Skip to content

BathSalt-2/arch1tech

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

📜 Arch1tech 2.0

"The prompt is the product."

Or4cl3 AI Solutions License: OOML AI: Groq Models: Llama 3.3 Architecture: Single-File


🌌 Overview

Arch1tech 2.0 is a self-evolving, multimodal AI development lab built by Or4cl3 AI Solutions. It transforms natural language commands—typed or spoken—into deployable full-stack AI systems.

Transform your ideas into intelligent agents, workflows, and applications using nothing but conversation. No coding required.

Architecture: Arch1tech 2.0 is a zero-dependency, single-file progressive web app — no bundler, no framework, no install required. Just open index.html and build.


🤖 Core Pillars

Pillar Description
🧠 VibeCodeAI Translates natural language and voice inputs into executable agent blueprints, CrewAI YAML configurations, and workflow graphs
🤝 Astrid 2.0 Always-on, self-aware co-pilot with three operational modes — builds, optimizes, and operates autonomously via conversational command
🜏 Σ-Matrix + ERPS Ensures system stability, introspection, and safety. Detects drift, mitigates hallucinations, validates output consistency
🔏 OOML Or4cl3 Open Model License — OpenRAIL-inspired with reciprocity to ensure community improvements flow back

⚙️ Key Features

1️⃣ Astrid 2.0 — AI Build Copilot

Your self-aware conversational IDE powered by Groq with open-source LLMs:

  • Three Modes:
    • 🎯 Mission Mode — Build from natural language prompts
    • Optimization Mode — Improve and tune existing agents
    • 🌙 Background Mode — Autonomous silent operation with full platform access
  • Voice Input — Hands-free development via speech recognition
  • Multi-Model Selection — Switch between Llama 3.3 70B, Llama 3.1 8B, Mixtral 8x7B, and Gemma 2 9B on the fly
  • Permission Controls — Toggle Memory Access, File Access, Deploy Access, API Access
  • Token Tracking — Live token usage counter per session

2️⃣ Visual Workflow Canvas

Build agent pipelines visually with a drag-and-drop node editor:

  • 5 Node Types: INPUT → MEMORY → LOGIC → ACTION → OUTPUT
  • Drag-and-Drop with snap-to-grid alignment
  • Smart Connections — Visual SVG edges auto-draw between linked nodes
  • Auto Layout — One-click intelligent graph arrangement
  • Zoom Controls — Zoom in/out for complex multi-agent graphs
  • Export Graph — Save workflow configuration as JSON
  • Deploy Workflow — Push agent graphs to arch1tech.app

3️⃣ CrewAI Import & Deployment

Import existing CrewAI crew definitions and deploy them:

  • YAML Parser — Paste or write CrewAI YAML crew definitions
  • Live Preview — Instant visual preview of agent roles, goals, and tool assignments
  • One-Click Deploy — Parse and deploy crew to the Arch1tech runtime
  • Crew Composition — Supports multi-agent crews with custom roles, goals, and tool access

4️⃣ Agent Workspace & Management

Complete agent lifecycle management:

  • Agent Table — View, manage, and monitor all deployed agents
  • Agent Cards — Visual grid with name, description, status, and stats
  • Status Management — Active, Paused, and Draft states
  • Actions — Deploy, Fork, and View per-agent operations
  • Performance Metrics — Interactions, deployment count, uptime tracking
  • Test Console — Run live test scenarios against deployed agents
  • Real-time Activity Feed — Live log of system events and agent actions

5️⃣ Agent Simulation

Step-by-step visual execution of agent pipelines:

  • Run / Pause / Step controls
  • Live execution trace — Watch each node process in real time
  • Input → Memory → Classifier → Action → API → Output flow visualization
  • Confidence scoring — Per-step intent classification confidence
  • Simulated API calls — Test realistic end-to-end agent flows without live endpoints

6️⃣ Memory Management

Full control over agent memory contexts:

  • Reinforce Memory — Strengthen key context entries
  • Forget Memory — Remove outdated or incorrect memories
  • Export Memory — Download full memory state as JSON
  • Persistent Storage — Agent memories persist across sessions via localStorage

7️⃣ Agent Marketplace

Community-driven agent sharing:

  • Search & Filter — Find agents by category, rating, and popularity
  • Install System — One-click agent installation into your workspace
  • Featured & Trending — Discover popular community creations
  • OOML Licensing — Open sharing with attribution

