Botastic

March 27, 2023 ยท View on GitHub

An AI framework for building cool things.

๐Ÿ‘€ Overview

architecture

๐Ÿ’ก Motivation

  • LLMs are fantastic for knowledge generation and reasoning as they're pre-trained on large amounts of publicly available data.
  • In practice, we need LLMs to be able to reason about our own data. Fine-tuning them is often impractical. Embedding data into LLMs is a viable alternative, but requires domain expertise, and considerable effort to dealing with prompt limitations
  • Botastic aims to make it easy to embed your own data into LLMs.
  • Botastic also aims to provide out-of-the-box solutions to satisfy common use cases.
  • It would to be easily integrated into existing systems and scalable to meet business requirements.
  • We're building AI-based applications, we need Botastic to simplify the process.

๐Ÿš€ Features

  • ๐Ÿ”Œ Connect to OpenAI's LLM
  • ๐Ÿ’พ Store and manage your own data as LLM's external brain
  • ๐ŸŒ Restful APIs
  • ๐Ÿ’ฌ Conversations management to help you build dialogue interface based applications
  • ๐Ÿค– Bot/Avatar management and builtin useful bots: emotion recognition, sentiment analysis, spam checker, etc. Please refer to [TBD].
  • ๐Ÿงฉ Customize Middlewares to extend Botastic's functionality
  • ๐Ÿ’พ Store vectors in Postgres and other databases
  • ๐Ÿ“ฆ Better strategies to split the data and build text chunks
  • ๐Ÿ“ค Export data
  • ๐Ÿ™‹โ€โ™‚๏ธ Support other LLMs
  • ๐Ÿ“š Fine-tune own models

Getting Started

  1. visit https://developers.pando.im/console/botastic to get your API key/secret, and create a new bot.
  2. run the following command to talk to Botastic
curl --location --request POST 'https://botastic-api.pando.im/api/conversations/oneway' \
--header 'X-BOTASTIC-APPID: YOUR_BOTASTIC_APP_ID' \
--header 'Content-Type: application/json' \
--data-raw '{
  "bot_id": YOUR_BOT_ID,
  "content": "How do you view the future of blockchain, and what benefits does it have for human being? Respond as short as possible like a Zen Master.",
  "category": "plain-text"
}'

Documentation

Please refer to Guide and API Reference for more details.

Showcases

Installation

Manually

Please refer to docs/install.md for manual installation.

Docker

TBD