app.py

import gradio as gr
from typing import TypedDict, Annotated
from huggingface_hub import InferenceClient, login
import random
from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint, HuggingFacePipeline
#from langchain.schema import AIMessage, HumanMessage
from langgraph.graph.message import add_messages
from langgraph.prebuilt import ToolNode, tools_condition
from langchain_core.messages import AnyMessage, HumanMessage, AIMessage
from langchain.tools import Tool
import os
import datasets
from langchain.docstore.document import Document
from langgraph.graph import START, StateGraph
from langchain_community.retrievers import BM25Retriever

HUGGINGFACEHUB_API_TOKEN = os.environ["HUGGINGFACEHUB_API_TOKEN"]
login(token=HUGGINGFACEHUB_API_TOKEN, add_to_git_credential=True)

# Load the dataset
guest_dataset = datasets.load_dataset("agents-course/unit3-invitees", split="train")

# Convert dataset entries into Document objects
docs = [
    Document(
        page_content="\n".join([
            f"Name: {guest['name']}",
            f"Relation: {guest['relation']}",
            f"Description: {guest['description']}",
            f"Email: {guest['email']}"
        ]),
        metadata={"name": guest["name"]}
    )
    for guest in guest_dataset
]


bm25_retriever = BM25Retriever.from_documents(docs)

def extract_text(query: str) -> str:
    """Retrieves detailed information about gala guests based on their name or relation."""
    results = bm25_retriever.invoke(query)
    if results:
        return "\n\n".join([doc.page_content for doc in results[:3]])
    else:
        return "No matching guest information found."

llm = HuggingFaceEndpoint(
    repo_id="HuggingFaceH4/zephyr-7b-beta",
    task="text-generation",
    max_new_tokens=512,
    do_sample=False,
    repetition_penalty=1.03,
)

model = ChatHuggingFace(llm=llm, verbose=True)

"""
def predict(message, history):
    history_langchain_format = []
    for msg in history:
        if msg['role'] == "user":
            history_langchain_format.append(HumanMessage(content=msg['content']))
        elif msg['role'] == "assistant":
            history_langchain_format.append(AIMessage(content=msg['content']))
    history_langchain_format.append(HumanMessage(content=message))
    gpt_response = model.invoke(history_langchain_format)
    return gpt_response.content
"""

def predict(message, history):
    # Convert Gradio history to LangChain message format
    history_langchain_format = []
    for msg in history:
        if msg['role'] == "user":
            history_langchain_format.append(HumanMessage(content=msg['content']))
        elif msg['role'] == "assistant":
            history_langchain_format.append(AIMessage(content=msg['content"]))
    
    # Add new user message
    history_langchain_format.append(HumanMessage(content=message))
    
    # Invoke Alfred agent with full message history
    response = alfred.invoke(
        input={"messages": history_langchain_format},
        config={"recursion_limit": 100}
    )
    
    # Extract final assistant message
    return response["messages"][-1].content

# setup agents

guest_info_tool = Tool(
    name="guest_info_retriever",
    func=extract_text,
    description="Retrieves detailed information about gala guests based on their name or relation."
)

tools = [guest_info_tool]
chat_with_tools = model.bind_tools(tools)

# Generate the AgentState and Agent graph
class AgentState(TypedDict):
    messages: Annotated[list[AnyMessage], add_messages]

def assistant(state: AgentState):
    return {
        "messages": [chat_with_tools.invoke(state["messages"])],
    }

## The graph
builder = StateGraph(AgentState)

# Define nodes: these do the work
builder.add_node("assistant", assistant)
builder.add_node("tools", ToolNode(tools))

# Define edges: these determine how the control flow moves
builder.add_edge(START, "assistant")
builder.add_conditional_edges(
    "assistant",
    # If the latest message requires a tool, route to tools
    # Otherwise, provide a direct response
    tools_condition,
)
builder.add_edge("tools", "assistant")
alfred = builder.compile()

demo = gr.ChatInterface(
    predict,
    type="messages"
)


demo.launch()