app.py

import os
from dotenv import load_dotenv
import httpx
import gradio as gr
from langchain.prompts import PromptTemplate
from langchain_huggingface import HuggingFaceEndpoint
from langchain_core.messages import BaseMessage, HumanMessage
from langgraph.graph import MessageGraph, END
from typing import Sequence

# Load environment variables
load_dotenv()
HF_TOKEN = os.getenv("HF_TOKEN")
WEATHER_TOKEN = os.getenv("WEATHER_TOKEN")

# Initialize the HuggingFace inference endpoint
llm = HuggingFaceEndpoint(
    repo_id="mistralai/Mistral-7B-Instruct-v0.3",
    huggingfacehub_api_token=HF_TOKEN.strip(),
    temperature=0.7,
    max_new_tokens=200
)

# Define nodes
def fetch_weather_node(city: str) -> str:
    url = f"https://api.openweathermap.org/data/2.5/weather?q={city}&appid={WEATHER_TOKEN}&units=metric"
    
    try:
        response = httpx.get(url)
        response.raise_for_status()
        weather_data = response.json()
        weather = weather_data['weather'][0]['main']
        temperature = weather_data['main']['temp']
        return f"The current weather in {city} is {weather} with a temperature of {temperature}°C."
    except Exception as e:
        return f"Error: {e}"

def generate_review_node(weather_info: str) -> str:
    response = llm(weather_info)
    return response

# Define the prompt template for generating weather reviews
review_prompt_template = """
You are an expert weather analyst. Based on the provided weather information, generate a detailed and insightful review. 
Weather Information: {weather_info}
Your review should include an analysis of the weather conditions and finish in 150 words. 
Review:
"""

# Create and configure the graph
builder = MessageGraph()

# Add nodes
builder.add_node("fetch_weather", fetch_weather_node)
builder.add_node("generate_review", generate_review_node)
builder.set_entry_point("fetch_weather")

# Define transitions
builder.add_edge("fetch_weather", "generate_review")
builder.set_finish_point("generate_review")

# Compile the graph
graph = builder.compile()

# Define the Gradio interface
def get_weather_and_review(city: str) -> str:
    if city:
        try:
            # Prepare the input for the graph
            weather_info = graph.invoke(HumanMessage(content=city))
            weather_info_text = weather_info[1].content
            
            # Generate the review using the refined prompt
            review_input = review_prompt_template.format(weather_info=weather_info_text)
            review = graph.invoke(HumanMessage(content=review_input))
            review_text = review[2].content

            return f"**Weather Information:**\n{weather_info_text}\n\n**AI Generated Weather Review:**\n{review_text}"

        except Exception as e:
            return f"Error generating weather review: {e}"
    else:
        return "Please enter a city name."

interface = gr.Interface(
    fn=get_weather_and_review,
    inputs=gr.Textbox(lines=2, placeholder="Enter the name of a city:", label="City"),
    outputs="text",
    title="City Weather Information with AI Review",
    description="Enter the name of a city to get current weather information and an AI-generated review based on that information."
)

if __name__ == "__main__":
    interface.launch()