main.py

import gradio
import numpy
import pandas
import sentence_transformers
import datasets
import faiss

model = sentence_transformers.SentenceTransformer('allenai-specter')

data = datasets.load_dataset("ccm/publications")['train'].to_pandas()

dimensionality = len(data['embedding'][0])
index = faiss.IndexFlatL2(dimensionality) 

vectors = numpy.stack(data['embedding'].to_list(), axis=0)

index.add(vectors)

def search(query):
    k=5
    query = numpy.expand_dims(model.encode(query), axis=0)
    _, I = top_five = index.search(query, k)
    top_five = data.loc[I[0]]
    search_results = ""

    for i in range(k):
        search_results += str(i+1) + ". "
        search_results += '"' + top_five["bibtex"].values[i]["title"] + '" '
        search_results += top_five["bibtex"].values[i]["citation"]
        if top_five["pub_url"].values[i] is not None:
            search_results += " [Full Paper](" + top_five["pub_url"].values[i] + ")"
        search_results += "\n"
    return search_results

with gradio.Blocks() as demo:
    query = gradio.Textbox(placeholder="Enter search terms...")
    btn = gradio.Button("Search")
    results = gradio.Markdown()
    btn.click(fn=search, inputs=[query], outputs=results)

demo.launch(debug=True)