new code

src/app.py

@@ -16,7 +16,7 @@ with demo:
 
     # Event handler for submitting text and generating response
     submit_button.click(pdf_chatbot.add_text, inputs=[chat_history, txt], outputs=[chat_history], queue=False).\
-        success(pdf_chatbot.generate_response, inputs=[chat_history, txt, uploaded_pdf], outputs=[chat_history, txt]).\
+        success(pdf_chatbot.generate_response, inputs=[chat_history, txt, uploaded_pdf], outputs=[chat_history,txt]).\
         success(pdf_chatbot.render_file, inputs=[uploaded_pdf], outputs=[show_img])
 
 if __name__ == "__main__":

src/pdfchatbot.py

@@ -14,7 +14,6 @@ import spaces
 from langchain_text_splitters import CharacterTextSplitter
 
 
-
 class PDFChatBot:
     def __init__(self, config_path="config.yaml"):
         """
@@ -37,6 +36,8 @@ class PDFChatBot:
         self.pipeline = None
         self.chain = None
         self.chunk_size = None
+        self.current_context = None
+        self.format_seperator="""\n\n--\n\n"""
         #self.chunk_size_slider = chunk_size_slider
 
     def load_config(self, file_path):
@@ -128,6 +129,46 @@ class PDFChatBot:
         )
         self.pipeline = HuggingFacePipeline(pipeline=pipe)
 
+    def create_organic_pipeline(self):
+        self.pipeline = pipeline(
+            "text-generation",
+            model=self.config.get("autoModelForCausalLM"),
+            model_kwargs={"torch_dtype": torch.bfloat16},
+            device="cuda",
+        )
+
+    def get_organic_context(self, query):
+        documents = self.vectordb.similarity_search_with_relevance_scores(query, k=self.k)
+        context = self.format_seperator.join([doc.page_content for doc, score in documents])
+        self.current_context = context
+        print(self.current_context)
+
+    def create_organic_response(self, history, query):
+        self.get_organic_context(query)
+        messages = [
+            {"role": "system", "content": "From the the contained given below, answer the question of user \n " + self.current_context},
+            {"role": "user", "content": query},
+        ]
+        prompt = self.pipeline.tokenizer.apply_chat_template(
+            messages,
+            tokenize=False,
+            add_generation_prompt=True
+        )
+        terminators = [
+            self.pipeline.tokenizer.eos_token_id,
+            self.pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
+        ]
+        temp = 0.1
+        outputs = pipeline(
+            prompt,
+            max_new_tokens=1024,
+            eos_token_id=terminators,
+            do_sample=True,
+            temperature=temp,
+            top_p=0.9,
+        )
+        return outputs[0]["generated_text"][len(prompt):]
+
     def create_chain(self):
         """
         Create a Conversational Retrieval Chain
@@ -153,8 +194,8 @@ class PDFChatBot:
         self.load_vectordb()
         self.load_tokenizer()
         self.load_model()
-        self.create_pipeline()
-        self.create_chain()
+        self.create_organic_pipeline()
+        #self.create_chain()
     @spaces.GPU
     def generate_response(self, history, query, file):
         """
@@ -176,11 +217,14 @@ class PDFChatBot:
             self.process_file(file)
             self.processed = True
 
-        result = self.chain({"question": query, 'chat_history': self.chat_history}, return_only_outputs=True)
+        """result = self.chain({"question": query, 'chat_history': self.chat_history}, return_only_outputs=True)
         self.chat_history.append((query, result["answer"]))
         for char in result['answer']:
             history[-1][-1] += char
-        return history, " "
+        return history, " """""
+
+        result = self.create_organic_response(history="",query=query)
+        return result,""
 
     def render_file(self, file,chunk_size):
         """