apply agent succesfully

app.py

@@ -220,7 +220,7 @@ async def query_3D_object(
     current_obj_path: str,
     embedding_dict: dict,
     top_k: int = 4,
-    method: Query3DObjectMethod = Query3DObjectMethod.SEMANTIC_SEARCH,
+    method: Query3DObjectMethod = Query3DObjectMethod.HYBRID_SEARCH,
 ) -> List:
     if query == "":
         raise gr.Error("Query cannot be empty!")
@@ -270,28 +270,31 @@ async def query_3D_object_by_hybrid_search_method(
     @function_tool
     def query_3D_object_by_keyword_search(query: str, match_code: str, top_k: int = 4):
         logger.info("Datum Agent is running query_3D_object_by_keyword_search")
-        logger.info(f"The 'match' function is:\n```{match_code}```")
-
-        def match(metadata: dict) -> bool:
-            """
-            This function should be generated by the match_code provided.
-            It will check if the metadata matches the query.
-            """
-            return True
+        logger.info(f"The 'match' function is:\n```\n{match_code}\n```")
 
+        # !!!IMPORTANT, create a new individual execution context for the match function
+        exec_globals = {}
         try:
-            exec(match_code, globals())
+            exec(match_code, exec_globals)
+            match = exec_globals[
+                "match"
+            ]  # get the match function from the execution context
             assert (
                 "def match(metadata: dict) -> bool:" in match_code
             ), "The match function is not defined correctly."
         except Exception:
-            raise gr.Error("Your query did not generate a valid match function.")
+            raise gr.Error(
+                "Your query did not generate a valid match function. Try your query again."
+            )
         matched_obj_paths = list(
             filter(
                 lambda obj_path: match(embedding_dict[obj_path]["metadata_dictionary"]),
                 embedding_dict,
             )
         )
+        logger.info(
+            f"Found {len(matched_obj_paths)} matching objects for the query `{query}`:\n```{matched_obj_paths}```"
+        )
 
         top_files = [x for x in matched_obj_paths[:top_k]]
         return top_files + [os.path.basename(x) for x in top_files]
@@ -393,14 +396,24 @@ Combine the `match` function with `query_3D_object_by_keyword_search` to filter
     @function_tool
     def query_3D_object_by_semantic_search(query: str, top_k: int = 4):
         logger.info("Datum Agent is running query_3D_object_by_semantic_search")
-        return query_3D_object_by_semantic_search_method(
+        response = query_3D_object_by_semantic_search_method(
             query, current_obj_path, embedding_dict, top_k
         )
+        logger.info(
+            f"Found {len(response) // 2} matching objects for the query `{query}`:\n```{response[: len(response) // 2]}```"
+        )
+        return response
 
     @function_tool
-    def search_3D_similarity_factory(selected_filepath: str, top_k: int = 4):
+    def search_3D_similarity_factory(
+        query: str, selected_filepath: str, top_k: int = 4
+    ):
         logger.info("Datum Agent is running search_3D_similarity_factory")
-        return search_3D_similarity(selected_filepath, embedding_dict, top_k)
+        response = search_3D_similarity(selected_filepath, embedding_dict, top_k)
+        logger.info(
+            f"Found {len(response) // 2} similar objects for the query `{query}`:\n```{response[: len(response) // 2]}```"
+        )
+        return response
 
     DATUM_AGENT_INSTRUCTIONS = """You are the Datum Agent: you retrieve the top-K most relevant 3D objects using three strategies.
 * Use `query_3D_object_by_semantic_search` for abstract or descriptive queries.
@@ -751,6 +764,13 @@ async def accumulate_and_embedding(input_files, file_list, embedding_dict):
     all_files = input_files
     new_files = input_files[len(file_list) :]
 
+    # # forwarding
+    # if os.environ.get("ENVIRONMENT") == "local" and os.path.exists("embedding_dict.pt"):
+    #     embedding_dict = torch.load(
+    #         "embedding_dict.pt", map_location=torch.device("cpu")
+    #     )  # load from local file
+    #     return all_files, gr.update(choices=all_files), embedding_dict
+
     # embedding
     for file_path in new_files:
         logger.info("Processing new upload file:", file_path)
@@ -786,6 +806,11 @@ async def accumulate_and_embedding(input_files, file_list, embedding_dict):
         embedding_dict[obj_path]["image_embedding"] = embeddings["image_embedding"]
         embedding_dict[obj_path]["text_embedding"] = text_embedding
 
+        # if os.environ.get("ENVIRONMENT") == "local":
+        #     # save to local file
+        #     torch.save(embedding_dict, "embedding_dict.pt")
+        #     logger.info("Saved embedding_dict to local file.")
+
     return all_files, gr.update(choices=all_files), embedding_dict