Update kig_core/planner.py

kig_core/planner.py

@@ -156,45 +156,45 @@ def generate_structured_issues(state: PlannerState) -> Dict[str, Any]:
 
     chain = prompt | issue_llm | output_parser
 
-try:
-    structured_issues_obj = invoke_llm(chain, {
-        "user_query": user_query,
-        "context": full_context
-    })
-    print(f"structured_issues_obj => type : {type(structured_issues_obj)}, value : {structured_issues_obj}")
-
-    # If the output is a dict with a key 'key_issues', extract it
-    if isinstance(structured_issues_obj, dict) and 'key_issues' in structured_issues_obj:
-        issues_data = structured_issues_obj['key_issues']
-    else:
-        issues_data = structured_issues_obj  # Assume it's already a list of dicts
-
-    # Always convert to KeyIssueInvoke objects
-    key_issues_list = [KeyIssueInvoke(**issue_dict) for issue_dict in issues_data]
-
-    # Ensure IDs are sequential if the LLM didn't assign them correctly
-    for i, issue in enumerate(key_issues_list):
-        issue.id = i + 1
-
-    logger.info(f"Successfully generated {len(key_issues_list)} structured key issues.")
-    final_message = f"Generated {len(key_issues_list)} Key Issues based on the query '{user_query}'."
-    return {
-        "key_issues": key_issues_list,
-        "messages": [AIMessage(content=final_message)],
-        "error": None
-    }
-except Exception as e:
-    logger.error(f"Failed to generate or parse structured key issues: {e}", exc_info=True)
-    # Attempt to get raw output for debugging if possible
-    raw_output = "Could not retrieve raw output."
     try:
-         raw_chain = prompt | issue_llm | StrOutputParser()
-         raw_output = invoke_llm(raw_chain, {"user_query": user_query, "context": full_context})
-         logger.debug(f"Raw output from failed JSON parsing:\n{raw_output}")
-    except Exception as raw_e:
-         logger.error(f"Could not even get raw output: {raw_e}")
-
-    return {"error": f"Failed to generate structured key issues: {e}. Raw output hint: {raw_output[:500]}..."}
+        structured_issues_obj = invoke_llm(chain, {
+            "user_query": user_query,
+            "context": full_context
+        })
+        print(f"structured_issues_obj => type : {type(structured_issues_obj)}, value : {structured_issues_obj}")
+    
+        # If the output is a dict with a key 'key_issues', extract it
+        if isinstance(structured_issues_obj, dict) and 'key_issues' in structured_issues_obj:
+            issues_data = structured_issues_obj['key_issues']
+        else:
+            issues_data = structured_issues_obj  # Assume it's already a list of dicts
+    
+        # Always convert to KeyIssueInvoke objects
+        key_issues_list = [KeyIssueInvoke(**issue_dict) for issue_dict in issues_data]
+    
+        # Ensure IDs are sequential if the LLM didn't assign them correctly
+        for i, issue in enumerate(key_issues_list):
+            issue.id = i + 1
+    
+        logger.info(f"Successfully generated {len(key_issues_list)} structured key issues.")
+        final_message = f"Generated {len(key_issues_list)} Key Issues based on the query '{user_query}'."
+        return {
+            "key_issues": key_issues_list,
+            "messages": [AIMessage(content=final_message)],
+            "error": None
+        }
+    except Exception as e:
+        logger.error(f"Failed to generate or parse structured key issues: {e}", exc_info=True)
+        # Attempt to get raw output for debugging if possible
+        raw_output = "Could not retrieve raw output."
+        try:
+             raw_chain = prompt | issue_llm | StrOutputParser()
+             raw_output = invoke_llm(raw_chain, {"user_query": user_query, "context": full_context})
+             logger.debug(f"Raw output from failed JSON parsing:\n{raw_output}")
+        except Exception as raw_e:
+             logger.error(f"Could not even get raw output: {raw_e}")
+    
+        return {"error": f"Failed to generate structured key issues: {e}. Raw output hint: {raw_output[:500]}..."}
 
 
 # --- Conditional Edges ---