Refine JSON response handling in batch analysis to exclude markdown code blocks and improve extraction logic

app/services/llm_tax_analyzer.py

@@ -817,7 +817,7 @@ Return your response as a SINGLE JSON object in this format:
                 messages=[
                     {
                         "role": "system",
-                        "content": "You are a Canadian tax expert. Analyze multiple transactions in batch and apply tax rules accurately. Always return valid JSON with all requested matches.",
+                        "content": "You are a Canadian tax expert. Analyze multiple transactions in batch and apply tax rules accurately. Return ONLY valid JSON - no markdown code blocks, no explanations, just pure JSON.",
                     },
                     {"role": "user", "content": prompt},
                 ],
@@ -844,17 +844,36 @@ Return your response as a SINGLE JSON object in this format:
             )
             logger.debug(f"Raw LLM response: {content[:500]}...")  # Log first 500 chars
 
-            # Parse the JSON response
+            # Parse the JSON response - handle various markdown code block formats
             json_str = content
+
+            # Check for markdown code blocks with various language identifiers
             if "```json" in content:
                 json_str = content.split("```json")[1].split("```")[0].strip()
+            elif "```javascript" in content:
+                json_str = content.split("```javascript")[1].split("```")[0].strip()
+            elif "```js" in content:
+                json_str = content.split("```js")[1].split("```")[0].strip()
             elif "```" in content:
-                json_str = content.split("```")[1].split("```")[0].strip()
+                # Generic code block - extract content between first ``` and last ```
+                parts = content.split("```")
+                if len(parts) >= 3:
+                    # Take the second part (index 1), which is between first and second ```
+                    json_str = parts[1].strip()
+                    # Remove language identifier if it's on the first line
+                    lines = json_str.split("\n", 1)
+                    if len(lines) > 1 and lines[0].strip() in [
+                        "json",
+                        "javascript",
+                        "js",
+                        "",
+                    ]:
+                        json_str = lines[1].strip()
 
             # Validate JSON string is not empty
             if not json_str:
                 logger.error("Extracted JSON string is empty")
-                logger.error(f"Original content was: {content}")
+                logger.error(f"Original content was: {content[:500]}")
                 return {}
 
             batch_analysis = json.loads(json_str)