Merge pull request #90 from seungwon2/master

[Fix] : fix no parent doc error

Author: dongjin-ml

genai/aws-gen-ai-kr/utils/rag_summit.py

@@ -614,56 +614,57 @@ def _get_hyde_response(query, prompt, llm_text):
     # ParentDocument based
     def get_parent_document_similar_docs(cls, **kwargs):
 
-        child_search_results = kwargs["similar_docs"]
-
-        parent_info, similar_docs = {}, []
-        for rank, (doc, score) in enumerate(child_search_results):
-            parent_id = doc.metadata["parent_id"]
-            if parent_id != "NA": ## For Tables and Images
-                if parent_id not in parent_info:
-                    parent_info[parent_id] = (rank+1, score)
-            else:
-                if kwargs["hybrid"]:
-                    similar_docs.append((doc, score))
+            child_search_results = kwargs["similar_docs"]
+            
+            parent_info, similar_docs = {}, []
+            for rank, (doc, score) in enumerate(child_search_results):
+                parent_id = doc.metadata["parent_id"]
+                if parent_id != "NA": ## For Tables and Images
+                    if parent_id not in parent_info:
+                        parent_info[parent_id] = (rank+1, score)
                 else:
-                    similar_docs.append((doc))
-                    
-        parent_ids = sorted(parent_info.items(), key=lambda x: x[1], reverse=False)
-        parent_ids = list(map(lambda x:x[0], parent_ids))
-
-        parent_docs = opensearch_utils.get_documents_by_ids(
-            os_client=kwargs["os_client"],
-            ids=parent_ids,
-            index_name=kwargs["index_name"],
-        )
-
-        if parent_docs["docs"]:
-            for res in parent_docs["docs"]:
-                doc_id = res["_id"]
-                doc = Document(
-                    page_content=res["_source"]["text"],
-                    metadata=res["_source"]["metadata"]
+                    if kwargs["hybrid"]:
+                        similar_docs.append((doc, score))
+                    else:
+                        similar_docs.append((doc))
+            
+            parent_ids = sorted(parent_info.items(), key=lambda x: x[1], reverse=False)
+            parent_ids = list(map(lambda x:x[0], parent_ids))
+            
+            if parent_ids:
+                parent_docs = opensearch_utils.get_documents_by_ids(
+                    os_client=kwargs["os_client"],
+                    ids=parent_ids,
+                    index_name=kwargs["index_name"],
                 )
-                if kwargs["hybrid"]:
-                    similar_docs.append((doc, parent_info[doc_id][1]))
-                else:
-                    similar_docs.append((doc))
-        
-        if kwargs["hybrid"]:
-            similar_docs = sorted(
-                similar_docs,
-                key=lambda x: x[1],
-                reverse=True
-            )
-        
-        if kwargs["verbose"]:
-            print("===== ParentDocument =====")
-            print (f'filter: {kwargs["boolean_filter"]}')
-            print (f'# child_docs: {len(child_search_results)}')
-            print (f'# parent docs: {len(similar_docs)}')
-            print (f'# duplicates: {len(child_search_results)-len(similar_docs)}')
 
-        return similar_docs
+                if parent_docs["docs"]:
+                    for res in parent_docs["docs"]:
+                        doc_id = res["_id"]
+                        doc = Document(
+                            page_content=res["_source"]["text"],
+                            metadata=res["_source"]["metadata"]
+                        )
+                        if kwargs["hybrid"]:
+                            similar_docs.append((doc, parent_info[doc_id][1]))
+                        else:
+                            similar_docs.append((doc))
+
+            if kwargs["hybrid"]:
+                similar_docs = sorted(
+                    similar_docs,
+                    key=lambda x: x[1],
+                    reverse=True
+                )
+            
+            if kwargs["verbose"]:
+                print("===== ParentDocument =====")
+                print (f'filter: {kwargs["boolean_filter"]}')
+                print (f'# child_docs: {len(child_search_results)}')
+                print (f'# parent docs: {len(similar_docs)}')
+                print (f'# duplicates: {len(child_search_results)-len(similar_docs)}')
+
+            return similar_docs
 
     @classmethod
     def get_rerank_docs(cls, **kwargs):