Add AI rules support for document processing and matching; enhance tax analysis with flag_for_review and auto_approve fields

app/services/document_processor.py

@@ -18,7 +18,11 @@ class DocumentProcessor:
         self.model = "meta-llama/llama-4-scout-17b-16e-instruct"  # Vision model
 
     async def process_file(
-        self, file_path: str, file_type: str, user_location: str = None
+        self,
+        file_path: str,
+        file_type: str,
+        user_location: str = None,
+        ai_rules: list = None,
     ) -> Dict[str, Any]:
         """Process uploaded file and extract receipt data
 
@@ -26,25 +30,27 @@ class DocumentProcessor:
             file_path: Path to the file to process
             file_type: Type of file (jpg, pdf, etc.)
             user_location: User's location string in format "State/Province, Country" (e.g., "Ontario, Canada")
+            ai_rules: List of AI rules for categorization (e.g., [{"condition": "vendor is Starbucks", "action": "Food"}])
         """
         try:
             if file_type.lower() in ["jpg", "jpeg", "png", "gif", "bmp"]:
-                return await self._process_image(file_path, user_location)
+                return await self._process_image(file_path, user_location, ai_rules)
             elif file_type.lower() == "pdf":
-                return await self._process_pdf(file_path, user_location)
+                return await self._process_pdf(file_path, user_location, ai_rules)
             else:
                 raise ValueError(f"Unsupported file type: {file_type}")
         except Exception as e:
             return {"error": str(e)}
 
     async def _process_image(
-        self, image_path: str, user_location: str = None
+        self, image_path: str, user_location: str = None, ai_rules: list = None
     ) -> Dict[str, Any]:
         """Extract data from image using Groq vision
 
         Args:
             image_path: Path to the image file
             user_location: User's location string in format "State/Province, Country" (e.g., "Ontario, Canada")
+            ai_rules: List of AI rules for categorization
         """
         try:
             # Encode image to base64
@@ -62,6 +68,16 @@ class DocumentProcessor:
             - Apply depreciation rules based on the user's location.
             """
 
+            # Build AI rules context for categorization
+            ai_rules_context = ""
+            if ai_rules and len(ai_rules) > 0:
+                ai_rules_context = "\n            CATEGORIZATION RULES (IMPORTANT - Apply these first):"
+                for idx, rule in enumerate(ai_rules, 1):
+                    condition = rule.get("condition", "")
+                    action = rule.get("action", "")
+                    ai_rules_context += f"\n            {idx}. If {condition} → set category to '{action}'"
+                ai_rules_context += "\n            - Apply these custom rules before using default categorization logic\n            - If multiple rules match, use the first matching rule\n            - If no rules match, use default categorization based on vendor type"
+
             # Create Groq vision prompt
             prompt = f"""
             Analyze this receipt image and extract the following information in JSON format:
@@ -89,9 +105,9 @@ class DocumentProcessor:
             - Total amount should be the final total including tax
             - Tax amount is separate tax line if available (if not clearly shown, calculate based on location)
             - Date should be the date on the receipt
-            - Categorize based on vendor type (Starbucks=Food, Shell=Transport, etc.)
             - Confidence score 0-1 based on how clear the receipt is
             - Currency should be the currency used on the receipt (e.g., "USD", "EUR", "CAD")
+            {ai_rules_context}
             {user_location_context}
             LOCATION & TAX RULES:
             - Extract location from receipt (look for store address, province/state, country)
@@ -166,18 +182,19 @@ class DocumentProcessor:
             return base64.b64encode(image_file.read()).decode("utf-8")
 
     async def _process_pdf(
-        self, pdf_path: str, user_location: str = None
+        self, pdf_path: str, user_location: str = None, ai_rules: list = None
     ) -> Dict[str, Any]:
         """Extract data from PDF by converting to image first
 
         Args:
             pdf_path: Path to the PDF file
             user_location: User's location string in format "State/Province, Country" (e.g., "Ontario, Canada")
+            ai_rules: List of AI rules for categorization
         """
         try:
             # For now, extract text from PDF and process as text
             text_content = self._extract_text_from_pdf(pdf_path)
-            return self._process_text_content(text_content, user_location)
+            return self._process_text_content(text_content, user_location, ai_rules)
 
         except Exception as e:
             return {"error": f"PDF processing error: {str(e)}"}
@@ -195,13 +212,14 @@ class DocumentProcessor:
             return ""
 
     def _process_text_content(
-        self, text_content: str, user_location: str = None
+        self, text_content: str, user_location: str = None, ai_rules: list = None
     ) -> Dict[str, Any]:
         """Process text content using Groq (fallback for PDFs)
 
         Args:
             text_content: Extracted text from PDF
             user_location: User's location string in format "State/Province, Country" (e.g., "Ontario, Canada")
+            ai_rules: List of AI rules for categorization
         """
         try:
             # Build user location context
@@ -216,6 +234,16 @@ class DocumentProcessor:
             - Apply depreciation rules based on the user's location.
             """
 
+            # Build AI rules context for categorization
+            ai_rules_context = ""
+            if ai_rules and len(ai_rules) > 0:
+                ai_rules_context = "\n            CATEGORIZATION RULES (IMPORTANT - Apply these first):"
+                for idx, rule in enumerate(ai_rules, 1):
+                    condition = rule.get("condition", "")
+                    action = rule.get("action", "")
+                    ai_rules_context += f"\n            {idx}. If {condition} → set category to '{action}'"
+                ai_rules_context += "\n            - Apply these custom rules before using default categorization logic\n            - If multiple rules match, use the first matching rule\n            - If no rules match, use default categorization based on vendor type"
+
             prompt = f"""
             Analyze this receipt text and extract the following information in JSON format:
             
@@ -247,9 +275,9 @@ class DocumentProcessor:
             - Total amount should be the final total including tax
             - Tax amount is separate tax line if available (if not clearly shown, calculate based on location)
             - Date should be the date on the receipt
-            - Categorize based on vendor type
             - Confidence score 0-1 based on clarity
             - Currency should be the currency used on the receipt (e.g., "USD", "EUR", "CAD")
+            {ai_rules_context}
             {user_location_context}
             LOCATION & TAX RULES:
             - Extract location from receipt (look for store address, province/state, country)