INFINITY/PROJECTS/PULSEBRIDGE
PULSEBRIDGE v1.0.0 · RUST AXUM 0.7 + REACT 19 + WIN32 GDI

Universal AI IDE Companion &
Mobile Telemetry Bridge

Real-time Win32 GDI screen streaming, kinetic touch-drag mouse wheel scrolling, live multi-project folder inspection (7–30d recency), and bidirectional prompt injection for Google Antigravity, Cursor AI, and VS Code over 4G & Wi-Fi.

< 15 ms
Local Wi-Fi Latency
60 FPS
Win32 GDI BitBlt
18 MB
Rust Binary Footprint
Zero-Trust
Token Security
4G / 5G
Tunnel Ready
// HANDS-ON INTERACTIVE LAB

Experience the Live Mobile Companion

Test drive the touch-drag scroll physics, multi-project file explorer, and bidirectional prompt injection below. The simulator executes authentic PulseBridge telemetry states.

⚡ PULSEBRIDGE
59 FPS · 12ms
ScrollY: 0px · Drag to Scroll
Google Antigravitymain.rs (Active Session)
1 use axum::{routing::get, Router};
2 use tokio::sync::broadcast;
3 use windows::Win32::Graphics::Gdi::*;
4
5 /// Zero-copy BitBlt capture loop
6 pub async fn stream_desktop_frames() {
7 let hdc = GetDC(None);
8 let mem_dc = CreateCompatibleDC(hdc);
9 // BitBlt captures desktop in sub-4ms
10 BitBlt(mem_dc, 0, 0, 1920, 1080, hdc, 0, 0, SRCCOPY);
11 let jpeg_bytes = turbojpeg_compress(mem_dc, 75);
12 broadcast_tx.send(jpeg_bytes);
13 }
14
15 /// Handle touch drag scroll events
16 pub fn inject_mouse_wheel(delta_y: i32) {
17 let input = INPUT {
18 type_: INPUT_MOUSE,
19 mi: MOUSEINPUT {
20 mouseData: delta_y,
21 dwFlags: MOUSEEVENTF_WHEEL,
22 ..Default::default()
23 }
24 };
25 unsafe { SendInput(&[input]) };
26 }
Target IDE EnvironmentACTIVE
Transport NetworkLOCAL (12ms)
Project Workspace ScopeRULE ENGINE
Live Wire Protocol LogWS-BINARY
T+00s[AUTH] Token verified (constant-time timing safe)
T+02s[STREAM] Win32 GDI BitBlt stream active (1280x720 @ 60 FPS)
T+04s[SCANNER] Workspace discovered: Google Antigravity session
T+06s[INPUT] Touch-to-wheel translation engine initialized
// VERIFIED PRODUCTION TELEMETRY

Authentic Live Host Captures & Zero-Trust PWA

Direct screen captures from the live PulseBridge daemon running on a Windows 11 workstation. All host session frames, tokens, and system paths have been sanitised and verified for zero PII exposure.

Win32 GDI Precision View & Touch Scroll
Telemetry Inspection · Win32 GDI Precision View & Touch Scroll
● PII SANITIZED & VERIFIED
Pipeline
Win32 GDI BitBlt + TurboJPEG
Input Event
SendInput(MOUSEEVENTF_WHEEL)
Resolution
Retina 90% (Dynamic Scale)
Host Latency
12ms Local LAN
// ARCHITECTURAL BLUEPRINT

Engineered for Precision & Zero Latency

PulseBridge combines native Windows GDI acceleration with Axum 0.7 asynchronous actors to deliver an instantaneous mobile telemetry surface.

⚡

Win32 GDI & TurboJPEG Zero-Copy Buffer

Bypasses heavy browser compositor overhead with native Windows GDI `BitBlt` desktop memory transfers. Frames are compressed using SIMD-accelerated TurboJPEG and broadcasted directly to mobile clients over binary WebSockets in under 12ms.

