Update app.py

app.py

@@ -225,70 +225,123 @@
 
 
 
-
 import gradio as gr  
-import boto3
-import json
-from botocore.exceptions import ClientError
-import os
+client = boto3.client(service_name='bedrock-runtime',region_name='us-east-1',aws_access_key_id=access_key_id,aws_secret_access_key=secret_access_key)
+
+prompt = """
+        <|begin_of_text|>
+        {history}
+        <|start_header_id|>user<|end_header_id|>
+        {input}
+        <|eot_id|>
+        <|start_header_id|>assistant<|end_header_id|>
+        """
+prompt_temp = PromptTemplate(input_variables=["history", "input"], template=template)
+
+def generate(
+    prompt_temp, temperature=0.2, max_gen_len=1024, top_p=0.95,
+):
+    temperature = float(temperature)
+    if temperature < 1e-2:
+        temperature = 1e-2
+    top_p = float(top_p)
+
+    generate_kwargs = dict(
+        temperature=temperature,
+        max_gen_len=max_gen_len,
+        top_p=top_p)
+
+    conversation = ConversationChain(
+        prompt=prompt_temp,
+        llm=llm,
+        verbose=True,
+        memory= ConversationBufferMemory(ai_prefix="AI Assistant")
+    )
+
+    chat_history = []
+
+    #result =conversation.predict(input="Hi there!")
+    
+    result = conversation({"input": message, "history":chat_history })
+    chat_history.append((message, result['response']))
+    return result['response']
+
+demo=gr.ChatInterface(qa_fn)
+
+
+demo.queue().launch(show_api=False)
+
+
+
+
+
+
 
-access_key_id = os.environ['aws_access_key_id']
-secret_access_key = os.environ['aws_secret_access_key']
 
 
-bedrock = boto3.client(service_name='bedrock-runtime',region_name='us-east-1',aws_access_key_id=access_key_id,aws_secret_access_key=secret_access_key)
+# import gradio as gr  
+# import boto3
+# import json
+# from botocore.exceptions import ClientError
+# import os
+
+# access_key_id = os.environ['aws_access_key_id']
+# secret_access_key = os.environ['aws_secret_access_key']
+
+
+# bedrock = boto3.client(service_name='bedrock-runtime',region_name='us-east-1',aws_access_key_id=access_key_id,aws_secret_access_key=secret_access_key)
 
 
-def invoke_llama3_8b(user_message):
-    try:
-        # Set the model ID, e.g., Llama 3 8B Instruct.
-        model_id = "meta.llama3-8b-instruct-v1:0"
+# def invoke_llama3_8b(user_message):
+#     try:
+#         # Set the model ID, e.g., Llama 3 8B Instruct.
+#         model_id = "meta.llama3-8b-instruct-v1:0"
          
          
-        # Embed the message in Llama 3's prompt format.
-        prompt = f"""
-        <|begin_of_text|>
-        <|start_header_id|>user<|end_header_id|>
-        {user_message}
-        <|eot_id|>
-        <|start_header_id|>assistant<|end_header_id|>
-        """
+#         # Embed the message in Llama 3's prompt format.
+#         prompt = f"""
+#         <|begin_of_text|>
+#         <|start_header_id|>user<|end_header_id|>
+#         {user_message}
+#         <|eot_id|>
+#         <|start_header_id|>assistant<|end_header_id|>
+#         """
          
-        # Format the request payload using the model's native structure.
-        request = {
-            "prompt": prompt,
-            # Optional inference parameters:
-            "max_gen_len": 1024,
-            "temperature": 0.6,
-            "top_p": 0.9,
-        }
+#         # Format the request payload using the model's native structure.
+#         request = {
+#             "prompt": prompt,
+#             # Optional inference parameters:
+#             "max_gen_len": 1024,
+#             "temperature": 0.6,
+#             "top_p": 0.9,
+#         }
          
-        # Encode and send the request.
-        response = bedrock.invoke_model(body=json.dumps(request), modelId=model_id)
+#         # Encode and send the request.
+#         response = bedrock.invoke_model(body=json.dumps(request), modelId=model_id)
          
-        # Decode the native response body.
-        model_response = json.loads(response["body"].read())
+#         # Decode the native response body.
+#         model_response = json.loads(response["body"].read())
          
-        # Extract and print the generated text.
-        response_text = model_response["generation"]
+#         # Extract and print the generated text.
+#         response_text = model_response["generation"]
 
-        return response_text
+#         return response_text
 
-    except ClientError:
-        print("Couldn't invoke llama3 8B")
-        raise
+#     except ClientError:
+#         print("Couldn't invoke llama3 8B")
+#         raise
 
 
     
-mychatbot = gr.Chatbot(
-    avatar_images=["./user.png", "./bot.png"], bubble_full_width=False, show_label=False, show_copy_button=True, likeable=True,)
+# mychatbot = gr.Chatbot(
+#     avatar_images=["./user.png", "./bot.png"], bubble_full_width=False, show_label=False, show_copy_button=True, likeable=True,)
 
-demo = gr.ChatInterface(fn=invoke_llama3_8b, 
-                        chatbot=mychatbot,
-                        title="llama3-Chat",
-                        retry_btn=None,
-                        undo_btn=None
-                       )
+# demo = gr.ChatInterface(fn=invoke_llama3_8b, 
+#                         chatbot=mychatbot,
+#                         title="llama3-Chat",
+#                         retry_btn=None,
+#                         undo_btn=None
+#                        )
 
-demo.queue().launch(show_api=False)
+# demo.queue().launch(show_api=False)