Skip to content

vin1397/prive

BranchesTags

Repository files navigation

██████╗ ██████╗ ██╗██╗   ██╗███████╗
██╔══██╗██╔══██╗██║██║   ██║██╔════╝
██████╔╝██████╔╝██║██║   ██║█████╗  
██╔═══╝ ██╔══██╗██║╚██╗ ██╔╝██╔══╝  
██║     ██║  ██║██║ ╚████╔╝ ███████╗
╚═╝     ╚═╝  ╚═╝╚═╝  ╚═══╝  ╚══════╝

⚡ Your Local AI Coding Companion

Chat. Code. Build. Ship.


TypeScript Node.js Ollama Qwen Platform License Status Tests


No subscriptions. No cloud dependency. No API bills.

Powered entirely by local AI models through Ollama.


🟣 Overview

Prive is an open-source, terminal-first AI coding assistant built for developers who demand complete control over their workflow and their data.

Unlike cloud-based tools, Prive runs 100% on your machine using local language models — Qwen, DeepSeek, Llama, and any Ollama-compatible model. No accounts. No API keys. No monthly bills. No telemetry. Just code.

✨ Features

🤖 Local AI Engine

  • Powered by Ollama
  • Qwen · DeepSeek · Llama · any model
  • Fully offline capable
  • Zero API keys or usage limits
  • Privacy-first by design

💬 Interactive Chat

  • Project-aware multi-turn conversations
  • Full session memory
  • Natural language workflows
  • Token-aware context trimming

📂 Workspace Intelligence

  • Read files into AI context
  • Analyze entire repositories
  • Generate architecture summaries
  • Fuzzy file search + content grep

✏️ Code Generation & Editing

  • Create, modify, and refactor code
  • Complete files only — no placeholders
  • Automatic backup on overwrite
  • Inline diff explanations

💻 Terminal Automation

  • Execute commands with real-time output
  • 21-pattern danger sandbox
  • Sequence execution with fail-fast
  • Smart stderr capture

🌿 Git Integration

  • Status, commit, push, pull
  • Branch create/switch/merge
  • Recent commit log
  • Auto stage on commit

🚀 Quick Start

1 · Install Ollama

# macOS / Linux
curl -fsSL https://ollama.ai/install.sh | sh
ollama serve

2 · Pull a model

# Recommended — best for coding
ollama pull qwen2.5-coder

# Alternatives
ollama pull deepseek-coder-v2
ollama pull qwen3
ollama pull codellama

3 · Install Prive

git clone https://github.com/vin1397/prive.git
cd prive
npm install
npm run build
npm link

4 · Run

cd ~/your-project
prive
 ██████╗ ██████╗ ██╗██╗   ██╗███████╗
 ...

  ⚡ Your Local AI Coding Companion · v0.1.0

╔══════════════════ ⚡ PRIVE DASHBOARD ══════════════════╗
║  ⚙  CPU           23%    ████░░░░░░░░░░░░  23%        ║
║  🧠  Memory        4.1/16 GB  ██░░░░░░░░░░  26%       ║
║  🌡  Temp          54°C                                ║
║  🤖  Model         qwen2.5-coder                       ║
║  📂  Project       your-project                        ║
╚════════════════════════════════════════════════════════╝

[CPU 23%·RAM 26%·54°C] ❯ _

💬 Usage

Chat — just type

❯ explain what this project does
❯ how do I add rate limiting to Express routes?
❯ what's the idiomatic way to handle errors in Go?

Load files into context

❯ /read src/auth.ts
❯ refactor this to use async/await throughout

Generate and write code

❯ /write src/utils/date.ts utility functions for date formatting

Debug an error

❯ /debug TypeError: Cannot read properties of undefined (reading 'map')

Git workflow

❯ /git status
❯ /git commit feat: add user authentication
❯ /git push

Run terminal commands

❯ /run npm install
❯ /run npm test

📋 Commands

Command Description
[message] Chat with the AI directly
/read <file> Read a file into AI context
/write <file> <task> Generate and write code to a file
/list List all source files
/tree [depth] Display the project file tree
/run <cmd> Execute a terminal command safely
/git status|commit|push|branch|log Git operations
/grep <pattern> Search file contents
/model [name] Show or switch the active model
/context [clear] Show or clear session context
/history View chat history
/memory [clear] View or clear saved sessions
/architect Analyze project architecture
/debug <error> Analyze an error with AI
/clear Clear the terminal
/exit Quit Prive

🤖 Agent System

Prive ships three specialized AI agents, each with its own focused system prompt and conversation state:

🏗️ Architect

Analyzes project structure, identifies architectural patterns, highlights concerns, and suggests concrete improvements.

