J.A.R.V.I.S. (Just A Retrieval-Augmented Virtual Intelligent System)

J.A.R.V.I.S. is an intelligent assistant designed to leverage Retrieval-Augmented Generation (RAG) for delivering accurate and context-aware responses. This project integrates technologies like Spring AI, Llama3, Vaadin, and PGvector to build a scalable, efficient, and secure AI-powered assistant.

---

Features

• Responsive Chat UI: Built with Vaadin for a modern and user-friendly interface.
• Retrieval-Augmented Generation (RAG): Combines contextual information and memory for precise responses.
• Private Language Model Hosting: Powered by Llama3 hosted via Ollama.
• Vector Search: Utilizes PostgreSQL with PGvector for fast and accurate similarity searches.
• Structured Ingestion Pipeline: Processes and stores documents with Spring AI ETL tools.
• Prompt Engineering: Implements robust system and user prompts for consistent and accurate interactions.

---

Tech Stack

• Frontend: Vaadin
• Backend: Java, Spring AI
• Language Model: Llama3 (via Ollama)
• Database: PostgreSQL with PGvector
• Containerization: Docker Compose
• ETL Tools: Spring AI components

---

Getting Started

Prerequisites

1. Docker and Docker Compose
2. Java 23 or later
3. PostgreSQL with the PGvector extension
4. Maven for building the project

---

Installation

1. Clone the Repository

   git clone https://github.com/the-sloth-dev/jarvis.git
   cd jarvis

2. Start PostgreSQL with PGvector

   docker-compose up -d pgvector

3. Start Ollama with Llama3

   docker-compose up -d ollama

   Pull the Llama3 model:

   docker exec -it ollama ollama pull llama3

4. Configure the Application Update the application.yml file with your environment variables:

   spring:
     datasource:
       url: jdbc:postgresql://localhost:5433/postgres
       username: postgres
       password: postgres
     ai:
       ollama:
         base-url: http://localhost:11434
       vectorstore:
         pgvector:
           index-type: HNSW
           distance-type: COSINE_DISTANCE
           dimensions: 768

5. Build and Run the Application

   mvn clean install
   java -jar target/jarvis-0.0.1-SNAPSHOT.jar

6. Access the Application Open http://localhost:8080 in your browser.

---

Usage

Trigger Document Ingestion

Place your documents in the directory specified in the configuration (/tmp/jarvis/documents by default) and trigger ingestion using the API:

curl -X POST http://localhost:8080/ingest/run

Chat with the Assistant

Use the chat interface to ask questions based on the ingested documents.

---

Key Components

System Prompt

Defines the assistant's behavior and rules for context and memory usage:

You are an intelligent assistant specialized in retrieval-augmented tasks.
Your primary goal is to provide accurate, concise, and relevant answers strictly based on the provided context and memory.

User Prompt

Guides each query to ensure responses adhere to system rules:

<context>
    {question_answer_context}
</context>
User question: {user_question}

---

Contributing

We welcome contributions to improve J.A.R.V.I.S.! Please submit pull requests or report issues via GitHub.

---

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Author: The Sloth Dev
"Building smarter systems, one step at a time!"