Add AI parsing capabilities for debt information and update database schema

- Implemented AI-powered parsing for debt emails using Google's Gemini model.
- Enhanced Postmark webhook to extract debt details including amount, vendor, description, due date, and legitimacy.
- Updated database schema to include new columns for AI-extracted data.
- Added environment variable requirements and updated package dependencies.
- Created migration script for new database columns and indexes.

supabase/migrations/20250607000500_add_ai_parsing_columns.sql

@@ -0,0 +1,16 @@
+-- Add new columns for AI-enhanced debt parsing
+-- Migration for improved debt information storage
+
+-- Add new columns to debts table
+ALTER TABLE debts ADD COLUMN IF NOT EXISTS description text;
+ALTER TABLE debts ADD COLUMN IF NOT EXISTS due_date timestamptz;
+ALTER TABLE debts ADD COLUMN IF NOT EXISTS metadata jsonb DEFAULT '{}'::jsonb;
+
+-- Create indexes for new columns
+CREATE INDEX IF NOT EXISTS idx_debts_due_date ON debts(due_date);
+CREATE INDEX IF NOT EXISTS idx_debts_metadata ON debts USING gin(metadata);
+
+-- Add comment for documentation
+COMMENT ON COLUMN debts.description IS 'AI-extracted description of what the debt is for';
+COMMENT ON COLUMN debts.due_date IS 'Due date extracted from the email, if mentioned';
+COMMENT ON COLUMN debts.metadata IS 'Additional metadata including isDebtCollection flag and other AI-extracted information';