app.py

"""
Requires:
/usr/local/bin/python3.13 -m pip install python-dotenv pydantic-ai imdbpy
OpenAI API Key
"""

from dataclasses import dataclass
import imdb
import asyncio
from pydantic import BaseModel
from pydantic_ai import Agent, RunContext
import gradio as gr
from pathlib import Path


from dotenv import load_dotenv
import os

script_folder = Path(__file__).parent
dotenv_path = script_folder/"Open_AI.env"
load_dotenv(dotenv_path=dotenv_path)
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")


# A pydantic basemodel for our output schema
class Review(BaseModel):
    why: str
    rating: int
    recommended: bool

# A class to externally connect to the IMDB API to fetch movie information
class IMDbConnection:
    def __init__(self):
        self.ia = imdb.IMDb()

    async def get_movie_info(self,title: str) -> dict:
        # Search for the movie using IMDB API
        print(f"Searching for movie: {title}")
        movies = self.ia.search_movie(title)
        movie = self.ia.get_movie(movies[0].movieID)
        system_prompt = f"""
Title: {movie.get("title")}\n
Rating: {movie.get("rating")}\n
Plot: {movie.get("plot")[0] if movie.get("plot") else None}
"""
        print(f"System prompt: {system_prompt}\n\n")
        return system_prompt

# A container containing dependencies for the agent
@dataclass
class MovieData:
    title: str
    imdb_conn: IMDbConnection

agent = Agent(
    "openai:gpt-4o-mini",
    deps_type=MovieData,
    output_type=Review,
)

# Dynamically generate the system prompt
@agent.system_prompt
async def get_movie_info(ctx: RunContext[Review]):

    # Given an input, the system prompt will be generated by querying the IMDB API
    return await ctx.deps.imdb_conn.get_movie_info(ctx.deps.title) #Fetch movie info

# create a gradio wrapper

async def run_agent(user_query: str, movie_title: str):
    deps = MovieData(title=movie_title, imdb_conn=IMDbConnection())
    result = await agent.run(user_query, deps=deps)
    return {
        "Why": result.output.why,
        "Rating": result.output.rating,
        "Recommend": result.output.recommended
    }

#run agent in gradio
def run_gradio(user_query, movie_title):
    return asyncio.run(run_agent(user_query, movie_title))

# build gradio ui
with gr.Blocks() as demo:
    gr.Markdown("## Movie Recommender Agent")
    user_query = gr.Textbox(label="What is your preference? (e.g. 'I like supernatural horror movies')")
    movie_title = gr.Textbox(label="Movie name (e.g. 'The Conjuring')")
    output = gr.JSON(label = "Agent Recommendation")

    submit_btn = gr.Button("Check Movie")
    submit_btn.click(fn=run_gradio, inputs=[user_query, movie_title], outputs=output)

#launch
if __name__ == "__main__":
    demo.launch(share=True)