AI Assistant Overview
The AI Assistant is your conversational partner for building and managing automations. Instead of manually configuring commands and datasets, you simply describe what you want in plain English, and the AI builds it for you.
What Can the AI Do?
Section titled “What Can the AI Do?”- Build automation projects from plain English descriptions
- Analyze web pages to find the right selectors and data structures
- Generate datasets automatically from page screenshots
- Create schedulers for recurring tasks
- Debug failed runs by analyzing logs
- Answer questions about your projects and data
- Handle login walls by prompting you for credentials
Accessing the AI Assistant
Section titled “Accessing the AI Assistant”Click AI Assistant (the home icon) in the sidebar.
Chat Settings
Section titled “Chat Settings”Each chat has configurable settings. Click the gear icon next to the chat title.
| Setting | Description | Default |
|---|---|---|
| Chat Title | Display name for the chat | First 50 chars of first message |
| Project Examples in Context | Number of recent projects to include as examples for the AI (0–49) | 0 |
| Message History | Number of past messages to include in context | Last 7 messages |
How the AI Builds Projects
Section titled “How the AI Builds Projects”- You describe what you want to automate.
- The AI analyzes the target page (if a URL is provided) using the
analyze_target_pagetool. - The AI generates a complete project configuration including commands, datasets, and settings.
- You review the generated configuration.
- You run the project.
AI Tools
Section titled “AI Tools”The AI has access to several tools:
| Tool | Description |
|---|---|
| analyze_target_page | Analyzes a web URL to extract page structure, elements, forms, APIs, sitemaps, and a screenshot |
| navigate_and_analyze | Navigates to a secondary page (e.g., a detail page) for additional analysis |
| request_user_input | Prompts you for credentials or information when the AI encounters a login wall or missing data |
Multi-Page Analysis
Section titled “Multi-Page Analysis”If you ask the AI to scrape data that isn’t on the current page (e.g., “extract product details”), the AI will:
- Analyze the listing page.
- Identify detail page links.
- Use
navigate_and_analyzeto visit a detail page. - Generate the complete extraction configuration.
AI Self-Correction
Section titled “AI Self-Correction”When AI Self-Correction is enabled in project settings:
- If a script fails during execution, the AI analyzes the error.
- It rewrites the problematic code.
- It retries the step (up to 3 times).
- You can see correction attempts in the Live Logs.
Handling Protected Pages
Section titled “Handling Protected Pages”If the AI encounters a login form, CAPTCHA, or 2FA:
- The AI uses the
request_user_inputtool. - A modal dialog appears in RTILA X asking for the required information.
- You enter the credentials.
- The AI uses them to authenticate and continues building the automation.
Chat Context
Section titled “Chat Context”The AI uses the following context when generating responses:
- System prompt — Base instructions for the AI
- Project examples — Recent project configurations (if configured)
- RAG knowledge base — Relevant training data retrieved from the local vector database
- Current project — The project associated with the chat (if any)
- Message history — Previous messages in the conversation
Chatting About a Specific Project
Section titled “Chatting About a Specific Project”To chat about a specific project:
- Go to Automation Projects.
- Select the project.
- Click Chat AI in the ribbon.
- A new chat opens, linked to that project. The AI has full context about the project.
Keyboard Shortcuts
Section titled “Keyboard Shortcuts”| Shortcut | Action |
|---|---|
| Enter | Send message |
| Shift+Enter | New line in message |