Introduction to RTILA X

What is RTILA X?

RTILA X is a local-first, AI-powered web automation platform that runs entirely on your own computer. It lets you record, build, and run browser automations — clicking buttons, filling forms, scraping data, crawling websites — without writing a single line of code.

The name “RTILA” comes from the Moroccan Darija word for “spider,” reflecting the software’s core ability to crawl the web and extract data, much like a spider weaves its web across the internet.

Unlike cloud-based scraping tools that send your data to remote servers, RTILA X processes everything locally. Your projects, your scraped data, your AI conversations — none of it ever leaves your machine unless you explicitly choose to send it somewhere.

What Makes RTILA X Different

Feature	RTILA X	Typical Cloud Scrapers
Runs 100% locally	✅ All processing on your machine	❌ Data sent to cloud servers
AI-powered building	✅ Describe what you want in plain English	❌ Manual configuration only
Standalone bot export	✅ Package automations as sellable executables	❌ Locked to the platform
Anti-bot evasion built in	✅ Humanoid mouse, proxies, CAPTCHA solving	❌ Requires separate tools
Checkpoint & resume	✅ Survives crashes and network drops	❌ Starts over on failure
30+ data destinations	✅ Databases, cloud storage, BI tools, email, and more	❌ Limited export options
Free tier available	✅ Free forever with 1 project	❌ Usually paid only
Any GGUF model support	✅ Use any Hugging Face GGUF model locally	❌ N/A
KV cache optimization	✅ Full control over inference memory	❌ N/A
Trigger Chains	✅ Build multi-step post-processing pipelines	❌ Single-step exports only
No monthly data limits	✅ Scrape as much as your hardware allows	❌ Per-page or per-credit pricing
Open architecture	✅ Tauri + Deno + PocketBase stack	❌ Proprietary black box

What You Can Do With RTILA X

• Scrape any website — Extract product listings, prices, reviews, contacts, job listings, real estate data, and more from any public webpage. RTILA X handles pagination, infinite scroll, and dynamic content automatically.

• Automate form filling — Log into websites, fill multi-step forms, submit applications, and handle CAPTCHAs. Perfect for repetitive data entry tasks that would otherwise take hours.

• Monitor changes — Track price changes, content updates, stock levels, or job postings on a schedule. Get notified via email, Slack, or Telegram when something changes.

• Crawl entire websites — Follow links across pages, respect robots.txt rules, and extract structured data from hundreds or thousands of pages with configurable concurrency and delays.

• Build and sell bots — Package your automations as standalone executables that run without RTILA installed. Sell them to clients or deploy them on servers. Password-protect them and revoke access remotely.

• Trigger downstream pipelines — Send scraped data to databases (PostgreSQL, MySQL, MongoDB), cloud storage (S3, GCS, Azure Blob), BI tools (Power BI, Tableau), messaging queues (RabbitMQ, Kafka), or 30+ other destinations automatically after each run.

• Build Trigger Chains — Create reusable multi-step post-processing pipelines with conditional logic (IF/ELSE), parallel execution blocks, and scheduled runs. Chain multiple triggers together so data flows through your entire pipeline automatically.

• Control from your phone — Use the built-in Telegram bot to trigger runs, check status, view logs, and download results from anywhere.

System Requirements

Requirement	Minimum	Recommended
OS	Windows 10, macOS 11+, Ubuntu 22.04+	Latest version of your OS
RAM	4 GB	8 GB+ (16 GB for local AI)
Disk Space	2 GB free	10 GB+ (more for AI models)
Browser	Microsoft Edge, Brave, or Google Chrome installed	Microsoft Edge (recommended)
Internet	Required for initial setup and target websites	Broadband connection
CPU	Dual-core 2.0 GHz	Quad-core 3.0 GHz+

For Local AI features (v8.3.2+): 6–8 GB RAM for 7B–9B parameter models at Q4 quantization. For larger models (13B+), 16 GB+ RAM and a dedicated GPU are recommended. KV cache quantization (Q8_0) can reduce memory usage by approximately 50%.

Supported Platforms

Windows

• Windows 10 (version 1809 or later)
• Windows 11 (all editions)
• Windows Server 2019+ (requires Visual C++ Redistributable, auto-installed)
• Installer format: .msi (NSIS-based)

macOS

• macOS 11 (Big Sur) and later
• Apple Silicon (M1, M2, M3, M4) — native support
• Intel Mac — native support
• Installer format: .dmg
• Minimum system version: 10.15 (Catalina) for runtime compatibility

Linux

• Ubuntu 22.04 LTS and later (recommended)
• Any distribution with a full Desktop GUI (GNOME recommended)
• Installer formats: .AppImage or .deb
• Requires: libwebkit2gtk, libgtk-3, libappindicator3

Technology Stack

RTILA X is built on a modern, high-performance technology stack:

Component	Technology	Purpose
Desktop Shell	Tauri 2.11 (Rust)	Lightweight native window, system tray, process management
Frontend UI	Vue 3.5 + Pinia	Reactive user interface with state management
Automation Engine	Deno + Playwright	Browser automation, command execution, stealth features
Database	PocketBase (embedded)	Local-first data storage for projects, sessions, logs
Browser Extension	CRXJS + Vue 3	Visual Builder, element picker, recording toolbar
Local AI	llama.cpp (bundled)	Runs GGUF models locally for AI-powered building
AI Proxy	OpenRouter API	Cloud AI access for users without local hardware
Tunneling	Cloudflared (bundled)	Public URL generation for remote execution
Relay Engine	Deno (rtila-relay sidecar)	Executes trigger chains and post-processing pipelines

How the Architecture Works

RTILA X uses a multi-process architecture with four main components working together:

1. Tauri Shell (Rust) — The native desktop application window. It manages the UI, spawns and monitors all child processes, handles system tray integration, and provides the bridge between the frontend and the operating system. The Tauri shell spawns the database, engine, relay, and tunnel as managed sidecar processes.

2. PocketBase Database — An embedded, single-file database that starts automatically on first launch. It stores all your projects, sessions, logs, notifications, chat history, browser profiles, global variables, trigger chains, and scheduler configurations. It runs on http://127.0.0.1:8880.

3. Deno Engine (rtila-engine) — The automation workhorse. Written in Deno (a secure JavaScript/TypeScript runtime), it uses Playwright to control browsers. It executes your commands, handles proxies, solves CAPTCHAs, simulates human behavior, and extracts data.

4. CRXJS Extension — A Chrome/Edge extension that provides the Visual Builder interface. When you open a project in the builder, this extension injects a side panel into the browser where you can visually pick elements, configure datasets, record actions, and preview extractions in real-time.

5. Relay Engine (rtila-relay) — A separate Deno process that handles trigger chain execution. When a project run completes, the relay picks up the trigger chain and executes each step sequentially (or in parallel where configured).

6. Cloudflared Tunnel — An optional component that creates a public URL for remote execution, allowing external services to trigger your automations via HTTP.

┌─────────────────────────────────────────────────┐
│              Tauri Shell (Rust)                  │
│  ┌─────────┐  ┌──────────┐  ┌──────────────┐  │
│  │ Vue 3   │  │  Pinia   │  │ System Tray  │  │
│  │   UI    │  │  Stores  │  │  + Updater   │  │
│  └────┬────┘  └────┬─────┘  └──────────────┘  │
│       │             │                           │
│  ┌────▼─────────────▼────────────────────────┐  │
│  │         Process Manager (Rust)             │  │
│  └──┬──────────┬──────────┬──────────┬───────┘  │
│     │          │          │          │           │
│  ┌──▼───┐  ┌──▼───┐  ┌──▼───┐  ┌──▼────┐     │
│  │PocketBase│ │Deno  │  │Deno  │  │Cloud- │     │
│  │  :8880 │  │Engine│  │Relay │  │flared │     │
│  └────────┘  └──┬───┘  └──────┘  └───────┘     │
│                 │                                 │
│            ┌────▼────┐                           │
│            │Playwright│                           │
│            │(Browser) │                           │
│            │+ CRXJS   │                           │
│            │Extension │                           │
│            └─────────┘                           │
└─────────────────────────────────────────────────┘

What’s New in v8.3.2

Version 8.3.2 is a major architectural update focused on AI reliability and performance:

RAG / Qdrant Vector Database — Removed

The local RAG (Retrieval-Augmented Generation) pipeline backed by Qdrant has been completely removed. Testing across hundreds of real-world automation tasks showed that RAG was hurting performance: it added 350–850ms latency per message, consumed 1.5–3 GB of RAM, and frequently retrieved irrelevant or misleading context. The full system prompt and command reference are now injected directly into every AI conversation.

Finetuned Models — Removed

Custom finetuned models (RTILA Lite 1.5 Assistant) have been removed due to maintenance burden (2–4 weeks per update cycle), hallucination issues at Q4 quantization, and insufficient context windows (8K–16K tokens). Any GGUF model from Hugging Face is now supported.

Local LLM Provider — Redesigned

The Local LLM provider now supports two modes:

• GGUF Model (Hugging Face) — Download and run any GGUF model via the bundled llama.cpp server
• Custom API Endpoint — Connect to any OpenAI-compatible API (Ollama, LM Studio, vLLM, etc.)

KV Cache Configuration — New

Full control over llama.cpp’s KV cache for optimal local performance:

• KV Cache Type K/V (F16, Q8_0, Q4_0, F32)
• Flash Attention toggle
• Parallel Slots (1–8)
• No MMap and MLock options

Default Context Increased to 32K

The default context window for local models increased from 8K to 32K tokens, required to fit the full system prompt and complete command reference.

Trigger Chains — New

A completely new system for building multi-step post-processing pipelines with a visual builder, IF/ELSE conditional blocks, parallel execution groups, scheduling, and run history tracking.