import gradio as gr
import pickle
import pandas as pd
from sklearn.metrics.pairwise import cosine_similarity

# Load model and dataset
with open("recommender_model.pkl", "rb") as f:
    model = pickle.load(f)

posts_df = pd.read_csv("posts_cleaned.csv")  # your full dataset with post content
post_embeddings = model["embeddings"]  # precomputed post embeddings
vectorizer = model["vectorizer"]      # for transforming user input

# Predict function
def recommend_from_input(user_text):
    user_vec = vectorizer.encode([user_text])
    sims = cosine_similarity(user_vec, post_embeddings)[0]
    top_idxs = sims.argsort()[-5:][::-1]
    top_posts = posts_df.iloc[top_idxs]["post_text"].tolist()
    return "\n\n".join(top_posts)

# Gradio UI
interface = gr.Interface(
    fn=recommend_from_input,
    inputs="text",
    outputs="text",
    title="AI Content Recommender",
    description="Enter a sample interest or post to receive recommendations"
)

interface.launch()