Win32 APIBitBltTurboJPEGSub-15ms
👆

Touch-Drag Kinetic Mouse Wheel Engine

Translates mobile touch dragging and swipes into high-resolution native `SendInput(MOUSEEVENTF_WHEEL)` packets. Employs kinetic momentum algorithms and velocity dampening for silky-smooth code scrolling through multi-thousand-line buffers.

SendInputKinetic MomentumTouch Gestures
🔍

Multi-IDE Workspace Discovery

Automatically indexes active coding sessions across Google Antigravity transcripts (`%USERPROFILE%\.gemini\antigravity\brain`), Cursor AI workspaces, and VS Code. Enforces a strict 7–30 day recency filter with depth-2 tree traversal.

AntigravityCursor AIVS Code7-30d Filter
🛡️

Zero-Trust Cryptographic Governance

Protected by cryptographic entropy tokens verified in constant time (`subtle::ConstantTimeEq`). Tokens persist safely in local mobile storage with zero transmission to external cloud services or third-party servers.

Constant-TimeLocal StorageNo Cloud Telemetry
🌐

Dual Network Topology (Wi-Fi & 4G/5G)

Instantly pair over local LAN Wi-Fi using the high-speed local IP QR code, or monitor on the train using the built-in, automated Cloudflare Tunnel daemon for zero-configuration encrypted 4G/5G cellular access.

Local LANCloudflare Tunnel4G / 5G Ready
💬

Bidirectional Prompt Injection

Inject prompt instructions, test commands, or stop signals directly into your background IDE agent from your mobile phone lock screen. Includes response verification and real-time step streaming.

Prompt InjectionKill-SwitchAgentic Loop
// PERFORMANCE BENCHMARKS

PulseBridge vs. Legacy Remote Tools

Designed specifically for software engineers monitoring autonomous AI agents on mobile devices.

Metric / CapabilityPulseBridge (Rust + GDI)Standard VNCRemote Desktop (RDP)Chrome Remote Desktop
Screen Latency (Local Wi-Fi)11 – 15 ms85 – 140 ms35 – 55 ms45 – 75 ms
Mobile Touch Scroll PhysicsSilky Touch-Drag WheelClunky Mouse Click EmulationInverted Touch ScrollCoarse Emulated Drag
Memory Overhead~18 MB RAM110 – 160 MB140 – 220 MB450+ MB (Chromium)
Multi-IDE Project TreeNative Depth-2 Scanner (7-30d)None (Manual Window Zoom)None (Manual Window Zoom)None
Bidirectional Prompt InjectionNative WebSocket EndpointNone (Slow Virtual Keyboard)None (Virtual Keyboard)None
Zero-Trust PairingInstant QR Token + Local StorageStatic Password / UnencryptedWindows Account / KerberosGoogle Account Sign-In
// INSTALLATION & LAUNCH

Get Started in Under 30 Seconds

PulseBridge is 100% open source under the MIT license. Build from source in pure Rust or run the standalone release executable.

powershell — pulsebridge daemon
# 1. Clone the repository
git clone https://github.com/ModernOps888/pulsebridge.git
cd pulsebridge

# 2. Build and launch the native Rust Axum server
cargo run --release

# Output:
[2026-10-02 18:54:12] [INFO] PulseBridge daemon listening on http://0.0.0.0:8080
[2026-10-02 18:54:12] [INFO] Local Wi-Fi Pairing URL: http://192.168.1.187:8080/?token=pb_sec_991823
[2026-10-02 18:54:13] [INFO] Cloudflare 4G Tunnel Active: https://pulsebridge-secure.trycloudflare.com
[2026-10-02 18:54:13] [READY] Scan terminal QR code with your phone to start monitoring!