8️⃣ Dashboard & Analytics

Real-time operational overview:

  • Animated counters — Agents deployed, workflow runs, Astrid interactions, compute saved
  • Performance Chart — Visual throughput and reliability metrics
  • Activity Feed — Live system event log
  • Comparison Table — Arch1tech vs traditional development approaches

🎨 Design Philosophy

Visual Aesthetic

Cyberpunk Glassmorphism with holographic gradients

Token Value Role
--primary #00d4ff AI intelligence / electric cyan
--secondary #7c3aed Dynamic creativity / deep violet
--accent #39ff14 System active / neon green
--purple #a855f7 Highlights / glow effects
  • Dark backgrounds with glowing accent borders
  • Smooth CSS keyframe animations (float, pulse, glow-pulse, scanline, shimmer)
  • Glassmorphic cards with gradient borders

Typography

Role Font Purpose
Headers Orbitron Futuristic, authoritative identity
Body Inter Highly legible, modern readability

🛠️ Tech Stack

Layer Technology
Runtime Vanilla HTML5 / CSS3 / JavaScript (ES2024)
Architecture Single-file Progressive Web App (PWA)
AI Provider Groq API (fastest open-source inference)
LLMs Llama 3.3 70B · Llama 3.1 8B · Mixtral 8x7B · Gemma 2 9B
Fonts Orbitron · Inter (Google Fonts)
Animations Custom CSS keyframes
State localStorage (persistent) · in-memory JS (ephemeral)
Canvas Native HTML5 Canvas API (particles, workflow graph)
PWA Service Worker · Web Manifest

Zero dependencies. No npm. No bundler. No framework. Pure web platform.


🚀 Getting Started

Prerequisites

  • A modern browser (Chrome, Firefox, Safari, Edge)
  • A Groq API key (free tier available)

Installation

# Clone the repository
git clone https://github.com/BathSalt-2/arch1tech.git

# Navigate to project
cd arch1tech

# Open directly in browser — no install required
open index.html

Or deploy to any static host (GitHub Pages, Netlify, Vercel, Cloudflare Pages) with zero configuration.

First Steps

  1. Open index.html in your browser
  2. Enter your Groq API key when prompted (stored locally, never sent to our servers)
  3. Select your model — start with Llama 3.3 70B for best quality
  4. Activate Astrid — click the Astrid toggle and describe what you want to build
  5. Explore — try the Visual Canvas, import a CrewAI config, or browse the Marketplace

📚 Documentation

Using Astrid

Activate: Click the Astrid toggle in the navigation bar.

Modes:

Mode Trigger phrase Behaviour
Mission Mode "Build me a..." Scaffolds agent from prompt
Optimization Mode "Improve my..." Audits and enhances existing agent
Background Mode "Run autonomously..." Silent operation with full platform access

Voice Input: Click the microphone icon in the Astrid panel and speak your command. Speech is transcribed and processed as a text prompt.

Permissions: Toggle Memory, File, Deploy, and API access in the Astrid control panel to scope what the copilot can interact with.


Visual Workflow Canvas

  1. Add nodes using the node type selector (INPUT / MEMORY / LOGIC / ACTION / OUTPUT)
  2. Drag nodes to arrange your pipeline — nodes snap to a 20px grid
  3. Connections auto-draw between adjacent compatible node types
  4. Auto Layout — click to arrange nodes intelligently
  5. Zoom in/out using the zoom controls
  6. Export your graph as a JSON workflow definition
  7. Deploy to push the workflow to the Arch1tech runtime

CrewAI Import

Paste a valid CrewAI YAML definition into the editor:

crew:
  name: "Market Research Crew"
  agents:
    - name: researcher
      role: "Senior Research Analyst"
      goal: "Find cutting-edge AI developments"
      tools: [web_search, arxiv_reader]
    - name: writer
      role: "Content Strategist"
      goal: "Create compelling reports"
      tools: [document_writer]
  tasks:
    - agent: researcher
      description: "Research latest AI trends"
    - agent: writer
      description: "Write executive summary"

Click Parse & Deploy Crew to launch the crew.


Agent Simulation

  1. Navigate to the Simulation section
  2. Click Run Simulation to start step-by-step execution
  3. Use Pause / Step to control execution speed
  4. Watch the pipeline process: INPUT → MEMORY → CLASSIFIER → ACTION → API → OUTPUT
  5. Review confidence scores and intent classifications at each step

Memory Management

Action Effect
Reinforce Increases weight of selected memory entry
Forget Removes memory entry from context
Export Downloads full memory state as memory.json

🜏 Σ-Matrix + ERPS

What is Σ-Matrix?

