Documentation & Knowledge Hub

System Guides, Architecture & Docs.

Everything you need to master Yaazh Web Platform, Chrome Extension, and the Voice-Controlled Desktop Automation Agent.

GETTING STARTED

Introduction to Yaazh AI

Yaazh AI (தமிழ் AI) is a state-of-the-art autonomous multimodal AI platform. It unites three synchronized experiences:

01. WEB PLATFORM

20+ Universal AI tools, deep reasoning models, research and translation suite.

02. CHROME EXTENSION

Autonomous web browsing, intelligent form filling, and DOM manipulation.

03. DESKTOP AGENT

Voice-controlled computer automation that sees screens and clicks controls.

DESKTOP AGENT

Desktop Agent Operating Pipeline

The Desktop Automation Agent operates using a high-precision 4-stage loop:

  1. 1
    Multilingual Voice & Text Input: Processes user intent in Tamil, Tanglish, or English via real-time speech recognition.
  2. 2
    Visual Planning & Coordinate Estimation: Captures an ephemeral screen frame, parses bounding boxes for UI buttons, inputs, and icons.
  3. 3
    Batched Action Execution: Simulates smooth mouse movement, keystrokes, and window navigation while monitoring screen feedback.
  4. 4
    Human Approval Gate: Pauses for manual confirmation before sensitive actions (file deletions, payments, form submission).
SHORTCUTS

Keyboard Shortcuts Cheatsheet

ACTIONPRIMARY SHORTCUTALTERNATIVEPURPOSE
New Chat / Task
CtrlShiftO
CtrlN
Start a fresh conversation and reset screen state
Voice Recording
CtrlSpace
—Toggle microphone recording for voice task
Toggle Sidebar
CtrlB
—Collapse or expand the left sidebar
History & Audit
CtrlH
Ctrl2
Open past conversation transcripts and action log
Plugins & MCP
CtrlP
Ctrl3
View and connect external MCP tool servers
Settings & Keys
Ctrl,
CtrlShiftS
Configure API keys, AI model, voice, and safety gates
Theme Switcher
CtrlShiftT
—Toggle between Living Moss Dark and Clean Light mode
Emergency Stop
Esc
—Immediately abort running task or close open sheet
SAFETY PROTOCOL

Safety Architecture & Killswitches

Autonomy is safe only when the human always maintains supreme authority over the machine. Yaazh incorporates 4 protective layers:

Human Takeover Guard

Any physical movement of your mouse or hardware keyboard immediately pauses automation, preventing cursor conflicts.

Emergency Killswitch (Esc)

Pressing Escape immediately terminates all active subprocesses, cancels pending clicks, and restores complete desktop focus.

Destructive Action Confirmation

Actions involving file deletion, formatting, or financial checkouts trigger an explicit on-screen modal requiring user confirmation.

System32 Protected Zone

Critical Windows system directories, registry hives, and security certificates are permanently locked from unauthorized agent writes.

CONFIGURATION

AI Models & Model Context Protocol (MCP)

1. Local Configuration (.env)

Configure your API credentials directly in your local environment file. Keys are encrypted and masked in the application UI:

# Model API Configuration (Stored locally in .env)
GEMINI_API_KEY="AIzaSy..."
ANTHROPIC_API_KEY="sk-ant-api03-..."
OPENAI_API_KEY="sk-proj-..."

# Local Model & Vision Settings
DEFAULT_VISION_MODEL="gemini-2.5-flash"
FALLBACK_VISION_MODEL="claude-sonnet-5"
LOCAL_OLLAMA_ENDPOINT="http://localhost:11434"

# Audio & Speech-to-Text
WHISPER_MODEL_SIZE="base"
ENABLE_TAMIL_TTS=true

2. MCP Server Integrations (mcp_config.json)

Connect external tool servers using the open Model Context Protocol to grant the agent filesystem access, headless browser scraping, or database query capabilities:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "C:\\Users\\Workspace"]
    },
    "puppeteer": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-puppeteer"]
    }
  }
}
VOICE INTERACTION

Tamil, Tanglish & Multilingual Voice Commands

Yaazh is natively optimized for Tamil, Tanglish, and Indian regional languages. Speak naturally without strict keyword syntax:

Tamil / TanglishFile System & Explorer Automation

“Downloads-ல இருக்குற PDF-ஐ Desktop-க்கு move பண்ணு”

Action: Locates recent PDF in Downloads folder and safely moves it to Desktop.
TamilBrowser & Web Search

“Chrome open பண்ணி இன்றைய வானிலை நிலவரத்தைப் பார்”

Action: Launches browser, navigates to search, extracts local weather forecast.
TanglishData Analysis & Spreadsheet

“இந்த Excel sheet-ல் உள்ள sales data-வை summarize பண்ணி chart காட்டு”

Action: Reads spreadsheet columns, computes totals and displays insights.
EnglishAutonomous Web Form Automation

“Fill out the college scholarship application form with my profile”

Action: Detects input fields, enters verified user details, pauses before submit for approval.
Ready to Experience Autonomy?

Start using Yaazh on Web, Chrome, or Desktop.

Explore all 20+ universal AI tools, automated workflow agents, and desktop computer automation features.

ProductsGet Started