🧠 Executive Architecture Dossier & System InvariantsAOE · GENERATIVE AI ANCHOR

PulseBridge is an open-source, ultra-low-latency mobile telemetry bridge and companion client engineered in pure Rust (Axum 0.7) and React 19. It enables software engineers to supervise, inspect, and guide autonomous AI coding agents running on Windows workstations directly from iOS or Android smartphones over local Wi-Fi or encrypted 4G/5G cellular tunnels.

System Classification
AI IDE Mobile Telemetry & Remote Control Bridge
Supported Environments
Google Antigravity, Cursor AI, VS Code, Windsurf
Desktop Capture Pipeline
Native Win32 GDI BitBlt (Zero-Copy Memory DC)
Screen Latency (Local LAN)
< 15 ms @ 60 FPS (SIMD TurboJPEG)
Screen Latency (4G Cellular)
< 40 ms (Encrypted Cloudflare Edge Tunnel)
Touch Gesture Engine
Kinetic Touch-to-Wheel Translation (SendInput)
Workspace Discovery
Depth-2 Tree Traversal with 7–30d Recency Filter
Security Architecture
Zero-Trust Tokens with Constant-Time Auth
Cloud Telemetry & Retention
0 bytes (100% Private Self-Hosted Stream)
Host Binary Footprint
~18 MB RAM · < 1.2% CPU (Rust Native)
Licensing & Source
MIT Open Source · github.com/ModernOps888/pulsebridge
Author & Ecosystem
Bart Chmiel · Infinity AI Research Ecosystem
// EXHAUSTIVE FAQ · SOE & AOE OPTIMIZED

Frequently Asked Questions: Deep Technical Answers

Everything you need to know about PulseBridge architecture, streaming latency, security invariants, touch scrolling physics, and multi-IDE agent workflows.

