app.py

from typing import Annotated, Sequence, TypedDict
import functools
import operator
from bertopic import BERTopic
from langchain.agents import AgentExecutor, create_openai_tools_agent
from langchain_core.messages import BaseMessage, HumanMessage
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.output_parsers.openai_functions import JsonOutputFunctionsParser
from langgraph.graph import END, StateGraph
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_experimental.tools import PythonREPLTool

# Initialize tools
tavily_tool = TavilySearchResults(max_results=5)
python_repl_tool = PythonREPLTool()

# Load BERTopic model
topic_model = BERTopic.load("topic_model")

# Function to create an agent
def create_agent(llm: ChatOpenAI, tools: list, system_prompt: str):
    prompt = ChatPromptTemplate.from_messages(
        [
            ("system", system_prompt),
            MessagesPlaceholder(variable_name="messages"),
            MessagesPlaceholder(variable_name="agent_scratchpad"),
        ]
    )
    agent = create_openai_tools_agent(llm, tools, prompt)
    executor = AgentExecutor(agent=agent, tools=tools)
    return executor

# Function to define an agent node
def agent_node(state, agent, name):
    result = agent.invoke(state)
    return {"messages": [HumanMessage(content=result["output"], name=name)]}

# Define the Viewer agent using the BERTopic model
def viewer_agent(state):
    review = state["messages"][-1].content
    index = topic_model.transform([review])[0][0]
    answer = topic_model.topic_labels_.get(index)
    return {"messages": [HumanMessage(content=answer, name="Viewer")]}

# Define AgentState
class AgentState(TypedDict):
    messages: Annotated[Sequence[BaseMessage], operator.add]
    next: str

# Create LLM for the supervisor
llm = ChatOpenAI(model="gpt-4-1106-preview")

# Define the system prompt for the supervisor
system_prompt = (
    "You are a supervisor tasked with managing a conversation between the following workers: Researcher, Coder, Viewer. "
    "Given the following user request, respond with the worker to act next. Each worker will perform a task and respond with their results and status. "
    "When finished, respond with FINISH. If the request seems like a product review or sentiment analysis, route it to the Viewer."
)

# Define options
options = ["FINISH", "Researcher", "Coder", "Viewer"]

# Define the function for routing
function_def = {
    "name": "route",
    "description": "Select the next role.",
    "parameters": {
        "title": "routeSchema",
        "type": "object",
        "properties": {
            "next": {
                "title": "Next",
                "anyOf": [
                    {"enum": options},
                ],
            }
        },
        "required": ["next"],
    },
}

# Define the prompt for the supervisor
prompt = ChatPromptTemplate.from_messages(
    [
        ("system", system_prompt),
        MessagesPlaceholder(variable_name="messages"),
        (
            "system",
            "Given the conversation above, who should act next? Or should we FINISH? Select one of: {options}",
        ),
    ]
).partial(options=str(options), members="Researcher, Coder, Viewer")

# Create the supervisor chain
supervisor_chain = (
    prompt
    | llm.bind_functions(functions=[function_def], function_call="route")
    | JsonOutputFunctionsParser()
)

# Define agents
research_agent = create_agent(llm, [tavily_tool], "You are a web researcher.")
research_node = functools.partial(agent_node, agent=research_agent, name="Researcher")

code_agent = create_agent(
    llm,
    [python_repl_tool],
    "You may generate safe python code to analyze data and generate charts using matplotlib.",
)
code_node = functools.partial(agent_node, agent=code_agent, name="Coder")

viewer_node = functools.partial(viewer_agent)

# Create the workflow
workflow = StateGraph(AgentState)
workflow.add_node("Researcher", research_node)
workflow.add_node("Coder", code_node)
workflow.add_node("Viewer", viewer_node)
workflow.add_node("supervisor", supervisor_chain)

# Add edges for each agent to report back to the supervisor
members = ["Researcher", "Coder", "Viewer"]
for member in members:
    workflow.add_edge(member, "supervisor")

# Add conditional edges
conditional_map = {k: k for k in members}
conditional_map["FINISH"] = END
workflow.add_conditional_edges("supervisor", lambda x: x["next"], conditional_map)

# Set the entry point
workflow.set_entry_point("supervisor")

# Compile the workflow
graph = workflow.compile()

# Testing the workflow
for s in graph.stream(
    {
        "messages": [
            HumanMessage(content="write a report of gopal who worked in 3 k technologies")
        ]
    }
):
    if "__end__" not in s:
        print(s)
        print("----")