Jarvis AI Platform

A Modular, Local-First, Open-Source AI Assistant Platform

Built with Java 21, Spring Boot 4, and Spring AI 2.0

---

---

---

🚀 What Is Jarvis?

Jarvis is not just a chatbot.

It is a modular AI orchestration platform that runs entirely on your own machine with complete privacy.

Your AI. Your Data. Your Machine.

✨ Key Differences from ChatGPT

• Your conversations never leave your computer
• Completely free (runs on Ollama locally)
• Remembers you across sessions (Phase 2 ✅)
• Searches your documents (Phase 3 ✅)
• Uses real-world tools (Phase 4 ✅)
• Speaks and listens (Phase 5 ✅)
• Plans and executes multi-step tasks (Phase 6 ✅)

---

📈 Current Status

Phase	Version	Feature	Status
Phase 1	v0.1.0	AI Chat + CLI + JWT	✅ Released
Phase 2	v0.2.0	Memory System	✅ Core Complete
Phase 3	v0.3.0	RAG Engine	✅ Core Complete
Phase 4	v0.4.0	Tool Engine	✅ Core Complete
Phase 5	v0.5.0	Voice Assistant	✅ Core Complete
Phase 6	v0.6.0	Agent System	✅ Core Complete
Phase 7	v1.0.0	Web UI	🔨 In Progress

---

✅ What Works Right Now

Feature	Status	Since
AI chat via CLI (streaming)	✅	v0.1.0
JWT authentication (Argon2id)	✅	v0.1.0
Session management + persistence	✅	v0.1.0
PostgreSQL (all messages saved)	✅	v0.1.0
Ollama local AI (primary)	✅	v0.1.0
Gemini cloud AI (fallback)	✅	v0.1.0
Provider abstraction	✅	v0.1.0
Redis session caching	✅	v0.2.0
Long-term memory storage	✅	v0.2.0
Automatic memory extraction	✅	v0.2.0
Memory injection into prompts	✅	v0.2.0
pgvector semantic memory search	✅	v0.2.0
Memory REST API	✅	v0.2.0
CLI memory commands	✅	v0.2.0
Document upload + processing	✅	v0.3.0
Document chunking + embedding	✅	v0.3.0
RAG semantic search	✅	v0.3.0
RAG context injection into prompt	✅	v0.3.0
Document REST API	✅	v0.3.0
DateTimeTool (timezone support)	✅	v0.4.0
CalculatorTool (exact math)	✅	v0.4.0
WeatherTool (OpenWeatherMap)	✅	v0.4.0
WebSearchTool (DuckDuckGo)	✅	v0.4.0
MCP Server (external clients)	✅	v0.4.0
Voice transcription (Whisper)	✅	v0.5.0
Text-to-speech (cross-platform)	✅	v0.5.0
Voice chat loop	✅	v0.5.0
Voice selection + speed control	✅	v0.5.0
GET /api/v1/settings	✅	v0.5.0
PATCH /api/v1/settings/voice	✅	v0.5.0
GET /api/v1/documents/{id}/status	✅	v0.5.0
ReACT agent system	✅	v0.6.0
Multi-step task execution	✅	v0.6.0
Agent streaming (SSE)	✅	v0.6.0
Agent REST API	✅	v0.6.0
CLI document commands	📋	v0.3.0
CLI tool commands	📋	v0.4.0
CLI voice commands	📋	v0.5.0
CLI agent commands	📋	v0.6.0

---

⚡ Quick Start

Prerequisites

Tool	Version	Purpose
Java	21+	Runtime
Docker	Latest	PostgreSQL + Redis
Ollama	Latest	Local AI Models

Step 1 — Clone Repository

git clone https://github.com/sujankim/jarvis-ai-platform.git
cd jarvis-ai-platform

Step 2 — Pull AI Models

# Chat model (~5GB)
ollama pull llama3.1:8b

# Embedding model (~274MB)
ollama pull nomic-embed-text

Step 3 — Configure Environment

cp .env.example .env

Edit .env:

JARVIS_JWT_SECRET=your-super-secret-key-at-least-32-characters

OPENWEATHER_API_KEY=your_openweather_key

GEMINI_API_KEY=your_gemini_key

Step 4 — Start Infrastructure

docker compose up -d

This starts:

• PostgreSQL 16 + pgvector
• Redis 7

Step 5 — Run Jarvis

cd server
./mvnw spring-boot:run

Step 6 — First-Time Setup

jarvis:> setup
jarvis:> login
jarvis:> chat

---

💻 CLI Commands

Authentication

login
logout
whoami
setup

Chat

chat
chat --new
ask -m "Hello Jarvis"

Memory

memory list
memory add
memory delete --number 3
memory clear

Sessions

session
new-session
switch-session -n 2

Tools (Phase 4)

tools
tool-test --tool datetime

System

status
doctor
jarvis-version
about
benchmark-latency --provider ollama --runs 10

---

🏗️ Architecture

Spring Shell CLI (jarvis:> prompt)
           │
  Spring Boot 4 AI Engine
           │
  ┌────────┴────────────────────┐
  │                             │
AiOrchestrator            Memory System
                          (Phase 2 ✅)
  │
PromptAssembler
  │
  ├── Working Memory
  ├── Long-term Memories (pgvector)
  ├── RAG Context (pgvector)
  └── Session History (Redis)
           │
    ProviderRouter
    │             │
OllamaProvider  GeminiProvider
(local)         (cloud fallback)
    │
  Tools (Phase 4 ✅)
  ├── DateTimeTool
  ├── CalculatorTool
  ├── WeatherTool
  └── WebSearchTool
    │
MCP Server → External Clients
    │
AgentOrchestrator (Phase 6 ✅)
  ├── AgentPlanner (ReACT)
  ├── AgentExecutor
  └── ToolRegistry
           │
PostgreSQL + pgvector     Redis

---

🛠️ Tech Stack

Layer	Technology
Language	Java 21 (LTS)
Framework	Spring Boot 4.0.6
AI Framework	Spring AI 2.0.0-M8
Web	Spring WebFlux
Security	Spring Security 7
Authentication	JWT (Argon2id)
Database	PostgreSQL 16
Vector Database	pgvector 0.7.4
Data Access	R2DBC + JDBC
Cache	Redis 7
Migrations	Flyway (V1–V16)
Local AI	Ollama
Chat Model	llama3.1:8b
Embeddings	nomic-embed-text
Cloud AI	Google Gemini
Tools Protocol	MCP (Spring AI)
CLI	Spring Shell 4
Mapping	MapStruct 1.6
API Docs	SpringDoc OpenAPI 3

---

🤝 Contributing

We welcome all contributions.

Good First Issues

Issue	Description	Phase
#3	Token count display	Phase 1
#4	Examples command	Phase 1
#6	Docker image	Phase 1
#11	Rate limiting	Phase 2
#51	CLI document commands	Phase 3
#52	PDF text extraction	Phase 3
#66	CLI tool commands	Phase 4
#69	CLI voice commands	Phase 5
#78	Voice integration tests	Phase 5
#84	CLI agent commands	Phase 6

See CONTRIBUTING.md for the setup guide.

---

📝 Articles

• Building a Local-First AI Assistant with Spring Boot 4
• Building Long-Term Memory with pgvector
• Implementing Semantic Memory Retrieval
• Building a Tool Engine with Spring AI
• Adding Voice with Whisper and Text-to-Speech
• Building an AI Agent System with the ReACT Pattern

---

🔒 Privacy

• No telemetry by default
• Ollama runs 100% locally
• Conversations never leave your machine
• All embeddings are stored locally in PostgreSQL
• MCP Server runs locally — no cloud dependency

---

📄 License

Licensed under the Apache License 2.0.

See LICENSE for details.

---

Built by Sujan and the Open Source Community ❤️

⭐ Star this repository if Jarvis helps you!