/architect

⚙️ Coder

Generates, refactors, and explains code. Always produces complete, working files — never stubs or placeholders.

/write · /read + chat

🐛 Debugger

Analyzes error messages, identifies root causes, and can apply fixes directly to files with auto-backup.

/debug

🧠 Memory System

All session data is stored locally inside your project:

.prive/
├── memory.json      # Last 50 conversation sessions
├── projects.json    # Known projects + last model used
└── settings.json    # Your preferences

Sessions are saved automatically on exit. Memory never leaves your machine.

🖥️ Live System Dashboard

Every time you launch Prive, a live system dashboard renders in your terminal:

╔══════════════════════ ⚡ PRIVE DASHBOARD ══════════════════════╗
║                                                               ║
║   ◆ SYSTEM ──────────────────────────────────────────         ║
║                                                               ║
║   ⚙  CPU           23%     ████░░░░░░░░░░░░  23%             ║
║   🧠  Memory        4.1/16   ██░░░░░░░░░░░░░  26%            ║
║   🌡  Temp          54°C                                      ║
║   🎮  GPU           12% · 2048/8192 MB                        ║
║   🌐  Network       192.168.1.10                              ║
║   ⏱  Uptime        6h 32m                                    ║
║   📊  Load avg      0.48  0.31  0.22    1m · 5m · 15m        ║
║   💻  Platform      macOS arm64                               ║
║   🔢  CPU cores     10                                        ║
║                                                               ║
║   ◆ PRIVE ───────────────────────────────────────────         ║
║                                                               ║
║   🤖  Model         qwen2.5-coder                             ║
║   📂  Project       my-project                                ║
║   🧩  Agents        Architect · Coder · Debugger              ║
║   🔒  Privacy       100% local · no telemetry                 ║
╚═══════════════════════════════════════════════════════════════╝

The prompt line updates live every 2 seconds:

[CPU 23%·RAM 26%·54°C·GPU 12%] ❯ _

🔧 Configuration

.prive/settings.json:

{
  "model": "qwen2.5-coder",
  "ollamaHost": "http://localhost:11434",
  "temperature": 0.7,
  "streamResponses": true,
  "autoSaveMemory": true,
  "maxHistoryLength": 50
}

Switch model on the fly:

❯ /model deepseek-coder-v2

🔌 Supported Models

Model Best For Context
qwen2.5-coder ⭐ Code generation, review, refactoring 32K
qwen3 ⭐ Reasoning + coding 32K
deepseek-coder-v2 ⭐ Complex algorithms, multi-language 64K
deepseek-r1 Reasoning, debugging 32K
codellama Code completion 16K
llama3.2 General questions, explanations 8K
mistral Fast general-purpose 8K

Any Ollama-compatible model works.

🧱 Project Structure

Prive/
│
├── src/
│   ├── cli/              # Entry point · command router · prompt
│   │   ├── index.ts
│   │   ├── commands.ts   # 15 commands with full handlers
│   │   └── prompt.ts     # Live system stats in prompt line
│   │
│   ├── ai/               # AI layer
│   │   ├── ollama.ts     # Streaming HTTP client
│   │   ├── chat.ts       # History + context trimming
│   │   ├── context.ts    # File → prompt builder
│   │   └── memory.ts     # Session persistence
│   │
│   ├── agents/           # Specialized AI agents
│   │   ├── coder.ts
│   │   ├── debugger.ts
│   │   └── architect.ts
│   │
│   ├── filesystem/       # I/O operations
│   │   ├── read.ts
│   │   ├── write.ts      # With auto-backup support
│   │   ├── search.ts     # Grep · glob · find
│   │   └── tree.ts
│   │
│   ├── terminal/         # Command execution
│   │   ├── execute.ts    # Streaming output
│   │   └── sandbox.ts    # 21 danger patterns
│   │
│   ├── git/              # Git operations
│   │   ├── status.ts
│   │   ├── commit.ts
│   │   ├── push.ts
│   │   └── branch.ts
│   │
│   ├── config/
│   │   ├── constants.ts
│   │   ├── models.ts     # 8 model definitions
│   │   └── settings.ts
│   │
│   └── utils/
│       ├── logger.ts     # Live dashboard · system metrics
│       ├── spinner.ts
│       └── helpers.ts
│
├── tests/
│   ├── ai.test.ts        # 13 tests
│   ├── fs.test.ts        # 24 tests
│   └── git.test.ts       # 26 tests  ← 63 total, all passing
│
├── docs/
│   ├── architecture.md
│   ├── commands.md
│   └── roadmap.md
│
└── .prive/               # Local memory store (gitignored)
    ├── memory.json
    ├── projects.json
    └── settings.json

