Beksar.AI — Desktop AI Agent

A powerful, privacy-first desktop AI agent with live screen reading, chess analysis, real-time streaming, and multi-provider support — built with Tauri + Rust + React.

---

✨ Features

Feature	Description
🧠 Multi-Provider AI	Switch between OpenAI (GPT-4o-mini), Anthropic (Claude 3.5) and Google (Gemini 1.5 Flash) on the fly
⚡ Real-Time Streaming	Tokens are streamed live from the AI to your screen via Tauri events — no waiting for the full response
🖥️ Live Screen Mode	Activate live screen reading so the agent can observe what's on your display and narrate its activity in the toolbar
♟️ Chess Analysis Mode	Watch a chess game while the agent analyzes the board live and suggests the best moves directly in the toolbar
⏸️ Pause & Resume	Pause the live session at any moment, then resume without losing context
🎙️ Voice Control	Toggle voice input mode — talk to the agent hands-free
🛡️ BYOK Privacy	Bring Your Own Key — API keys are stored only in local storage, never transmitted to any backend
🔐 Firebase Auth	Email/password registration and login powered by Firebase Authentication
🌑 Native Desktop	Lightweight native binary for macOS, Windows and Linux via Tauri — no Electron, no Chromium bloat

---

🖼️ App Structure

Landing Page (/) → Login (/login) → Agent (/agent)

• Landing — Hero page with features overview and quick-start CTAs
• Login — Firebase email/password auth (sign up or log in)
• Agent — Full-featured chat interface with live toolbar and activity feed

---

🏗️ Tech Stack

Layer	Technology
Frontend	React 19, TypeScript 5, Tailwind CSS v4, Framer Motion, Lucide Icons
Backend	Rust (Tauri v2), reqwest, tokio, futures-util
AI Providers	OpenAI API, Anthropic API, Google Gemini API
Auth	Firebase Authentication
Build Tool	Vite 7
Packaging	Tauri bundler (macOS .dmg, Windows .msi, Linux .AppImage)

---

🚀 Getting Started

Prerequisites

• Node.js ≥ 18
• Rust stable toolchain
• Tauri v2 system dependencies for your OS
• An API key from at least one provider: OpenAI, Anthropic, or Google AI Studio

1 — Clone

git clone https://github.com/sardorazimov/beksar.git
cd beksar

2 — Install frontend dependencies

npm install

3 — Run in development mode

npm run tauri dev

This starts the Vite dev server and opens the Tauri native window simultaneously.

4 — Enter your API key

Click the Settings (⚙️) button in the sidebar, select your AI provider, and paste your API key. Keys are stored in browser localStorage — they never leave your device.

5 — Build for production

npm run tauri build

The packaged installer is placed in src-tauri/target/release/bundle/.

---

🔧 How It Works

Chat & Streaming

1. You type a message and press Send (or Enter).
2. The React frontend calls the Tauri stream_prompt command via invoke().
3. The Rust backend sends the request to the selected AI provider with SSE streaming enabled.
4. Each token is parsed from the SSE stream and emitted as a stream_chunk Tauri event.
5. The frontend listens and appends each chunk to the streaming response in real time.
6. When the stream ends (stream_end event), the response is committed to the message history.

Live Screen Mode

1. Click Canlı Ekran (Live Screen) in the toolbar.
2. The agent activates screen observation mode and logs its activity in the collapsible panel below the toolbar.
3. In Chess Mode (♟ Satranç Modu), the agent focuses on the chess board and displays real-time move suggestions in the toolbar badge.
4. Click Beklet (Pause) to freeze the agent without stopping the session; click Devam Et (Resume) to continue.
5. Click Durdur (Stop) to end the live session entirely.

---

📁 Project Structure

beksar/
├── src/                        # React frontend
│   ├── App.tsx                 # Router (Landing → Login → Agent)
│   ├── firebase.ts             # Firebase auth helpers
│   └── pages/
│       ├── Landing.tsx         # Welcome / features page
│       ├── Login.tsx           # Firebase auth form
│       └── Agent.tsx           # Main agent UI (chat + toolbar + live screen)
│
├── src-tauri/                  # Tauri / Rust backend
│   ├── src/
│   │   ├── main.rs             # Binary entry point → calls lib::run()
│   │   ├── lib.rs              # Tauri setup, command registration
│   │   ├── openai_stream.rs    # OpenAI SSE streaming implementation
│   │   ├── gemini.rs           # Google Gemini SSE streaming
│   │   ├── anthropic.rs        # Anthropic Claude SSE streaming
│   │   ├── agent/              # Agent core & conversation memory
│   │   ├── providers/          # LLM provider abstraction trait
│   │   └── tools/              # Tool implementations (file_read, …)
│   ├── Cargo.toml
│   └── tauri.conf.json         # Window size, bundle config
│
├── index.html
├── package.json
└── vite.config.ts

---

🛡️ Security & Privacy

• BYOK (Bring Your Own Key) — Your API keys are stored exclusively in the browser's localStorage on your local machine. They are sent directly from your device to the AI provider — Beksar never proxies them through any server.
• No telemetry — Beksar does not collect usage data, analytics, or any personal information.
• Tool approval — If future tool-use capabilities require file system access, a confirmation dialog appears before execution.
• Do not commit your .env file — it is listed in .gitignore.

---

🗺️ Roadmap

• Actual screen capture integration (Tauri screenshot plugin)
• Voice-to-text transcription (Whisper API)
• Conversation history persistence (SQLite via Tauri)
• Plugin system for custom tools
• Multi-window support

---

📄 License

MIT License — see LICENSE for details.