AI prompt processing Worker

A simple worker which fetches a prompt job every 2s and processes it through an AI endpoint.

Tested only with LM Studio model API

This is a multirepo project, and depends on the Async Queue.

Features

• Async: Process jobs from a queue.
• Multiple instance There's no limit on the amount of simultaneous active workers, each will fetch a different Job.
• Private/Public Each worker can be assigned to users(Giving priority), once there's no exclusive jobs for the worker... It will fetch non assigned prompts(Public)

Instalation and configuration

git clone https://github.com/Angel-del-dev/Async-AI-Prompt-Local-Worker.git async-ai-worker
cd async-ai-worker
cp .env-example .env
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

--psql
-- Create a worker
|> insert into workers(created_at) values (current_timestamp);

Structure and execution

Dependencies

• sqlalchemy
• pandas
• python-dotenv
• psycopg2-binary
• requests

Worker startup

(venv) |> python3 main.py

Prerequisites

• Python (Version3.13.5)
• PostgreSQL engine.
• LM Studio (Either GUI or headless)
• AI Model (Installed via LM Studio)

Notes

LM Studio API seems to give better results when Enable thinking is disabled in inference/Custom Fields

Environment variables (.env)

Project startup depends on .env variables.

Var	Description	Type	Required	Domain
DB_HOST	Database host.	string	true	DB
DB_NAME	Database name.	string	true	DB
DB_USER	Database username.	string	true	DB
DB_PASSWORD	Database password.	string	true	DB
DB_PORT	Database port.	int	true	DB
WORKER	UUID obtained when inserting the new worker in the DB.	string	true	ENV
MODEL	Model code given by LM Studio(Ex: qwen/qwen3.5-9b, google/gemma-4-e4b).	string	true	AI
MODEL_URL	LM Studio local API Url.	string	true	LM_STUDIO