🏗 Architecture

graph TD
    %% Nodes
    Dev([💻 Developer]) -->|Prompt| CLI[🛠️ Prive CLI]
    
    CLI --> Agent[🤖 Agent Layer]
    
    Agent -->|Read / Write| FS[📂 Filesystem]
    Agent -->|Execute| Term[🖥️ Terminal]
    Agent --> AI[🧠 AI Layer]
    
    AI <--> Ollama[🦙 Ollama]
    Ollama <--> Models[[🔮 Qwen / DeepSeek]]

    %% Style Configurations
    style Dev fill:#238636,stroke:#2ea44f,stroke-width:2px,color:#fff
    style CLI fill:#1f6feb,stroke:#388bfd,stroke-width:2px,color:#fff
    style Agent fill:#8957e5,stroke:#a371f7,stroke-width:2px,color:#fff
    style FS fill:#6e7681,stroke:#8b949e,stroke-width:1px,color:#fff
    style Term fill:#6e7681,stroke:#8b949e,stroke-width:1px,color:#fff
    style AI fill:#da3633,stroke:#f85149,stroke-width:2px,color:#fff
    style Ollama fill:#d29922,stroke:#f2cc60,stroke-width:2px,color:#000
    style Models fill:#0e4429,stroke:#26a641,stroke-width:2px,color:#fff
Loading

🛣️ Roadmap

Phase 1 — Foundation ✅

  • 🟣 Interactive Chat
  • 🟣 Ollama Integration
  • 🟣 Terminal UI with live dashboard
  • 🟣 Qwen / DeepSeek / Llama support

Phase 2 — Workspace ✅

  • 🟣 File Reading & Writing (with backup)
  • 🟣 Project Search (glob + grep + fuzzy)
  • 🟣 Terminal Commands Integration (sandbox)

Phase 3 — VCS & Memory ✅

  • 🟣 Git Integration (status · commit · push · branch · log)
  • 🟣 Local Memory Storage (session history + project registry)

Phase 4 — Agents ✅

  • 🟣 Multi-Agent Architecture (Coder · Debugger · Architect)
  • 🟣 Autonomous Task Planner (chain agents across multi-step goals)

Phase 5 — Studio ⚡ In Progress

  • 🟣 Desktop Studio UI (Electron — chat, file tree, git, terminal, metrics)
  • 🔲 Voice Development Integration (local Whisper — planned v0.6)
  • ⬜ Generate images

Phase 6 — Ecosystem ⚡ In Progress

  • 🟣 Plugin system (.prive/plugins/ — load custom .mjs commands)
  • 🟣 MCP server (JSON-RPC · compatible with Claude Desktop, Cursor)
  • 🟣 VS Code extension (chat panel, explain, refactor, debug, commit)
  • 🔲 Remote workspace support

🔒 Privacy

Principle Detail
🔒 Zero telemetry No analytics, no tracking, no crash reports
🏠 Local inference All AI runs via Ollama on your machine
💾 Local storage .prive/ in your project — never synced
👁️ Open source Audit every line
✈️ Offline capable Works with no internet connection

🛠️ Development

# Install dependencies
npm install

# Dev mode (ts-node, no build needed)
npm run dev

# Build TypeScript
npm run build

# Run all 63 tests
npm test

# Watch mode
npm run test:watch

🤝 Contributing

Contributions, bug reports, and pull requests are welcome.

  1. Fork the repository
  2. Create a feature branch: git checkout -b feat/my-feature
  3. Write tests for new functionality
  4. Ensure all tests pass: npm test
  5. Submit a pull request

Built For Prive💜 by ~Vin

Prive — Your Local AI Coding Companion

MIT License · Open Source · Local First · Privacy Focused

About

Prive — Your Local AI Coding Companion. Chat, code, and build with local AI.

Resources

Readme

License

MIT license
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages
⚡