Glossary

This glossary defines all key terms used throughout the RTILA X documentation. Understanding these terms will help you navigate the documentation and use RTILA X more effectively.

Complete Glossary

Term	Definition
AI Assistant	RTILA X’s built-in AI chat interface that helps you build automations by describing what you want in plain English.
Automation	A sequence of browser actions (clicks, fills, navigations) performed automatically by RTILA X.
Browser Profile	An isolated browser environment with its own cookies, cache, extensions, and stealth settings.
Browser Profile ID	A unique identifier for a browser profile, used to assign profiles to projects.
CAPTCHA	A challenge-response test used by websites to verify that a user is human. RTILA X can solve CAPTCHAs automatically using 2Captcha.
Checkpoint	A saved snapshot of automation progress that allows resuming after interruption.
Cloudflare Turnstile	A JavaScript-based CAPTCHA challenge used by Cloudflare-protected websites.
Command	A single action in an automation (e.g., click, fill, goto, extract_data).
Command Flow	The sequence of commands in a project, executed from top to bottom.
CRXJS	The browser extension framework used to build RTILA X’s Visual Builder extension.
CSS Selector	A pattern that identifies HTML elements using CSS syntax (e.g., .product-card, #main-content).
Dataset	A definition of what data to extract from a web page, including item selectors and property definitions.
Deno	The JavaScript/TypeScript runtime used by RTILA X’s automation engine.
Element Picker	A visual tool in the Visual Builder that lets you select elements on a page by clicking.
GGUF	A binary model format used by llama.cpp for local AI model inference.
Headless Mode	A browser mode where the browser runs without a visible window.
Humanoid Behavior	Simulated human-like mouse movements, typing, and interactions to avoid bot detection.
Humanoid Sensitivity	The level of randomization applied to humanoid behavior: Low, Medium, or High.
Item Selector	The CSS or XPath selector that identifies the repeating container element in a dataset.
KV Cache	Key-Value cache. Stores the attention state of processed tokens in AI models for faster subsequent processing.
License Key	A unique key that activates RTILA X and unlocks features.
Local LLM	A locally-run AI model using llama.cpp. No internet connection needed after download.
MCP (Model Context Protocol)	An open standard protocol that allows AI models to interact with external tools. The RTILA engine binary (rtila-engine) doubles as an MCP server when started with --mcp, exposing 17 tools for page analysis, project execution, and data access to any MCP-compatible client.
run_id	A unique identifier returned by async MCP browser tools. Used to poll rtila_get_run_status for completion and rtila_get_run_result for results.
Async Contract	The three-step pattern used by MCP browser tools: (1) start the work and receive a run_id, (2) poll rtila_get_run_status until finished, (3) get the result.
Agent Skills	Markdown files that teach Claude Code how to drive the RTILA MCP server correctly, encoding the async polling loop, report interpretation, and diagnosis patterns.
OpenRouter	A service that provides access to hundreds of AI models through a single API key.
Playwright	The browser automation framework used by RTILA X.
PocketBase	The lightweight backend database framework used by RTILA X.
Profile	A user workspace within RTILA X with its own projects, data, and settings.
Project	The fundamental unit of work in RTILA X, containing commands, datasets, settings, and triggers.
Property	A single field in a dataset that defines what data to extract (e.g., title, price, URL).
Property Type	The type of data to extract: text, html, attribute, property, count, page_url, selector_path, list, or index.
Proxy	An intermediary server that routes your internet traffic through a different IP address.
Proxy Level	How proxies are applied: Context Level (per-session) or Browser Level (shared).
Queue	The job queue system that manages automation execution.
Recording	The process of capturing your browser actions as automation commands.
Relay Engine	The Deno process that executes trigger chains and post-processing pipelines.
Runtime Variable	A variable created during automation execution, available to subsequent commands.
Script Transformation	A custom JavaScript transformation applied to extracted data during extraction.
Selector	A pattern that identifies HTML elements on a page (CSS, XPath, text, or variable).
Session	One complete execution of a project, including start time, end time, and status.
Sitemap	An XML file that lists all pages on a website, used for URL discovery.
Standalone Bot	A packaged automation that runs independently without RTILA X installed.
Stealth	Features that help your automation avoid detection by anti-bot systems.
Tauri	The framework used to build RTILA X’s desktop application shell.
Trigger	A post-execution action that sends scraped data to an external destination.
Trigger Chain	A reusable multi-step post-processing pipeline with conditional logic and scheduling.
Trigger Chain Run	One execution of a trigger chain, with status tracking and retry support.
Trigger Chain Schedule	A schedule that automatically runs a trigger chain at specified intervals.
Transformation	A processing step that modifies extracted data (trim, replace, cast, etc.).
Variable	A named value that can be used in commands and datasets.
Variable Substitution	The process of replacing ${variable_name} placeholders with actual values.
Visual Builder	The browser-based interface for building and editing automations.
WAF (Web Application Firewall)	A security system that protects websites from automated access.
Webhook	An HTTP callback that triggers an action when an event occurs.

Removed Terms (v8.3.2)

The following terms were used in previous versions of RTILA X but have been removed in v8.3.2. They are documented here for historical reference.

Term	Previous Definition	Reason for Removal
RAG (Retrieval-Augmented Generation)	A system that stored RTILA documentation in a vector database and retrieved relevant context for AI requests.	Removed because it reduced accuracy, added latency, and consumed significant resources. Replaced by full-context injection.
Qdrant	The vector database used to store RAG embeddings.	Removed along with the RAG system. No longer needed.
Embedding	A numerical representation of text used for vector search in RAG.	Removed along with the RAG system. No longer needed.
Finetuned Model	Custom AI models trained specifically for RTILA X command generation (e.g., RTILA Lite 1.5 Assistant).	Removed due to maintenance burden, hallucination issues, and insufficient context windows. Replaced by any GGUF model from Hugging Face.
Custom Provider	A previous AI provider option that allowed connecting to custom OpenAI-compatible endpoints.	Merged into the unified Local LLM provider as the “Custom API Endpoint” mode.
Top K	A RAG setting that controlled how many retrieved chunks were injected into AI context.	Removed along with the RAG system. No longer needed.
Similarity Threshold	A RAG setting that controlled the minimum similarity score for retrieved chunks.	Removed along with the RAG system. No longer needed.
RTILA Lite 1.5 Assistant	The previous finetuned AI model (9B parameters) optimized for RTILA command generation.	Removed. Replaced by any GGUF model from Hugging Face.
RTILA Lite 1.5 Assistant Q4	The Q4 quantized version of the finetuned model.	Removed along with the finetuned model.
RTILA Lite 1.5 Assistant Q8	The Q8 quantized version of the finetuned model.	Removed along with the finetuned model.

Migration Notes

If you see references to removed terms in old documentation or tutorials:

Old Term	New Equivalent
RAG	Full-context injection (automatic, no configuration needed)
Qdrant	No longer needed. Removed.
Finetuned Model	Any GGUF model from Hugging Face
Custom Provider	Local LLM → Custom API Endpoint mode
Top K	No longer needed. Full context is always injected.
Similarity Threshold	No longer needed. Full context is always injected.
RTILA Lite 1.5 Assistant	Use any GGUF model (e.g., Qwen 2.5 7B, Llama 3.1 8B)