Update app.py

app.py

@@ -1,9 +1,13 @@
+import os
 import requests
 import chainlit as cl
-from langchain.prompts import PromptTemplate
-from langchain.chains import LLMChain
+from langchain import hub
+from langchain.agents import AgentExecutor, load_tools
+from langchain.agents.format_scratchpad import format_log_to_str
+from langchain.agents.output_parsers import ReActJsonSingleInputOutputParser
+from langchain.tools.render import render_text_description
 from langchain_community.llms import HuggingFaceHub
-import os
+from langchain_community.utilities import SerpAPIWrapper
 
 # Define the API endpoint URL
 API_URL = "https://api-inference.huggingface.co/models/mistralai/Mixtral-8x7B-Instruct-v0.1"
@@ -11,26 +15,6 @@ API_URL = "https://api-inference.huggingface.co/models/mistralai/Mixtral-8x7B-In
 # Replace this with your actual Hugging Face API token
 HUGGINGFACE_API_TOKEN = os.getenv("HUGGINGFACEHUB_API_TOKEN")
 
-# Function to query the Hugging Face Inference API
-def query_hf_api(inputs):
-    headers = {"Authorization": f"Bearer {HUGGINGFACE_API_TOKEN}"}
-    payload = {
-        "inputs": inputs,
-        "parameters": {"temperature": 0.7},
-        "options": {"use_cache": False}
-    }
-    response = requests.post(API_URL, headers=headers, json=payload)
-    return response.json()
-
-# Example usage
-def get_response(question):
-    result = query_hf_api(question)
-    if isinstance(result, list) and result:
-        # Assuming the first element of the list is the answer
-        return result[0]
-    else:
-        return ""
-
 # Load the LLM (Hugging Face Hub)
 llm = HuggingFaceHub(
     repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
@@ -43,43 +27,45 @@ llm = HuggingFaceHub(
     },
 )
 
-# Define the prompt template
-template = """
-You are a helpful AI assistant. Provide the answer for the following question:
+# Setup tools
+tools = load_tools(["serpapi", "llm-math"], llm=llm)
 
-Question: {question}
+# Setup ReAct style prompt
+prompt = hub.pull("hwchase17/react-json")
+prompt = prompt.partial(
+    tools=render_text_description(tools),
+    tool_names=", ".join([t.name for t in tools]),
+)
 
-Answer:
-"""
-prompt = PromptTemplate(template=template, input_variables=["question"])
+# Define the agent
+chat_model_with_stop = llm.bind(stop=["\nObservation"])
+agent = (
+    {
+        "input": lambda x: x["input"],
+        "agent_scratchpad": lambda x: format_log_to_str(x["intermediate_steps"]),
+    }
+    | prompt
+    | chat_model_with_stop
+    | ReActJsonSingleInputOutputParser()
+)
 
-# Create the LLMChain
-llm_chain = LLMChain(llm=llm, prompt=prompt)
+# Instantiate AgentExecutor
+agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
 
 @cl.on_chat_start
 async def main():
-    cl.user_session.set("llm_chain", llm_chain)
+    cl.user_session.set("agent_executor", agent_executor)
     return "API-based Mixtral-Instruct"
 
 @cl.on_message
 async def run(message: cl.Message):
-    cb = cl.AsyncLangchainCallbackHandler(
-        stream_final_answer=True, answer_prefix_tokens=["Answer:"]
-    )
-
-    # Retrieve the LLMChain from the user session
-    llm_chain = cl.user_session.get("llm_chain")
+    # Retrieve the AgentExecutor from the user session
+    agent_executor = cl.user_session.get("agent_executor")
 
     question = message.content
-    result = llm_chain.run(question)
-
-    # Extract the generated response from the LLMChain output
-    if "Answer:" in result:
-        answer = result.split("Answer:")[1].strip()
-    else:
-        answer = result
+    result = agent_executor.invoke({"input": question})
 
-    if answer:
-        await cl.Message(content=answer).send()
+    if result["output"]:
+        await cl.Message(content=result["output"]).send()
     else:
         await cl.Message(content="I'm sorry, I couldn't find an answer.").send()