feat: OCR via tesseract, dedup, category selection for expense receipts

- Dockerfile: install tesseract-ocr so Pillow+pytesseract can OCR receipt images
- operational_store: JSON-serialize raw_data before passing to asyncpg JSONB
- receipt_parser: add SHA256 hash + date extracted from filename timestamps
- expenses_agent: deduplicate receipts by hash before creating expense records
- expenses_agent: fetch all expensable Odoo products, pass list to LLM for
  category selection (Meals, Flights, etc.) per receipt
- expenses_agent: pass date_hint from filename (e.g. 20260509_180857.jpg -> 2026-05-09)
  as fallback when OCR text is unavailable
- expenses_tools: add get_expense_products() to fetch all expensable products

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

agent_service/agents/expenses_agent.py

@@ -115,11 +115,39 @@ class ExpensesAgent(BaseAgent):
         sheet_id = sheet_result.record_id
         actions = [f'Created expense sheet "{sheet_name}" (ID {sheet_id})']
 
-        product_id = await self._et.get_default_expense_product()
+        # Fetch all expensable products once for category selection
+        expense_products = await self._et.get_expense_products()
+        default_product_id = expense_products[0]['id'] if expense_products else None
+        product_map = {p['id']: p['name'] for p in expense_products}
 
-        for receipt in receipts:
+        # Deduplicate receipts by SHA256 hash — same image uploaded twice
+        seen_hashes: set = set()
+        unique_receipts = []
+        for r in receipts:
+            h = r.get('sha256')
+            if h:
+                if h in seen_hashes:
+                    logger.info('expenses_agent: skipping duplicate receipt %s', r.get('filename'))
+                    actions.append(f"Skipped duplicate: {r.get('filename', 'receipt')}")
+                    continue
+                seen_hashes.add(h)
+            unique_receipts.append(r)
+
+        for receipt in unique_receipts:
             parsed = await self._parse_receipt_text(
-                receipt.get('text', ''), receipt.get('filename', 'receipt'))
+                receipt.get('text', ''), receipt.get('filename', 'receipt'),
+                expense_products=expense_products,
+                date_hint=receipt.get('date_from_name'),
+            )
+            # Pick product by name match returned from LLM, fall back to default
+            product_id = default_product_id
+            chosen_name = parsed.get('product_name', '')
+            if chosen_name:
+                for p in expense_products:
+                    if p['name'].lower() == chosen_name.lower():
+                        product_id = p['id']
+                        break
+
             expense_result = await self._et.create_expense(
                 sheet_id=sheet_id,
                 employee_id=employee_id,
@@ -127,13 +155,13 @@ class ExpensesAgent(BaseAgent):
                 total_amount=float(parsed.get('amount', 0.0)),
                 date=str(parsed.get('date') or _date.today().isoformat()),
                 product_id=product_id,
-                description=str(parsed.get('description', '')),
             )
             if expense_result.success:
+                cat = product_map.get(product_id, 'Expense')
                 actions.append(
                     f"Added: {parsed.get('vendor', 'Unknown vendor')} "
                     f"${float(parsed.get('amount', 0)):.2f} "
-                    f"on {parsed.get('date', 'today')}"
+                    f"({cat}) on {parsed.get('date', 'today')}"
                 )
                 if receipt.get('b64'):
                     await self._et.attach_receipt(
@@ -151,20 +179,39 @@ class ExpensesAgent(BaseAgent):
         self._actions_taken = actions
         return actions
 
-    async def _parse_receipt_text(self, text: str, filename: str) -> dict:
+    async def _parse_receipt_text(self, text: str, filename: str,
+                                   expense_products: list = None,
+                                   date_hint: str = None) -> dict:
+        today = _date.today().isoformat()
         fallback = {'vendor': filename, 'amount': 0.0,
-                    'date': _date.today().isoformat(), 'description': filename}
-        if not text or text.startswith('['):
-            return fallback
+                    'date': date_hint or today, 'product_name': ''}
+        ocr_failed = not text or text.startswith('[')
 
-        prompt = (
-            'Extract expense details from the following receipt text. '
-            'Return ONLY valid JSON with these keys: '
-            '"vendor" (string), "amount" (number, the total charged), '
-            '"date" (string YYYY-MM-DD, use today if absent), '
-            '"description" (string, brief expense type).\n\n'
-            f'Receipt text (first 2000 chars):\n{text[:2000]}\n\nJSON only:'
-        )
+        product_list = ''
+        if expense_products:
+            names = [p['name'] for p in expense_products]
+            product_list = ', '.join(f'"{n}"' for n in names)
+
+        if ocr_failed:
+            # No OCR text — still try to classify category from filename/date
+            if not product_list:
+                return fallback
+            prompt = (
+                f'A receipt photo named "{filename}" could not be read by OCR. '
+                f'Based only on the filename, pick the most likely expense category '
+                f'from this list: [{product_list}]. '
+                f'Return ONLY valid JSON: {{"product_name": "..."}}'
+            )
+        else:
+            prompt = (
+                'Extract expense details from the following receipt text. '
+                'Return ONLY valid JSON with these keys:\n'
+                '"vendor" (string, merchant name),\n'
+                '"amount" (number, the total amount charged — look for "Total", "Amount Due", "Grand Total"),\n'
+                f'"date" (string YYYY-MM-DD, use {date_hint or today} if not found),\n'
+                f'"product_name" (string, pick the best match from [{product_list}] or empty string).\n\n'
+                f'Receipt text (first 2000 chars):\n{text[:2000]}\n\nJSON only:'
+            )
         try:
             resp = await self._llm.submit(
                 [{'role': 'user', 'content': prompt}],
@@ -177,8 +224,8 @@ class ExpensesAgent(BaseAgent):
                 return {
                     'vendor': str(data.get('vendor', filename)),
                     'amount': float(data.get('amount', 0.0)),
-                    'date': str(data.get('date', _date.today().isoformat())),
-                    'description': str(data.get('description', '')),
+                    'date': str(data.get('date') or date_hint or today),
+                    'product_name': str(data.get('product_name', '')),
                 }
         except Exception as exc:
             logger.warning('Receipt parse failed for %s: %s', filename, exc)