Docker Local Agents

A production-ready, containerized FastAPI application with background task processing, AI model serving, and persistent storage. This project leverages Docker Compose to orchestrate a modern stack for scalable, automated workflows.

---

Features

• FastAPI: High-performance Python web API.
• Celery: Distributed task queue for background and scheduled jobs.
• PostgreSQL: Reliable relational database.
• Redis: Fast in-memory broker for Celery.
• Ollama: Local AI model server for LLM inference.
• CrewAI: Agentic workflow orchestration (via crewai and crewai-tools).
• Automated Model Management: Scripts for managing Ollama models.

---

Architecture

[FastAPI] <--> [PostgreSQL]
     |              ^
     v              |
 [Celery Worker]    |
     |              |
     v              |
   [Redis] <--------+
     |
     v
 [Ollama]

• FastAPI serves as the main API and interacts with the database and Ollama.
• Celery Worker runs background and scheduled tasks, including database checks and agentic workflows.
• Ollama provides local LLM inference, managed via helper scripts.
• Celery Beat schedules periodic tasks (e.g., database checks every minute).

---

Getting Started

Prerequisites

• Docker Desktop
• Git

Quickstart

1. Clone the repository

   git clone <your-repo-url>
   cd <your-repo-directory>

2. Build and start all services

   docker compose up --build

3. (Optional) Pull a default Ollama model

   docker exec ollama ollama pull qwen3:1.7b

4. Access the API

   • FastAPI: http://localhost:8000
   • Ollama: http://localhost:11434

5. Stop all services

   docker compose down

---

Service Details

FastAPI

• Main entrypoint: app/main.py
• Auto-reloads on code changes (if configured)
• Connects to PostgreSQL and Redis

Celery

• Worker and Beat services for background and scheduled tasks
• Example tasks: add, multiply, periodic DB check with agentic workflow

Ollama

• Local LLM server, managed via ollama-commands.sh and startup-ollama.sh
• Pull, run, and manage models interactively

PostgreSQL

• User: postgres
• Password: postgres
• Database: tododb
• Data persisted in Docker volume

Redis

• Used as broker and result backend for Celery

---

Model Management

Use the provided scripts to manage Ollama models:

• List models

  ./ollama-commands.sh list

• Pull a model

  ./ollama-commands.sh pull <model-name>

• Run a model interactively

  ./ollama-commands.sh run <model-name>

• Remove a model

  ./ollama-commands.sh remove <model-name>

• Show logs

  ./ollama-commands.sh logs

---

Environment Variables

Set in docker-compose.yml:

• CELERY_BROKER_URL and CELERY_RESULT_BACKEND: Redis connection
• DATABASE_URL: PostgreSQL connection
• OLLAMA_HOST and OLLAMA_PORT: Ollama service

---

Development

• Python dependencies in requirements.txt
• Dockerfile for reproducible builds
• All code in the app/ directory (see main.py, models.py, database.py, and agentic logic in todos_crew/)

---

Example API Usage

• Add your FastAPI endpoints in app/main.py
• Celery tasks in tasks.py (e.g., add_task, multiply_task, check_database_task)
• Scheduled DB checks trigger agentic workflows using CrewAI

---

Notes

• Data for PostgreSQL and Ollama is persisted in Docker volumes.
• The system is designed for local development and experimentation with agentic workflows and LLMs.
• You can extend the agent logic in app/todos_crew/.

---