app.py

# -*- coding: utf-8 -*-

# Install Cohere for embeddings

import cohere
import numpy as np
import pandas as pd
import gradio as gr
import os
from sklearn.metrics.pairwise import cosine_similarity
from annoy import AnnoyIndex
import warnings
warnings.filterwarnings('ignore')
pd.set_option('display.max_colwidth', None)

data_df = pd.read_csv('functions_data.csv')


data_df['docstring'].fillna('not specified', inplace=True)

# Paste your API key here. Remember to not share publicly
key = os.environ.get('API_KEY')
api_key = key

# Create and retrieve a Cohere API key from dashboard.cohere.ai/welcome/register
co = cohere.Client(api_key)


search_index = AnnoyIndex(4096, 'angular')
search_index.load('code.ann') # super fast, will just mmap the file

def get_code(query):
    # Get the query's embedding
    query_embed = co.embed(texts=[query],
                    model="large",
                    truncate="LEFT").embeddings

    # Retrieve the nearest neighbors
    similar_item_ids = search_index.get_nns_by_vector(query_embed[0],1)
    
   
    return data_df.iloc[similar_item_ids[0]]['function_body'], data_df.iloc[similar_item_ids[0]]['file_path']
examples = ['compute diffusion of given data']
inputs = gr.Textbox(label='query')
outputs = [gr.Textbox(label='matched function'), gr.Textbox(label='File path')]
title = "Search Code"
description = "Semantically search codebase using Cohere embed API. This demo uses Open AI point cloud codebase https://github.com/openai/point-e as an example"
iface = gr.Interface(fn=get_code, inputs=inputs, outputs=outputs, description = description, examples=examples, title=title)
iface.launch()