presentation_api.py

import uuid
from fastapi import FastAPI
from fastapi.responses import StreamingResponse
from fastapi.middleware.cors import CORSMiddleware
from langchain_core.messages import BaseMessage, HumanMessage, trim_messages
from langchain_core.tools import tool
from langchain_openai import ChatOpenAI
from langgraph.checkpoint.memory import MemorySaver
from langgraph.prebuilt import create_react_agent
from pydantic import BaseModel
from typing import Optional
import json
from sse_starlette.sse import EventSourceResponse
from datetime import datetime
from fastapi import APIRouter

router = APIRouter(
    prefix="/presentation",
    tags=["presentation"]
)


@tool
def plan(input: dict) -> str:
    """Create a presentation plan with numbered slides and their descriptions.
    Args:
        input: Dictionary containing presentation details
    Returns:
        A dictionary with slide numbers as keys and descriptions as values
    """
    return "plan created"

@tool
def create_slide(slideno: int, content: str) -> str:
    """Create a single presentation slide.
    Args:
        slideno: The slide number to create
        content: The content for the slide
    Returns:
        Confirmation of slide creation
    """
    return f"slide {slideno} created"

memory = MemorySaver()
model = ChatOpenAI(model="gpt-4-turbo-preview", streaming=True)
prompt = ChatPromptTemplate.from_messages([
    ("system", """You are a Presentation Creation Assistant. Your task is to help users create effective presentations.
    Follow these steps:
    1. First use the plan tool to create an outline of the presentation
    2. Then use create_slide tool for each slide in sequence
    3. Guide the user through the presentation creation process
    
    Today's date is {datetime.now().strftime('%Y-%m-%d')}"""),
    ("placeholder", "{messages}"),
])

def state_modifier(state) -> list[BaseMessage]:
    try:
        formatted_prompt = prompt.invoke({
            "messages": state["messages"]
        })
        return trim_messages(
            formatted_prompt,
            token_counter=len,
            max_tokens=16000,
            strategy="last",
            start_on="human",
            include_system=True,
            allow_partial=False,
        )
    except Exception as e:
        print(f"Error in state modifier: {str(e)}")
        return state["messages"]

# Create the agent with presentation tools
agent = create_react_agent(
    model,
    tools=[plan, create_slide],
    checkpointer=memory,
    state_modifier=state_modifier,
)

class ChatInput(BaseModel):
    message: str
    thread_id: Optional[str] = None

@router.post("/chat")
async def chat(input_data: ChatInput):
    thread_id = input_data.thread_id or str(uuid.uuid4())
    
    config = {
        "configurable": {
            "thread_id": thread_id
        }
    }
    
    input_message = HumanMessage(content=input_data.message)
    
    async def generate():
        async for event in agent.astream_events(
            {"messages": [input_message]}, 
            config,
            version="v2"
        ):
            kind = event["event"]
            
            if kind == "on_chat_model_stream":
                content = event["data"]["chunk"].content
                if content:
                    yield f"{json.dumps({'type': 'token', 'content': content})}\n"

            elif kind == "on_tool_start":
                tool_input = event['data'].get('input', '')
                yield f"{json.dumps({'type': 'tool_start', 'tool': event['name'], 'input': tool_input})}\n"
            
            elif kind == "on_tool_end":
                tool_output = event['data'].get('output', '')
                yield f"{json.dumps({'type': 'tool_end', 'tool': event['name'], 'output': tool_output})}\n"

    return EventSourceResponse(
        generate(),
        media_type="text/event-stream"
    )

@app.get("/health")
async def health_check():
    return {"status": "healthy"}

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)