fmeiraf / LLM-budget-assistant

0 stars 0 forks source link

LLM Budgeting Assistant

Live at: 💸 banqstream.com

This is an attempt to create a finance/budgeting assistant. This will be using LLMs to parse raw text coming from account statements and feeding this into a database used for further analysis. Among the features on this app:

Running it locally

The application is all built using docker containers. There are 3 containers, where pdadmin is only used for development.

OPENAI_API_KEY="..."
POSTGRES_PASSWORD="..."
POSTGRES_USER=".."
PGADMIN_DEFAULT_EMAIL="..."
PGADMIN_DEFAULT_PASSWORD="..."
DISCORD_BOT_TOKEN=".."
DISCORD_GUILD_ID=..
OPENAI_MODEL_NAME="gpt-3.5-turbo"
ENVIRONMENT="PROD"

Running on GCP (or other cloud services)

Setting up first installation

Install Docker

Installing docker-compose

Add firewall rule

Making custom domain to work

Instal nginx (for more reference):

Installing SSL:

For updates in the enviroment specs

To make sure new changes in dependencies are correctly added to requirements ( and are easily usable afterwards..):

After new additions run:

pip list --format=freeze > requirements.txt

For fresh enviroment installations:

pip install -r requirements.txt