LLM Slack Chat Bot

Functions

Agentic work flow by Langgraph
RAG for the personal data
Short-term Context Retention
Multi modal (image, text only for now)
use OpenAI model
permission for each user
custom prompt for each user
fetch youtube transcription by Tor proxy

Usage

ask the bot on DM or by tagging the bot
whenever LLM model create next token, it's updated on the answer message in realtime
if you ask again on the thread of the slack message, chatbot keep the short-term context
attach image if you want to ask about the image

Requirements

python3.11
install requirements.txt
create slack app and add proper permissions
set .env file by referring to .env_sample

Slack configuration

Run

python chatbot.py

Deploy to server

configure deploy/config.sh and run the below this requires docker and docker-compose on the server. tested on ec2 server

cd deploy
sh deploy_server.sh

Plan

local chatbot for device automation
personal AI assistant
company chatbot (not updated here)
RAG for source code. let user can decide scope like cursor IDE

Reference