Update app.py

app.py

@@ -135,6 +135,20 @@ if "llm" not in st.session_state:
         max_tokens=2048
     )
 
+# Custom logger for yt-dlp to redirect logs to Streamlit
+class StreamlitLogger:
+    def debug(self, msg):
+        st.text(f"[yt-dlp DEBUG] {msg}")
+    
+    def info(self, msg):
+        st.info(f"[yt-dlp INFO] {msg}")
+    
+    def warning(self, msg):
+        st.warning(f"[yt-dlp WARNING] {msg}")
+    
+    def error(self, msg):
+        st.error(f"[yt-dlp ERROR] {msg}")
+
 # Sidebar for URL and YouTube input
 with st.sidebar:
     st.header("Enter Web URL")
@@ -220,9 +234,11 @@ def fetch_youtube_captions_api(video_id, api_key):
 
         # Note: Downloading captions requires OAuth 2.0 authentication
         st.warning(
-            "Downloading captions requires OAuth 2.0 authentication, which is not supported in this environment. "
-            "English captions are available but cannot be fetched with an API key alone. "
-            "Consider setting up OAuth 2.0 for full functionality (see documentation)."
+            "English captions are available for this video but cannot be fetched with an API key alone. "
+            "Downloading captions requires OAuth 2.0 authentication, which is not supported in Hugging Face Spaces without user interaction. "
+            "To fetch captions:\n"
+            "- Test locally with OAuth 2.0 setup (see instructions in the documentation).\n"
+            "- Or try a video with transcripts available (e.g., https://www.youtube.com/watch?v=dQw4w9WgXcQ)."
         )
         return None
 
@@ -268,7 +284,7 @@ def extract_subtitles_with_ytdlp(video_url):
             return None
 
         with yt_dlp.YoutubeDL(ydl_opts) as ydl:
-            ydl.params['logger'] = st
+            ydl.params['logger'] = StreamlitLogger()
             info = ydl.extract_info(video_url, download=False)
             available_subs = info.get('subtitles', {})
             auto_subs = info.get('automatic_captions', {})
@@ -319,12 +335,18 @@ def process_content(text, embeddings, source):
     docs = text_splitter.create_documents([text], metadatas=[{"source": source}])
     if docs:
         st.text(f"Document metadata: {docs[0].metadata}")
+    else:
+        st.error("No documents created from the content.")
+        return None
     vectorstore = FAISS.from_documents(docs, embeddings)
     st.text(f"Vector store created with {len(docs)} documents.")
     return vectorstore
 
 # Function to create QA chain
 def create_qa_chain(vectorstore, llm):
+    if vectorstore is None:
+        st.error("Vector store is not initialized. Cannot create QA chain.")
+        return None
     retriever = vectorstore.as_retriever(search_kwargs={"k": 2})
     qa_chain = RetrievalQAWithSourcesChain.from_chain_type(
         llm=llm,
@@ -447,7 +469,11 @@ if ask_clicked and query:
                 try:
                     if "qa_chain" not in st.session_state or st.session_state.qa_chain is None:
                         st.session_state.qa_chain = create_qa_chain(st.session_state.vectorstore, st.session_state.llm)
+                        if st.session_state.qa_chain is None:
+                            st.error("Failed to create QA chain.")
+                            st.stop()
                     
+                    st.text(f"Querying with question: {query}")
                     result = st.session_state.qa_chain({"question": query}, return_only_outputs=True)
 
                     if not result.get("answer"):