Update app.py

app.py

@@ -1,18 +1,34 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
 from transformers import pipeline
+from typing import List, Dict, Tuple
+import re
+import os
+
+# Set up the retriever pipeline
+retriever = pipeline('retrieval', model='facebook/rag-token-nq')
+
+# Load your PDF document
+pdf_path = "apexcustoms.pdf"
+with open(pdf_path, 'rb') as f:
+    pdf_text = f.read().decode('utf-8', errors='ignore')
+
+# Split the PDF text into chunks
+split_pattern = r'\n\n'
+doc_chunks = re.split(split_pattern, pdf_text)
+
+# Create the retriever input
+corpus = [{"text": chunk} for chunk in doc_chunks]
 
 """
 For more information on huggingface_hub Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
 """
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
-# Load the question-answering pipeline
-qa_pipeline = pipeline("question-answering", model="deepset/roberta-base-squad2")
 
 def respond(
     message,
-    history: list[tuple[str, str]],
+    history: List[Tuple[str, str]],
     system_message,
     max_tokens,
     temperature,
@@ -28,6 +44,11 @@ def respond(
 
     messages.append({"role": "user", "content": message})
 
+    # Retrieve relevant context from the PDF
+    retrieval_output = retriever(message, corpus, top_k=3)
+    retrieved_contexts = [passage['text'] for passage in retrieval_output['retrieved_passages']]
+    context = ' '.join(retrieved_contexts)
+
     response = ""
 
     for message in client.chat_completion(
@@ -36,12 +57,12 @@ def respond(
         stream=True,
         temperature=temperature,
         top_p=top_p,
+        context=context,  # Include the retrieved context
     ):
         token = message.choices[0].delta.content
 
-        response = qa_pipeline(question=message, context=system_message)["answer"]
         response += token
-        return response
+        yield response
 
 """
 For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface