Skip to content

Latest commit

 

History

History
65 lines (45 loc) · 1.49 KB

README.md

File metadata and controls

65 lines (45 loc) · 1.49 KB
Raw

Chainstack AI Telegram bot

AI-powered Telegram bot all knowing about Chainstack docs. It uses GPT 3.5/4 and context stored using Actvieloop's Deep Lake as a vector database.

  • index.py is run to index the data from the docs.
  • main.py is the bot script itself.

Quickstart

  • Clone the repo
git clone https://github.com/soos3d/chainstack-tg-ai-bot.git
  • Create a new Python virtual environment
python3 -m venv chainstack-ai-tg
  • Activate the virtual environment
source chainstack-ai-tg/bin/activate
  • Install dependencies
pip install -r requirements.txt
  • Add Telegram bot token and username, API keys, and Deep Lake account path to .env.example and rename .env
# Bot config
CHAINSTACK_BOT_TOKEN="BOT_TOKEN_FROM_BOTFATHER" # Chainstack_AI_bot token
CHAINSTACK_BOT_USERNAME="@USERNAME"

# OpenAI
OPENAI_API_KEY="OPENAI_KEY"
EMBEDDINGS_MODEL="text-embedding-ada-002"
LANGUAGE_MODEL="gpt-3.5-turbo" # gpt-4 gpt-3.5-turbo-0613

# Scrape settings
SITE_MAP="https://docs.chainstack.com/sitemap.xml"
URLS_FILTER="https://docs.chainstack.com/reference/"

# Deeplake vector DB
ACTIVELOOP_TOKEN="ACTIVELOOP_TOKEN"
DATASET_PATH="hub://USER_ID/custom_dataset" # "./local_chainstack_docs_db"  # Edit with your user ID if you want to use the cloud db. or use the `./` for a local instance.
  • Run the index file to scrape the docs and store them
python3 index.py
  • Run the main file to start the bot
python3 main.py