Improve error handling when documents not searched with stack trace

- Stop extract OCR content from PDFs
- Only use agent knowledge base when user not provided

src/khoj/database/adapters/__init__.py

@@ -1463,12 +1463,15 @@ class EntryAdapters:
         file_filters = EntryAdapters.file_filter.get_filter_terms(query)
         date_filters = EntryAdapters.date_filter.get_query_date_range(query)
 
-        user_or_agent = Q(user=user)
+        owner_filter = Q()
+
+        if user != None:
+            owner_filter = Q(user=user)
         if agent != None:
-            user_or_agent |= Q(agent=agent)
+            owner_filter |= Q(agent=agent)
 
         if len(word_filters) == 0 and len(file_filters) == 0 and len(date_filters) == 0:
-            return Entry.objects.filter(user_or_agent)
+            return Entry.objects.filter(owner_filter)
 
         for term in word_filters:
             if term.startswith("+"):
@@ -1504,7 +1507,7 @@ class EntryAdapters:
                 formatted_max_date = date.fromtimestamp(max_date).strftime("%Y-%m-%d")
                 q_filter_terms &= Q(embeddings_dates__date__lte=formatted_max_date)
 
-        relevant_entries = Entry.objects.filter(user_or_agent).filter(q_filter_terms)
+        relevant_entries = Entry.objects.filter(owner_filter).filter(q_filter_terms)
         if file_type_filter:
             relevant_entries = relevant_entries.filter(file_type=file_type_filter)
         return relevant_entries
@@ -1519,13 +1522,18 @@ class EntryAdapters:
         max_distance: float = math.inf,
         agent: Agent = None,
     ):
-        user_or_agent = Q(user=user)
+        owner_filter = Q()
 
+        if user != None:
+            owner_filter = Q(user=user)
         if agent != None:
-            user_or_agent |= Q(agent=agent)
+            owner_filter |= Q(agent=agent)
+
+        if owner_filter == Q():
+            return Entry.objects.none()
 
         relevant_entries = EntryAdapters.apply_filters(user, raw_query, file_type_filter, agent)
-        relevant_entries = relevant_entries.filter(user_or_agent).annotate(
+        relevant_entries = relevant_entries.filter(owner_filter).annotate(
             distance=CosineDistance("embeddings", embeddings)
         )
         relevant_entries = relevant_entries.filter(distance__lte=max_distance)

src/khoj/processor/content/pdf/pdf_to_entries.py

@@ -67,7 +67,7 @@ class PdfToEntries(TextToEntries):
                     bytes = pdf_files[pdf_file]
                     f.write(bytes)
                 try:
-                    loader = PyMuPDFLoader(f"{tmp_file}", extract_images=True)
+                    loader = PyMuPDFLoader(f"{tmp_file}", extract_images=False)
                     pdf_entries_per_file = [page.page_content for page in loader.load()]
                 except ImportError:
                     loader = PyMuPDFLoader(f"{tmp_file}")

src/khoj/routers/api_chat.py

@@ -859,7 +859,7 @@ async def chat(
                     defiltered_query = result[2]
         except Exception as e:
             error_message = f"Error searching knowledge base: {e}. Attempting to respond without document references."
-            logger.warning(error_message)
+            logger.error(error_message, exc_info=True)
             async for result in send_event(
                 ChatEvent.STATUS, "Document search failed. I'll try respond without document references"
             ):