diff --git a/daras_ai_v2/vector_search.py b/daras_ai_v2/vector_search.py
index 302e5cda9..038bb269f 100644
--- a/daras_ai_v2/vector_search.py
+++ b/daras_ai_v2/vector_search.py
@@ -84,7 +84,7 @@ class DocSearchRequest(BaseModel):
     scroll_jump: int | None
 
     doc_extract_url: str | None
-
+    cache_knowledge: typing.Optional[bool] = True
     embedding_model: typing.Literal[tuple(e.name for e in EmbeddingModels)] | None
     dense_weight: float | None = Field(
         ge=0.0,
@@ -146,9 +146,14 @@ def get_top_k_references(
     else:
         selected_asr_model = google_translate_target = None
 
-    file_url_metas = flatmap_parallel(doc_or_yt_url_to_metadatas, input_docs)
+    cache_knowledge = request.cache_knowledge
+    file_url_metas = flatmap_parallel(
+        lambda f_url: doc_or_yt_url_to_metadatas(f_url, cache_knowledge), input_docs
+    )
     file_urls, file_metas = zip(*file_url_metas)
 
+    logger.debug(f"file_urls: {file_urls}")
+
     yield "Creating knowledge embeddings..."
 
     embedding_model = EmbeddingModels.get(
@@ -269,7 +274,9 @@ def get_vespa_app():
     return Vespa(url=settings.VESPA_URL)
 
 
-def doc_or_yt_url_to_metadatas(f_url: str) -> list[tuple[str, FileMetadata]]:
+def doc_or_yt_url_to_metadatas(
+        f_url: str, cache_knowledge: bool = True
+) -> list[tuple[str, FileMetadata]]:
     if is_yt_dlp_able_url(f_url):
         entries = yt_dlp_get_video_entries(f_url)
         return [
@@ -287,10 +294,10 @@ def doc_or_yt_url_to_metadatas(f_url: str) -> list[tuple[str, FileMetadata]]:
             for entry in entries
         ]
     else:
-        return [(f_url, doc_url_to_file_metadata(f_url))]
+        return [(f_url, doc_url_to_file_metadata(f_url, cache_knowledge))]
 
 
-def doc_url_to_file_metadata(f_url: str) -> FileMetadata:
+def doc_url_to_file_metadata(f_url: str, cache_knowledge: bool = True) -> FileMetadata:
     from googleapiclient.errors import HttpError
 
     f = furl(f_url.strip("/"))
@@ -314,7 +321,20 @@ def doc_url_to_file_metadata(f_url: str) -> FileMetadata:
     else:
         try:
             if is_user_uploaded_url(f_url):
-                r = requests.head(f_url)
+
+                if cache_knowledge:
+                    logger.debug(f"using cached metadata for {f_url}")
+                    name = f.path.segments[-1]
+                    return FileMetadata(
+                        name=name,
+                        etag=None,
+                        mime_type=mimetypes.guess_type(name)[0],
+                        total_bytes=0,
+                    )
+                else:
+                    logger.debug(f"fetching latest metadata for {f_url}")
+                    r = requests.head(f_url)
+
             else:
                 r = requests.head(
                     f_url,
@@ -402,10 +422,10 @@ def get_or_create_embedded_file(
     """
     lookup = dict(
         url=f_url,
-        metadata__name=file_meta.name,
-        metadata__etag=file_meta.etag,
-        metadata__mime_type=file_meta.mime_type,
-        metadata__total_bytes=file_meta.total_bytes,
+        # metadata__name=file_meta.name,
+        # metadata__etag=file_meta.etag,
+        # metadata__mime_type=file_meta.mime_type,
+        # metadata__total_bytes=file_meta.total_bytes,
         max_context_words=max_context_words,
         scroll_jump=scroll_jump,
         google_translate_target=google_translate_target or "",
diff --git a/recipes/DocExtract.py b/recipes/DocExtract.py
index 0fa063379..951808784 100644
--- a/recipes/DocExtract.py
+++ b/recipes/DocExtract.py
@@ -225,6 +225,7 @@ def run_v2(
                 message="Updating sheet...",
             )
         else:
+            # TODO: implement this as well
             file_url_metas = yield from flatapply_parallel(
                 doc_or_yt_url_to_metadatas,
                 request.documents,