Skip to content

nguyenhhluong/work

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🌌 HEIFI Neural Interface: Matrix v3

HEIFI is a high-fidelity, enterprise-grade command center that bridges Large Language Models with infrastructure orchestration. Designed for the "Power Operator," it offers a unified interface for AI reasoning and autonomous SSH workflows.

HEIFI Interface

🎨 Design Language & Theming

The HEIFI Matrix v3 aesthetic is built on a "Premium Industrial" design language, prioritizing high-density information without visual clutter.

Color Palette

  • Obsidian Backbone: #000000 (Pure Black) for maximum contrast and OLED optimization.
  • Neural Accents: #1d9bf0 (Grok Blue) for primary interactive elements and active neural links.
  • Semantic Feedback:
    • Success: #10b981 (Emerald) for successful handshakes and stable links.
    • Error: #ef4444 (Rose Red) for critical faults and security breaches.
  • Typography:
    • Sans: Inter for clean, readable UI labels.
    • Mono: Geist Mono for code, terminal streams, and low-level logical identifiers.

Visual Effects

  • Glassmorphism 2.0: Uses deep backdrop-blur(45px) combined with low-opacity backgrounds (rgba(10, 10, 10, 0.15)) to create a sense of depth and focus.
  • Ambient Mesh: A fixed radial gradient system (radial-gradient) simulates an ambient glow emanating from the corners of the interface, reducing eye strain in dark environments.
  • Glow-Pulse Feedback: Interactive cards use a custom glowPulse animation that shifts border-color and box-shadow based on system state.

🚀 Core Technology Stack

Intelligence Layer

  • Google Gemini 3 API: Specifically Gemini 3 Pro Preview for its superior tool-calling accuracy and Gemini 2.5 Flash for high-speed advisory tasks.
  • Thinking Budget: Leverages the model's reasoning tokens (up to 32k) for "Deep Mode" architectural planning.
  • Local Neural Bridge: Native integration for Ollama and LM Studio via a custom local proxy.

Orchestration & Communication

  • Real-time Streams: Socket.io handles the low-latency duplex stream between the browser and remote SSH targets.
  • SSH Logic: Powered by the ssh2 library on the backend, supporting RSA/Ed25519 keys and password-based auth.
  • Terminal UI: Xterm.js with FitAddon and WebLinksAddon for a native-feeling TTY experience.

Frontend Architecture

  • React 19: Utilizing the latest concurrent rendering features.
  • Tailwind CSS: Custom configuration extending Grok-style spacing and animations.
  • React Virtuoso: High-performance virtualization for infinite chat history and agent reasoning logs.
  • Lucide Icons: Feather-weight vector iconography for clear operational signals.

🏗️ Architectural Breakdown

1. The Autonomous Loop

When in Agent Mode, HEIFI enters a recursive reasoning loop:

  1. Perception: The model analyzes the user's objective and current file system state.
  2. Planning: It emits functionCall objects (e.g., list_dir, read_file).
  3. Validation: The UI presents these actions to the operator (Human-in-the-Loop).
  4. Execution: The backend executes the command via the established SSH socket.
  5. Synthesis: The output is fed back into the model to refine the next step.

2. Security Enclave

  • Zero-Knowledge Auth: Credentials (SSH keys/passwords) are held in ephemeral memory and never persisted to a database.
  • RSA-4096 Encryption: All device-flow links (GitHub/Copilot) use standard OAuth 2.0 security protocols.
  • Termination Purge: Clicking "Terminate Node" triggers a complete memory sweep of the socket session on the backend.

🛠️ Operational Commands

Command Description
npm run dev Initialize local neural hub and next.js server.
npm run build Compile the matrix for production deployment.
docker-compose up Launch the fully containerized HEIFI stack.

HEIFI: The thin layer between human intent and machine execution.

About

a simple chat to learn

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors