app.py

import gradio as gr
import faiss
import numpy as np
from sentence_transformers import SentenceTransformer

# Load the model once
model = SentenceTransformer('all-MiniLM-L6-v2')

# Global storage for documents and index
global_docs = []
global_index = None

# Load documents from uploaded file
def load_documents(file_obj):
    docs = [line.strip() for line in file_obj if line.strip()]
    return docs

# Build FAISS index
def build_index(docs):
    embeddings = model.encode(docs).astype(np.float32)
    index = faiss.IndexFlatL2(embeddings.shape[1])
    index.add(embeddings)
    return index, embeddings

# Semantic search
def semantic_search(query, top_k=3):
    if not global_index or not global_docs:
        return "Please upload a file first."
    query_embedding = model.encode([query]).astype(np.float32)
    distances, indices = global_index.search(query_embedding, top_k)
    results = [
        f"Rank {rank + 1}:\nDocument: {global_docs[i]}\nL2 Distance: {distances[0][rank]:.4f}\n"
        for rank, i in enumerate(indices[0])
    ]
    return "\n".join(results)

# Handle file upload
def upload_and_index(file):
    global global_docs, global_index
    contents = file.read().decode("utf-8").splitlines()
    global_docs = [line.strip() for line in contents if line.strip()]
    global_index, _ = build_index(global_docs)
    return "Document indexed successfully!"

# Gradio Interface
with gr.Blocks() as demo:
    gr.Markdown("## 🔍 Semantic Search in Academic Papers")
    
    file_input = gr.File(label="Upload Academic Paper (.txt)", file_types=['.txt'])
    upload_button = gr.Button("Upload & Index")
    upload_output = gr.Textbox(label="Status")

    query_input = gr.Textbox(label="Enter Search Query")
    search_button = gr.Button("Search")
    search_output = gr.Textbox(label="Top 3 Results")

    upload_button.click(upload_and_index, inputs=file_input, outputs=upload_output)
    search_button.click(semantic_search, inputs=query_input, outputs=search_output)

demo.launch()