app.py

import gradio as gr
from huggingface_hub import HfApi
import os
import datasets
from datasets import load_dataset, Dataset
import pandas as pd

TOKEN = os.getenv("TOKEN")
DATASET_ID = os.getenv("DATASET_ID")
datasets.set_caching_enabled(False)
hf_api = HfApi(token=TOKEN)

def create_dataset():
    # creates the dataset if it doesn't exist
    try:
        hf_api.create_repo(repo_id = DATASET_ID, repo_type = "dataset", private=True)
        print("created dataset")
        initial_dataset()
    except:
        pass

def initial_dataset():
    # creates initial dataset repository and uploads csv file
    df = pd.DataFrame(columns=["question", "answer"])
    df.loc[len(df)] = ["Find question here", "Find answer here"]  # Add first row
    # persist initial dataset
    df.to_csv("persisted_dataset.csv", index=False)
    # push to hub
    hf_api.upload_file(path_or_fileobj="persisted_dataset.csv",
    path_in_repo = "./persisted_dataset.csv",
    repo_id = DATASET_ID, repo_type="dataset")


def write_to_dataset(input):
    # if the dataset is not created, submit button will crash to we create it
    create_dataset()
    # load the dataset, append input and placeholder
    dataset = load_dataset(DATASET_ID, data_files="persisted_dataset.csv", use_auth_token=TOKEN)
    df = pd.DataFrame(dataset["train"])
    df = df.append({'question': input, "answer":"ANSWER HERE"}, ignore_index=True)
    df = df.drop_duplicates()
    # persist it to local dataset and push it back to Hub
    df.to_csv("persisted_dataset.csv", index=False)
    hf_api.upload_file(path_or_fileobj="persisted_dataset.csv",
    path_in_repo = "./persisted_dataset.csv",
    repo_id = DATASET_ID,
    repo_type="dataset")
    # remove the persisted dataset for privacy
    # as of now we don't persist the dataset anyway so let's see if it will stay without removing
    # os.remove("persisted_dataset.csv")
    return df

def read_dataset():
    # read the dataset for second tab, will crash if it doesn't exist 
    create_dataset()
    dataset = load_dataset(DATASET_ID, data_files="persisted_dataset.csv", use_auth_token=TOKEN, download_mode='force_redownload')
    # read it to a pandas df
    df = pd.DataFrame(dataset["train"])
    # return only answered questions
    return df[df['answer'] != 'ANSWER HERE']

def render_answers():
    # render dataframe into sequentially written text
    # used for textbox in second tab
    df = read_dataset()
    new_df = df.copy()
    new_df = new_df.sort_index(ascending=False) 
    new_df['question'] = '❓ ' + new_df['question']
    
    new_df['answer'] = '🙋 ' + new_df['answer']
    
    str_df = '\n'.join(new_df.apply(lambda x: x['question'] + '\n' + x['answer'] + '\n\n\n', axis=1))
    
    return str_df


with gr.Blocks() as demo:
    gr.Markdown("## Ask me anything, I'm not going to lie! 👋 ")
    gr.Markdown("In this app, you can write me something anonymous and read my answers to your inputs. ✍️ ")
    gr.Markdown("Let's spread love and be respectful 💃🏻🕺🏻")
    with gr.Accordion("Open this toggle to see how you can build your own Ask Me Anything app to receive and answer questions ⬇️", open=False):
        gr.Markdown("Duplicate this Space by clicking three dots and then `Duplicate this Space` and provide TOKEN and DATASET_ID.")
        gr.Markdown("Provide your Hugging Face write token that you can get [here](https://huggingface.co/settings/tokens).")
        gr.Markdown("For DATASET_ID, provide something like `merve/answers` it's a dataset repository that will be created through this Space automatically, where your questions will be privately populated.")
        gr.Markdown("To answer questions, simply edit the `persisted_dataset.csv` file in the dataset repository, edit `ANSWER HERE` parts beside every question.")
        gr.Markdown("The app will not show unanswered questions by default.")
 
    with gr.Tab("Write me something ✍️ "):
        input = gr.Textbox(label="What do you want to say to me?")
        submit_btn = gr.Button("Submit")
        submit_btn.click(fn=write_to_dataset, inputs=[input])
    with gr.Tab("Read My Answers"):
        answers = render_answers()
        gr.Markdown(answers)
        refresh_btn = gr.Button("Refresh")
        refresh_btn.click(fn=render_answers)

demo.launch(debug=True)