• 📖 About the Project
  • 🛠 Built With
    • Tech Stack
    • Key Features
• 💻 Getting Started
  • Prerequisites
  • Setup
  • Install
  • Database & Seeds
  • AI Model Setup (Ollama)
  • Usage
• 👥 Authors
• 🔭 Future Features
• 🤝 Contributing
• ⭐️ Show your support
• 🙏 Acknowledgements
• 📝 License

Oracle API is a technical orchestrator that bridges user requests with local Large Language Models. Built on Rails 8, it utilizes a service-oriented architecture to manage AI queries via Ollama, ensuring a responsive UI through asynchronous job processing and real-time state synchronization.

🔗 Video Presentation: Oracle API (Ruby on Rails + Ollama)

🔗 Frontend Repository: Oracle Chat (Next.js)

• Asynchronous AI Inference: Uses Sidekiq to process heavy AI requests without blocking the main thread.
• Sidekiq-Rails Sync: Real-time synchronization via Redis to update the conversation state once the LLM finishes inference.
• Service-Oriented Design: Dedicated Service Objects to encapsulate the complexity of Ollama API interactions.
• JWT Authentication: Secure and stateless user access control.
• Request Idempotency: Powered by Kredis to prevent duplicate generations during network instability.
• Smart Seeding: Includes predefined System Prompts and test users to accelerate development environment setup.

(back to top)

Follow these steps to set up the backend orchestrator locally.

• Ruby 3.3.0+
• PostgreSQL
• Redis
• Ollama (Installed and running)

Clone the repository:

git clone [https://github.com/Nkaleth/oracle_api.git](https://github.com/Nkaleth/oracle_api.git)
cd oracle_api

Install dependencies:

bundle install

Configure your database.yml and run:

rails db:prepare
rails db:seed

1. Ensure Ollama is running: ollama serve
2. Download the preferred model:

ollama run llama3 # or your specific model

Start the background workers and the server:

# Terminal 1: Background Jobs
bundle exec sidekiq

# Terminal 2: Rails Server
rails s

The API will be available at http://localhost:3000.

(back to top)

👤 Nilton Segura

• GitHub: @Nkaleth
• Twitter: @NoeSeguraL
• LinkedIn: Nilton Segura

(back to top)

• Smart Pagination (Pagy): Activate the pre-installed Pagy gem to handle large message histories efficiently.
• Fine-grained Authorization: Implement Pundit or CanCanCan for secure resource access control (RBAC).
• Vector Search (RAG): Integrate a vector database to provide the AI with local documentation context.
• Multi-Model Support: Dynamic model selection per conversation thread.
• Rate Limiting: Throttling for AI inference endpoints using rack-attack.

(back to top)

Feel free to fork this project and submit pull requests. For major changes, please open an issue first to discuss what you would like to change.

(back to top)

If you find this backend architecture useful, please give it a star! 🌟

(back to top)

• The Rails team for the Rails 8 revolution and the rock-solid Action Cable architecture.
• The Sidekiq community for providing the gold standard in high-performance background processing.
• The Ollama community for making local LLM deployment accessible and efficient for developers.

(back to top)

This project is MIT licensed.

(back to top)