Update app.py

app.py

@@ -3,36 +3,61 @@ from model_loader import load_embedding_model, load_llm
 from transcript_handler import chunk_text, embed_chunks, create_faiss_index
 from qa_engine import query_faiss, build_prompt
 
+# Load models
 embedder = load_embedding_model()
 llm = load_llm()
 
-index = None
-chunks = []
-
-def upload_transcript(file):
-    global index, chunks
-    text = file.read().decode("utf-8")
-    chunks = chunk_text(text)
-    embeddings, chunks = embed_chunks(chunks, embedder)
-    index = create_faiss_index(embeddings)
-    return "Transcript uploaded and indexed successfully!"
-
-def chat_with_transcript(query):
-    if not index:
-        return "Please upload a transcript first."
-    context = query_faiss(query, index, embedder, chunks)
-    prompt = build_prompt(context, query)
-    response = llm(prompt)[0]['generated_text'].split("Answer:")[-1].strip()
-    return response
-
+# Main Gradio app
 with gr.Blocks() as demo:
     gr.Markdown("# 📄 Chat with a Transcript (Open Source + Free!)")
+    
+    # State variables for storing index and chunks per session
+    index_state = gr.State(None)
+    chunks_state = gr.State([])
+
     transcript_input = gr.File(label="Upload Transcript (.txt)")
     upload_button = gr.Button("Upload and Process")
     query_input = gr.Textbox(label="Ask a question about the transcript")
     answer_output = gr.Textbox(label="Answer")
 
-    upload_button.click(upload_transcript, inputs=[transcript_input], outputs=[])
-    query_input.submit(chat_with_transcript, inputs=[query_input], outputs=[answer_output])
+    def upload_transcript(file, chunks_state):
+        try:
+            text = file.read().decode("utf-8")
+            if not text.strip():
+                return "Error: Uploaded file is empty.", None, []
+            
+            chunks = chunk_text(text)
+            if not chunks:
+                return "Error: No chunks generated from the transcript.", None, []
+            
+            embeddings, chunks = embed_chunks(chunks, embedder)
+            if embeddings.size == 0:
+                return "Error: Failed to generate embeddings.", None, []
+            
+            index = create_faiss_index(embeddings)
+            return "Transcript uploaded and indexed successfully!", index, chunks
+        except Exception as e:
+            return f"Error processing transcript: {str(e)}", None, []
+
+    def chat_with_transcript(query, index_state, chunks_state):
+        if index_state is None:
+            return "Please upload a transcript first."
+        context = query_faiss(query, index_state, embedder, chunks_state)
+        prompt = build_prompt(context, query)
+        response = llm(prompt)[0]['generated_text']
+        if "Answer:" not in response:
+            return "Error: Unable to parse the model's response."
+        return response.split("Answer:")[-1].strip()
+
+    upload_button.click(
+        upload_transcript,
+        inputs=[transcript_input, chunks_state],
+        outputs=[answer_output, index_state, chunks_state]
+    )
+    query_input.submit(
+        chat_with_transcript,
+        inputs=[query_input, index_state, chunks_state],
+        outputs=[answer_output]
+    )
 
-demo.launch()
+demo.launch()