app.py

import os
import json
import operator
import streamlit as st
import tempfile
from typing import TypedDict, Annotated, Sequence
from langchain_openai import ChatOpenAI
from langchain_core.tools import tool
from langchain_core.messages import HumanMessage, AIMessage
from langchain_core.utils.function_calling import convert_to_openai_tool
from langgraph.graph import StateGraph, END

# Environment Setup
os.environ['OPENAI_API_KEY'] = os.getenv("OPENAI_API_KEY")

# Model Initialization
model = ChatOpenAI(temperature=0)

# Define the tool
@tool
def multiply(first_number: int, second_number: int):
    """Multiplies two numbers together and returns the result."""
    return first_number * second_number

# Bind tool to model
model_with_tools = model.bind(tools=[convert_to_openai_tool(multiply)])

# State Setup
class AgentState(TypedDict):
    messages: Annotated[Sequence, operator.add]

graph = StateGraph(AgentState)

# Model Invocation
def invoke_model(state):
    """
    Invoke the model and handle tool invocation logic.
    """
    # Extract the question as a string
    question = state['messages'][-1].content if isinstance(state['messages'][-1], HumanMessage) else state['messages'][-1]
    response = model_with_tools.invoke(question)

    # If the response is plain text (no tool calls)
    if isinstance(response, str):
        return {"messages": [AIMessage(content=response)]}

    # If no tool calls exist
    if not response.additional_kwargs.get("tool_calls", []):
        return {"messages": [AIMessage(content=response.content)]}

    # If tool calls are present, return the full response
    return {"messages": [response]}

graph.add_node("agent", invoke_model)

# Tool Invocation
def invoke_tool(state):
    """
    Invoke the 'multiply' tool if it's called by the model.
    """
    tool_calls = state['messages'][-1].additional_kwargs.get("tool_calls", [])
    for tool_call in tool_calls:
        if tool_call.get("function", {}).get("name") == "multiply":
            arguments = json.loads(tool_call.get("function").get("arguments"))
            result = multiply.invoke(arguments)
            return {"messages": [AIMessage(content=f"Tool Result: {result}")]}
    return {"messages": [AIMessage(content="No valid tool input provided.")]}

graph.add_node("tool", invoke_tool)
graph.add_edge("tool", END)
graph.set_entry_point("agent")

# Router Logic
def router(state):
    """
    Decide whether to invoke a tool or return the response.
    """
    tool_calls = state['messages'][-1].additional_kwargs.get("tool_calls", [])
    return "tool" if tool_calls else END

graph.add_conditional_edges("agent", router, {"tool": "tool", END: END})
app_graph = graph.compile()

# Save graph visualization as an image
with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmpfile:
    graph_viz = app_graph.get_graph(xray=True)
    tmpfile.write(graph_viz.draw_mermaid_png())
    graph_image_path = tmpfile.name

# Streamlit Interface
st.title("Simple Tool Calling Demo")

# Display the workflow graph
st.image(graph_image_path, caption="Workflow Visualization")

tab1, tab2 = st.tabs(["Try Multiplication", "Ask General Queries"])

# Multiplication Tool Tab
with tab1:
    st.subheader("Try Multiplication")
    col1, col2 = st.columns(2)

    with col1:
        first_number = st.number_input("First Number", value=0, step=1)
    with col2:
        second_number = st.number_input("Second Number", value=0, step=1)

    if st.button("Multiply"):
        question = f"What is {first_number} * {second_number}?"
        try:
            output = app_graph.invoke({"messages": [HumanMessage(content=question)]})
            st.success(output['messages'][-1].content)
        except Exception as e:
            st.error(f"Error: {e}")

# General Query Tab
with tab2:
    st.subheader("General Query")
    user_input = st.text_input("Enter your question here")

    if st.button("Submit"):
        if user_input:
            try:
                # Pass the user input as a HumanMessage
                result = app_graph.invoke({"messages": [HumanMessage(content=user_input)]})
                st.write("Response:")
                st.success(result['messages'][-1].content)
            except Exception as e:
                st.error(f"Error: {e}")
        else:
            st.warning("Please enter a valid input.")

# Sidebar for References
st.sidebar.title("References")
st.sidebar.markdown("1. [LangGraph Tool Calling](https://github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/LangGraph_02_ToolCalling.ipynb)")