AzimMuradov / mini-oracle

Mini-oracle can answer your questions. It's a full-stack machine-learning application that uses Hugging Face, Streamlit, Telegram Bot API, and Railway.

mini-oracle

Mini-oracle can give you answers for two types of questions: Who is _? and What is _?.

It's a full-stack machine-learning application:

• backend:
  • hugging-face bert-large-uncased-whole-word-masking-finetuned-squad model combined with wikipedia REST API
• frontends:
  • streamlit framework, you can use the application here - mini-oracle.streamlit.app
  • telegram bot API, you can use the bot here - @mini_oracle_bot

How to use?

Streamlit (web)

If you prefer a web interface, use the streamlit application.

1. Choose the type of the question:

2. Write the question content:

3. Press the submit button and get results:

Telegram Bot

If you prefer a telegram bot interface, use the @mini_oracle_bot.

1. Start the bot:

2. Execute the /who somebody or /what something commands:

If you need any help, you can always execute the /help command:

What to do if I got an error?

There are several types of the errors you may get:

Empty query

You should not leave the question content empty.

Other errors

This is the most certainly not your fault, you should probably wait (1-3 minutes) and try again. If it doesn't help (even after several tries), you should let me know.

How to build from scratch?

Backend - python + hugging-face + flask + wikipedia API

1. Acquire Hugging Face User Access Token (https://huggingface.co/docs/hub/security-tokens) and set it to the HUGGING_FACE_BEARER_TOKEN environment variable.
2. Acquire Wikimedia Personal API token (https://api.wikimedia.org/wiki/Documentation/Getting_started) and set it to the WIKI_BEARER_TOKEN environment variable.
3. Set your email (that you used when registering on the Wikimedia API Portal) to the WIKI_EMAIL environment variable.
4. Go to backend directory.
5. Install dependencies: pip install -r requirements.txt.
6. Run gunicorn -w 2 backend.src.main:app. Follow the logs to access your server.

Remember, there are only one active endpoint: <server_address>/answers?question_type=<question_type>&question_content=<question_content>.

Frontend (WEB) - python + streamlit

1. Go to frontend-streamlit directory.
2. Optionally change the backend API endpoint (by default it is set to mini-oracle.up.railway.app).
3. Install dependencies: pip install -r requirements.txt.
4. Run streamlit run frontend_streamlit/src/main.py. Follow the logs to access your site.

Frontend (Telegram Bot) - kotlin + telegram bot API

1. Create your bot using @BotFather bot (https://core.telegram.org/bots#how-do-i-create-a-bot) and set the token of your newly created bot to the TELEGRAM_BOT_TOKEN environment variable.
2. Go to frontend-telegram-bot directory.
3. Optionally change the backend API endpoint (by default it is set to mini-oracle.up.railway.app).
4. Run ./gradlew run (or .\gradlew.bat run for Windows). Your bot must be up and running.