perf: do not store nor serialize the unfiltered docsstore

Signed-off-by: thiswillbeyourgithub <[email protected]>

Author: thiswillbeyourgithub

wdoc/utils/filters.py

@@ -19,7 +19,7 @@
 def filter_docstore(
     loaded_embeddings: VectorStore,
     cli_kwargs: dict,
-) -> Tuple[VectorStore, bytes]:
+) -> VectorStore:
     if "filter_metadata" in cli_kwargs:
         filter_meta = create_metadata_filter(
             loaded_embeddings=loaded_embeddings,
@@ -59,14 +59,15 @@ def filter_cont(cont: str) -> bool:
         logger.warning("Your filter matched all stored documents!")
     assert good, "No documents in the vectorstore match the given filter"
 
-    # directly remove the filtered documents from the docstore
-    # but first store the docstore before altering it to allow
-    # unfiltering in the prompt
-    start_time = time.time()
-    unfiltered_docstore_bytes = loaded_embeddings.serialize_to_bytes()
-    serialize_time = time.time() - start_time
-    logger.debug(f"Serializing unfiltered docstore took {serialize_time:.3f} seconds")
+    # commented because it's taking quite long
+    # # first store the docstore before altering it to allow
+    # # unfiltering in the prompt
+    # start_time = time.time()
+    # unfiltered_docstore_bytes = loaded_embeddings.serialize_to_bytes()
+    # serialize_time = time.time() - start_time
+    # logger.debug(f"Serializing unfiltered docstore took {serialize_time:.3f} seconds")
 
+    # directly remove the filtered documents from the docstore
     start_time = time.time()
     status = loaded_embeddings.delete(ids_to_del)
     delete_time = time.time() - start_time
@@ -87,7 +88,7 @@ def filter_cont(cont: str) -> bool:
         loaded_embeddings.index_to_docstore_id
     ), "Something went wrong when deleting filtered out documents"
 
-    return loaded_embeddings, unfiltered_docstore_bytes
+    return loaded_embeddings
 
 
 def create_metadata_filter(

wdoc/wdoc.py

@@ -606,6 +606,9 @@ def print_exception(exc_type, exc_value, exc_traceback):
         else:
             self.loaded_docs = None  # will be loaded when embeddings are loaded
 
+        # flag to know if we already filtered or not
+        self._is_vectorstore_filtered = False
+
         if self.__import_mode__:
             logger.debug(
                 "Ready to query or summarize, call your_instance.query_task(your_question)"
@@ -791,19 +794,18 @@ def _query_or_search_task(self, query: str) -> dict:
         }
 
         # parse filters as callable for faiss filtering
-        if not hasattr(self, "unfiltered_docstore_bytes"):
+        if not self._is_vectorstore_filtered:
             if (
                 "filter_metadata" in self.cli_kwargs
                 or "filter_content" in self.cli_kwargs
             ):
                 from wdoc.utils.filters import filter_docstore
 
-                self.loaded_embeddings, self.unfiltered_docstore_bytes = (
-                    filter_docstore(
-                        loaded_embeddings=self.loaded_embeddings,
-                        cli_kwargs=self.cli_kwargs,
-                    )
+                self.loaded_embeddings = filter_docstore(
+                    loaded_embeddings=self.loaded_embeddings,
+                    cli_kwargs=self.cli_kwargs,
                 )
+                self._is_vectorstore_filtered = True
 
         assert query.strip(), "Cannot accept empty query"
         assert all(