🎯 Open Source • Local AI • Privacy First

PROJECT-JASP
Your Local AI Assistant

Runs fully on your machine — voice or text, with memory and smart routing. 100% private, 100% local.

2.8k+ Stars
Active Development
MIT License

How It Works

Simple 3-step process to get intelligent AI assistance

1

You Speak or Type

Talk to V-JASP or type to T-JASP. Upload images for analysis. It's that simple.

2

AI Understands

JASP analyzes your request, retrieves relevant memories, and selects the perfect AI model for the job.

3

Get Smart Response

Receive intelligent, context-aware responses. JASP learns from each conversation to serve you better.

100% Private

Everything runs locally on your device

Lightning Fast

No network delays, instant responses

Gets Smarter

Learns your preferences over time

Core Features

Powerful capabilities that make PROJECT-JASP your ideal local AI assistant

100% Local Processing

All AI models run directly on your device. Your data never leaves your machine, ensuring complete privacy and security.

Local processing means no internet connection required for AI operations, zero data transmission to external servers, and complete control over your computational resources.

Multi-Model Architecture

Three specialized AI models working together: Instruction for routing, LLM for chat, and VLM for vision tasks.

The instruction model analyzes your request and routes it to the appropriate specialized model, ensuring optimal performance for text, voice, and image processing tasks.

Advanced Memory System

Short-term and long-term memory capabilities allow JASP to remember context and learn from interactions.

Our memory system includes conversational context retention, preference learning, and knowledge accumulation across sessions while maintaining your privacy through local storage.

Voice Input & Output

Natural voice conversations with high-quality speech recognition and synthesis.

V-JASP includes real-time speech recognition, natural voice synthesis, and hands-free operation capabilities for seamless voice interactions.

Image Analysis

Upload and analyze images with advanced vision model capabilities for visual understanding.

The vision model can analyze uploaded images, extract text, identify objects, and provide detailed descriptions of visual content. Available in both V-JASP and T-JASP.

Smart Request Routing

Intelligent system automatically routes requests to the most appropriate AI model.

The routing model analyzes request types, content, and context to automatically select the best model for optimal performance and accuracy.

Choose Your Version

Two powerful variants designed for different use cases

V-JASP

Voice Assistant

Complete voice-controlled AI assistant with hands-free operation and speech synthesis.

Voice Input & Output
Screenshot Capture
Hands-free Operation
Text Chat Interface
Download V-JASP

T-JASP

Text Assistant

Clean text-based chat interface for focused AI conversations with image upload capabilities.

Text Chat Interface
Image Upload & Analysis
Fast & Lightweight
All Memory Features
Download T-JASP

AI Model Architecture

Specialized models optimized for different tasks

Instruction Model

Analyzes requests and determines the best processing approach

Technical Specifications:

  • • Model: Qwen3-4B-Instruct-2507
  • • Parameters: 4 Billion
  • • Purpose: Request routing and analysis
  • • Specialization: Intent detection

Language Model

Handles text generation and conversational responses

Technical Specifications:

  • • Model: DeepSeek-V3-Base
  • • Parameters: 21 Billion
  • • Purpose: Text generation
  • • Specialization: Conversational AI

Vision Model

Processes images and screenshots for visual understanding

Technical Specifications:

  • • Model: Qwen2.5-VL-7B-Instruct
  • • Parameters: 7 Billion
  • • Purpose: Visual processing
  • • Specialization: Image analysis

Semantic Memory Architecture

Advanced vector database system with intelligent semantic understanding

JSON Vector Database

Semantic memory stored in efficient JSON format with vector embeddings for intelligent retrieval.

JASP-Embed-7B Model

Advanced embedding model converts conversations and knowledge into semantic vectors for context-aware retrieval.

Semantic Similarity Search

Intelligent content retrieval using cosine similarity to find contextually relevant memories.

Local Vector Processing

All vector computations and semantic searches run locally on your device for maximum privacy.

Semantic Memory Flow

How memories are processed and retrieved

Input → Embedding JASP-Embed-7B
Text/conversation converted to 768-dimensional vectors
JSON Vector Store Local Storage
Semantic vectors stored in JSON format with metadata
Similarity Search Cosine Similarity
Finds most relevant memories using vector similarity
Context Retrieval Top-K Results
Retrieves top relevant memories for AI context
Technical Specifications
Vector Dimensions: 768
Similarity Metric: Cosine
Storage Format: JSON
Indexing: FAISS-like

Ready to Get Started?

Join thousands of users who have embraced local AI with PROJECT-JASP. Download today and experience the future of private AI assistance.

100% Free • Open Source • No Registration Required