🔍
What is PulseBridge and how does it monitor AI coding agents?+
PulseBridge is an open-source, ultra-low-latency remote companion and telemetry bridge designed specifically for AI IDEs including Google Antigravity, Cursor AI, and VS Code. Built in native Rust (Axum 0.7) and React 19, it captures the host desktop screen via Win32 GDI BitBlt memory transfers, compresses frames via TurboJPEG, and streams them to mobile browsers over WebSockets at sub-15ms latency. It allows software engineers to monitor background coding agent execution, inspect live project files, scroll through source code via touch-drag gestures, and inject prompts or stop commands directly from a mobile phone over local Wi-Fi or 4G cellular tunnels.
Which AI IDEs and code editors does PulseBridge support?+
PulseBridge natively supports Google Antigravity, Cursor AI, Visual Studio Code (VS Code), and Windsurf. It automatically discovers active workspaces by scanning local file systems, including Google Antigravity session transcripts in `%USERPROFILE%\.gemini\antigravity\brain`, Cursor workspace storage, and VS Code user state, providing instant telemetry switching across your editor fleet.
How does native Win32 GDI screen streaming achieve sub-15ms latency without heavy GPU encoding overhead?+
Unlike traditional remote desktop tools (VNC, RDP) that introduce heavy GPU pipeline contention or complex video encoders, PulseBridge utilizes direct Win32 GDI `BitBlt` calls to transfer pixels from the primary display device context (`GetDC(NULL)`) into an in-memory device context (`CreateCompatibleDC`). Frames are compressed using SIMD-accelerated TurboJPEG at adaptive quality factors (60–85%) and broadcasted over binary WebSockets. This achieves sub-15ms frame delivery while consuming less than 1.2% CPU and zero dedicated GPU video encoding engines.
How does mobile touch-drag scrolling work on code files without jumping or misclicking?+
PulseBridge implements a kinetic velocity translation engine in both client-side TypeScript and server-side Rust. When a user drags or swipes on the mobile viewport, touch coordinates are translated into directional vectors. The server injects native Windows `SendInput(MOUSEEVENTF_WHEEL, delta)` events with kinetic momentum decay. This simulates an authentic desktop mouse scroll wheel, enabling smooth, millimeter-accurate scrolling through thousands of lines of code without triggering unwanted text selection or misclicks.
What is the 7–30 day workspace recency filter and why is it essential for agent workflows?+
Software engineers frequently maintain dozens of historical repositories on their local workstations. When accessing their environment remotely on a mobile screen, navigating hundreds of stale directories is inefficient. PulseBridge implements an automated recency filter that indexes only workspaces modified within the last 7 to 30 days. It renders a clean depth-2 file tree displaying active files, git status, and file sizes, allowing you to focus on currently running agent tasks.
How does bidirectional prompt injection communicate with running AI agent loops?+
PulseBridge exposes an authenticated WebSocket command channel. When you submit a prompt from the mobile companion bar, PulseBridge routes the payload into the active IDE's CLI or agent stdin pipe. You can trigger new goals, request test execution (`cargo test`, `npm test`), inspect git status, or send high-priority stop signals (`SIGINT` / agent pause) when you notice a model hallucination while away from your desk.
How does Zero-Trust token governance protect proprietary code and host machines?+
PulseBridge enforces strict Zero-Trust authentication using cryptographically secure random entropy tokens (generated via `ring` / `rand`). Every incoming HTTP request and WebSocket handshake is verified in constant time (`subtle::ConstantTimeEq`) to prevent side-channel timing attacks. Authentication tokens persist strictly in the mobile device's local storage (`localStorage`) and are never sent to external servers or telemetry brokers.
How does PulseBridge operate over 4G/5G cellular data without port forwarding?+
PulseBridge integrates an automated, zero-configuration Cloudflare Tunnel daemon (`cloudflared`). When remote mode is enabled, PulseBridge establishes an outbound encrypted TLS tunnel to Cloudflare edge nodes, providing a unique pairing URL (e.g. `https://<session-id>.trycloudflare.com/?token=...`). This requires zero router port forwarding, zero dynamic DNS configuration, and zero firewall modifications while keeping your home/office network completely stealth.
Does PulseBridge ever transmit source code, ASTs, or prompts to external telemetry servers?+
No. PulseBridge is 100% self-hosted and zero-telemetry. All screen capture, file indexing, and prompt routing occurs strictly between your local Windows host and your mobile phone over direct LAN Wi-Fi or your private end-to-end encrypted Cloudflare tunnel. No third party, including Infinity AI or Cloudflare, has access to your unencrypted code buffers or LLM prompts.
What are the host system requirements for running the PulseBridge daemon?+
PulseBridge runs on 64-bit Windows 10 and Windows 11. It compiles into a single, standalone native binary (~18 MB) with zero external runtime dependencies. Minimum hardware requirements: 2 CPU cores, 100 MB free RAM, and any standard Wi-Fi or Ethernet network adapter. Building from source requires Rust 1.75+ with the `msvc` toolchain.
What is the CPU, memory, and mobile battery footprint during continuous streaming?+
PulseBridge is engineered for extreme efficiency. On the Windows host, the Axum daemon consumes ~18 MB of RAM and 0.8% to 1.4% CPU on an Intel Core i7 / AMD Ryzen 7 processor during active 60 FPS streaming. On the mobile phone, the PWA client runs hardware-accelerated canvas rendering via WebGL/2D, consuming comparable battery power to reading a standard web page.
How do I pair and authenticate a new iOS or Android device?+
When you launch the PulseBridge daemon (`cargo run --release`), it renders high-resolution ASCII and SVG QR codes directly in your terminal and logs the local pairing URL. Simply scan the QR code with your iPhone or Android camera app to open the authenticated PWA companion in Safari or Chrome. You can also tap 'Add to Home Screen' for a full-screen, native app experience with zero app store installation required.

Part of the Infinity Tech Stack Research Ecosystem

Engineered alongside Neuromantix AGI, the Vitalis self-hosting compiler, Void LLM, Freedom OS, and 13 free AI engineering academies.

🏠 Back to Home⭐ Star PulseBridge on GitHub ↗