bug fixes

Author: kazmer97

lib/idp_common_pkg/idp_common/assessment/granular_service.py

@@ -22,6 +22,7 @@
 from idp_common import image, metrics, s3, utils
 from idp_common.assessment.models import AssessmentResult, AssessmentTask
 from idp_common.assessment.strands_executor import execute_assessment_tasks_parallel
+from idp_common.assessment.strands_service import _convert_field_path_to_string
 from idp_common.config.models import IDPConfig
 from idp_common.config.schema_constants import (
     SCHEMA_ITEMS,
@@ -305,7 +306,7 @@ def _traverse(
                                 task_id=f"task_{task_counter[0]}",
                                 task_type="attribute",
                                 field_path=item_path,
-                                field_name=prop_name,
+                                field_name=_convert_field_path_to_string(item_path),
                                 field_schema=items_schema,
                                 confidence_threshold=threshold,
                                 parent_assessment_dict=assessment_list,  # type: ignore
@@ -328,7 +329,7 @@ def _traverse(
                         task_id=f"task_{task_counter[0]}",
                         task_type="attribute",
                         field_path=field_path,
-                        field_name=prop_name,
+                        field_name=_convert_field_path_to_string(field_path),
                         field_schema=prop_schema,
                         confidence_threshold=threshold,
                         parent_assessment_dict=parent_dict,

lib/idp_common_pkg/idp_common/extraction/agentic_idp.py

@@ -226,7 +226,7 @@ def extraction_tool(
         # Note: The @tool decorator passes data as a dict, not as a model instance
         # We need to validate it manually using the Pydantic model
         extraction_model = model_class.model_validate(extraction)  # pyright: ignore[reportAssignmentType]
-        extraction_dict = extraction_model.model_dump()
+        extraction_dict = extraction_model.model_dump(mode="json")
         agent.state.set(key="current_extraction", value=extraction_dict)
         logger.debug(
             "Successfully stored extraction in state",
@@ -255,7 +255,8 @@ def apply_json_patches(
         patched_data = apply_patches_to_data(current_data, patches)
         validated_patched_data = model_class(**patched_data)
         agent.state.set(
-            key="current_extraction", value=validated_patched_data.model_dump()
+            key="current_extraction",
+            value=validated_patched_data.model_dump(mode="json"),
         )
 
         return {
@@ -269,9 +270,9 @@ def make_buffer_data_final_extraction(agent: Agent) -> str:
             agent.state.get("intermediate_extraction")
         )
 
-        agent.state.set("current_extraction", valid_extraction.model_dump())
+        agent.state.set("current_extraction", valid_extraction.model_dump(mode="json"))
 
-        return f"Successfully made the existing extraction the same as the buffer data {str(valid_extraction.model_dump())[100:]}..."
+        return f"Successfully made the existing extraction the same as the buffer data {str(valid_extraction.model_dump(mode='json'))[100:]}..."
 
     return extraction_tool, apply_json_patches, make_buffer_data_final_extraction
 
@@ -607,7 +608,7 @@ def _prepare_prompt_content(
     if existing_data:
         prompt_content.append(
             ContentBlock(
-                text=f"Please update the existing data using the extraction tool or patches. Existing data: {existing_data.model_dump()}"
+                text=f"Please update the existing data using the extraction tool or patches. Existing data: {existing_data.model_dump(mode='json')}"
             )
         )
 
@@ -870,7 +871,7 @@ async def structured_output_async(
         ),
     )
     if existing_data:
-        agent.state.set("current_extraction", existing_data.model_dump())
+        agent.state.set("current_extraction", existing_data.model_dump(mode="json"))
 
     response, result = await _invoke_agent_for_extraction(
         agent=agent,
@@ -902,7 +903,7 @@ async def structured_output_async(
                 ContentBlock(
                     text=f"""
                 You have successfully extracted the following data:
-                {json.dumps(result.model_dump(), indent=2)}
+                {json.dumps(result.model_dump(mode="json"), indent=2)}
 
                 Please take one final careful look at this extraction:
                 1. Check each field against the source document
@@ -933,7 +934,9 @@ async def structured_output_async(
             state={
                 "current_extraction": None,
                 "images": {},
-                "existing_data": existing_data.model_dump() if existing_data else None,
+                "existing_data": existing_data.model_dump(mode="json")
+                if existing_data
+                else None,
                 "extraction_schema_json": schema_json,  # Store for schema reminder tool
             },
             conversation_manager=SummarizingConversationManager(
@@ -951,7 +954,7 @@ async def structured_output_async(
 
         # Check if patches were applied during review
         updated_extraction = agent.state.get("current_extraction")
-        if updated_extraction != result.model_dump():
+        if updated_extraction != result.model_dump(mode="json"):
             # Patches were applied, validate the new extraction
             try:
                 result = data_format(**updated_extraction)