject}\n\n${body_text_version}` }
   - Return { success: true } or a 4xx error with message

3. On ContactDetail, add a 'Send email' button that opens a Dialog with: Subject input, Body textarea, Send button. On send, call supabase.functions.invoke('send-email-to-contact', { body: {...} }) and refresh the ActivityTimeline after success.

Enable Google OAuth sign-in alongside the existing email+password flow.

1. In Google Cloud Console, create an OAuth 2.0 Client ID (type: Web application). Add these authorized redirect URIs:
   - https://[your-lovable-preview-subdomain].lovable.app/auth/v1/callback
   - https://[your-custom-domain.com]/auth/v1/callback

2. In Cloud tab → Users & Auth → Social providers → enable Google. Paste the client ID and client secret.

3. In Cloud tab → Users & Auth → URL Configuration: set Site URL to your production domain. Add both the preview domain and the production domain to Redirect URLs.

4. On the Login page, add a 'Continue with Google' Button that calls `supabase.auth.signInWithOAuth({ provider: 'google', options: { redirectTo: window.location.origin + '/dashboard' } })`. Place it above the email/password form with an 'or' divider.

5. Confirm AuthGuard handles the OAuth callback correctly — after redirect, supabase.auth.getSession() should return the user automatically.

Add a pipeline forecast widget to the /dashboard page.

1. Create a Postgres RPC function `pipeline_forecast()` that returns:
   SELECT stage, COUNT(*) as deal_count, SUM(value_cents) as total_cents
   FROM deals
   WHERE owner_id = auth.uid() AND stage NOT IN ('won','lost')
   GROUP BY stage

2. Apply weighted win-rates: lead=10%, qualified=30%, proposal=60%, won=100%. Compute weighted_value_cents = total_cents * win_rate per stage.

3. On the Dashboard, call this RPC via supabase.rpc('pipeline_forecast'). Render the result as a stacked horizontal bar chart using recharts. Show each stage as a colored segment matching the Kanban column colors (gray/blue/amber). Include a legend and a 'Weighted value' total below the chart.

4. Refresh the chart every 60 seconds via a setInterval inside a useEffect. Show a skeleton placeholder while loading.

Convert the CRM from per-user ownership to multi-tenant org-scoped ownership so multiple reps can share the same pipeline.

1. Create two new tables:
   - `orgs` — id (uuid pk), name (text not null), created_at (timestamptz)
   - `members` — id (uuid pk), org_id (uuid references orgs(id)), user_id (uuid references auth.users(id)), role (text default 'member' check in ('owner','member')), joined_at (timestamptz default now()). UNIQUE(org_id, user_id).

2. Add org_id (uuid references orgs(id)) to contacts, companies, deals, activities. Backfill with a placeholder org for the current user.

3. Create a SECURITY DEFINER plpgsql function `is_org_member(p_org_id uuid)` that returns true when auth.uid() appears in members for that org. Use LANGUAGE plpgsql (not SQL) to prevent inlining.

4. Rewrite RLS on contacts, companies, deals, activities to use: `is_org_member(org_id)` for SELECT and `is_org_member(org_id) AND EXISTS(SELECT 1 FROM members WHERE org_id = t.org_id AND user_id = auth.uid())` for INSERT/UPDATE/DELETE.

5. Add an org switcher dropdown to the AppLayout sidebar — reads orgs the user belongs to from members join, writes active_org_id to React context. All queries use the active org_id from context.

Add AI-powered lead scoring to the /contacts page.

1. Create Edge Function `score-contact` at supabase/functions/score-contact/index.ts. It should:
   - Accept POST with { contact_id }
   - Fetch the contact row and their last 5 activities from Lovable Cloud
   - Call the Lovable AI panel (Gemini 3 Flash — the default model in Cloud tab → AI) with a prompt: 'Given this contact data and recent activities, rate the lead quality from 1 to 10 and give a 1-sentence reason. Return JSON: {score: number, reason: string}'
   - Parse the JSON response
   - UPDATE contacts SET ai_score = score, ai_score_reason = reason, ai_scored_at = now() WHERE id = contact_id
   - Return { score, reason }

2. Add two columns to contacts: ai_score (int), ai_score_reason (text), ai_scored_at (timestamptz).

3. On the Contacts list table, add an 'AI Score' column showing a colored badge (1-4 red, 5-7 amber, 8-10 green) or a 'Score' button if ai_score is null. Clicking Score calls the Edge Function for that contact.

4. Add a 'Score all unscored' button at the top of the Contacts page that iterates through contacts where ai_score IS NULL and calls score-contact for each, with a 1-second delay between calls to avoid rate limits.