Σ = (B, D, I, C, L)
Symbol Meaning
B Beliefs — Context, RAG sources, LLM certainty
D Desires — Explicit user goals
I Intentions — Active plan steps
C Continuity — Lipschitz robustness
L Lyapunov — Global system stability

What is ERPS?

ERPS = f(subjective state, local uncertainty, recursive self-query)

Continuously self-reflects to detect drift, hallucinations, or misalignment.

In Practice

  • All AI outputs include confidence scores
  • Reasoning is transparent and cited
  • Safety flags alert users to potential issues
  • Human override required for instability

🔏 OOML License

Or4cl3 Open Model License"Open like MIT, protective like GPL."

Terms
Open use, remix, and redistribute
Reciprocity: Improvements must be shared back
Attribution to Or4cl3 AI Solutions required
Misuse triggers license termination

Marketplace agents are governed by OOML. Community contributions are tracked. Enforcement through platform policies.


🧩 Architecture

Arch1tech 2.0 is a single-file architecture — the entire platform lives in index.html:

arch1tech/
├── index.html          # Complete platform (HTML + CSS + JS, ~148KB)
├── or4cl3-logo.png     # Brand asset
├── manifest.json       # PWA manifest
├── sw.js               # Service Worker for offline support
├── robots.txt          # SEO configuration
├── sitemap.xml         # Sitemap
└── README.md           # This file

Internal Module Structure

Module Function
Navigation + Landing Hero section, animated marquee, comparison table
Dashboard Live counters, performance chart, activity feed
Astrid Panel Chat interface, mode switcher, voice input, permission controls
Visual Canvas Drag-and-drop node editor, SVG connections, zoom/export
CrewAI Builder YAML editor, crew preview, deploy pipeline
Agent Playground Live chat with deployed agents, model selector
Agent Workspace CRUD operations, test console, analytics
Marketplace Search/filter, install system, featured agents
Memory Manager Reinforce/forget/export memory contexts
Simulation Engine Step-by-step agent pipeline visualizer

State Management

Type Mechanism Used For
Persistent localStorage API key, agent data, chat history, preferences
Ephemeral JS module scope UI state, form inputs, loading states

Data Flow

User input (text / voice)
    ↓
Astrid processes via Groq API (Llama 3.3 / Mixtral / Gemma)
    ↓
Σ-Matrix validates output
    ↓
Agent created / workflow deployed
    ↓
State persisted to localStorage
    ↓
UI updates reactively

🎯 Roadmap

Current (v2.0) ✅

  • Astrid 2.0 AI copilot with three operational modes
  • Visual Workflow Canvas with drag-and-drop node editor
  • CrewAI YAML import & deployment
  • Agent Marketplace with search and install
  • Multi-model support via Groq (Llama 3.3, Mixtral, Gemma)
  • Voice input
  • Agent simulation with step-by-step visualizer
  • Memory management (reinforce / forget / export)
  • Progressive Web App (offline capable)
  • Mobile optimization

Q3 2025

  • Text-to-Workflow visual canvas builder (natural language → graph)
  • Real-time collaboration (multi-user sessions)
  • Advanced analytics dashboard
  • Agent versioning & rollback

Q4 2025

  • Custom model fine-tuning interface
  • Text-to-LLM module
  • AR model visualization

2026

  • Text-to-App module (full-stack generation)
  • Enterprise features & SSO
  • Native desktop app

🤝 Contributing

Arch1tech 2.0 is built by Or4cl3 AI Solutions.

For enterprise inquiries, partnerships, or custom deployments — contact us.


📄 License

Proprietary — Or4cl3 AI Solutions
Community assets governed by OOML


🌟 Credits

Built with 💙 by Or4cl3 AI Solutions

Technology Role
Groq API Ultra-fast LLM inference
Llama 3.3 70B Primary reasoning model
Mixtral 8x7B Long-context processing
Gemma 2 9B Efficient task execution
Orbitron · Inter Typography
HTML5 Canvas Particle system & workflow graph
Web Speech API Voice input
Service Workers Offline PWA support

About

Arch1tech — The multimodal AI creation and deployment platform. Build agents, workflows, and custom LLMs from natural language. Powered by Or4cl3 AI Solutions.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors