Skip to main content

​
Built-in Tools

Complete reference for all built-in tools available in Xagent, organized by category.

​
Basic Tools

Search the internet for current information. Capabilities:
  • Find recent news and information
  • Research topics online
  • Get current data
  • Fact-check statements
Parameters:
  • query (required) - Search query
  • num_results (optional) - Number of results to return (default: 10)
  • include_content (optional) - Whether to include full page content
Returns:
  • Search results with titles, links, and snippets
  • Full page content (if requested)
  • Source information
Uses Google Custom Search API and Zhipu AI search

​
python_executor

Execute Python code safely for data analysis and computation. Capabilities:
  • Run Python code
  • Data analysis with pandas
  • Visualization with matplotlib
  • Mathematical computations
  • Data transformations
Parameters:
  • code (required) - Python code to execute
  • libraries - Available libraries: pandas, numpy, matplotlib, etc.
Returns:
  • Output from code execution
  • Error messages if code fails
  • Displayed figures (saved as images)
Security: Executes in isolated environment with syntax checking

​
Knowledge Tools

​
list_knowledge_bases

List all available knowledge bases. Capabilities:
  • See all knowledge bases in your workspace
  • Get document and chunk counts
  • Check knowledge base status
Parameters:
  • None (automatically uses accessible knowledge bases)
Returns:
  • List of knowledge base names
  • Document counts
  • Chunk counts
  • Embedding model information

​
search_knowledge_base

Search for relevant documents in your knowledge bases. Capabilities:
  • Find relevant information
  • Semantic search with embeddings
  • Keyword search
  • Hybrid search (semantic + keyword)
Parameters:
  • query (required) - Search query
  • collections (optional) - Specific knowledge bases to search
  • search_type (optional) - “dense”, “sparse”, or “hybrid” (default: “hybrid”)
  • top_k (optional) - Maximum results per knowledge base (default: 5)
  • min_score (optional) - Minimum relevance score (default: 0.3)
Returns:
  • Relevant document chunks
  • Relevance scores
  • Source document names
  • Section/page references

​
File Tools

​
read_file

Read content from text files. Capabilities:
  • Read text files (TXT, MD, CSV, JSON, etc.)
  • Parse PDF documents
  • Extract text from DOCX
  • Read Excel files
Parameters:
  • file_path (required) - Path to file in workspace
Returns:
  • File content as text
  • Document structure for complex formats
  • Metadata (page count, sections, etc.)

​
write_file

Write content to files. Capabilities:
  • Create new text files
  • Save results
  • Generate reports
  • Export data
Parameters:
  • file_path (required) - Path where to save the file
  • content (required) - Content to write
Returns:
  • Success status
  • Saved file path
  • File size

​
list_files

List files and directories in workspace. Capabilities:
  • Browse workspace structure
  • Find files by pattern
  • Check directory contents
Parameters:
  • directory_path (optional) - Directory to list (default: workspace root)
  • pattern (optional) - File pattern to match
Returns:
  • List of files and directories
  • File sizes and types
  • Directory structure

​
edit_file

Edit existing file content. Capabilities:
  • Replace text
  • Insert new content
  • Delete sections
  • Multiple replacements at once
Parameters:
  • file_path (required) - Path to file
  • operations (required) - List of edit operations
    • operation - “replace”, “insert”, or “delete”
    • old_text - Text to replace (for replace operation)
    • new_text - New text (for replace/insert operations)
Returns:
  • Success status
  • Number of changes made
  • Updated file content preview

​
document_parser

Parse various document formats. Supported Formats:
  • PDF (.pdf)
  • Word (.doc, .docx)
  • PowerPoint (.pptx)
  • Excel (.xlsx, .xls)
  • Images (.png, .jpg) for OCR
  • HTML (.html, .htm)
Parameters:
  • file_path (required) - Path to document
  • parse_method (optional) - Parsing method (default, pypdf, pdfplumber, etc.)
Returns:
  • Extracted text content
  • Document structure
  • Tables (for spreadsheet files)
  • Metadata

​
Vision Tools

​
understand_images

Analyze images and answer questions about them. Capabilities:
  • Understand image content
  • Answer specific questions
  • Extract information
  • OCR and text recognition
  • Chart and graph analysis
Parameters:
  • images (required) - List of image paths or URLs
  • question (optional) - Specific question about images
Returns:
  • Detailed answer to question
  • Description of image content
  • Detected text (OCR)
  • Object information
Limit: Maximum 10 images per call Requires: Vision model (multimodal LLM)

​
describe_images

Generate detailed descriptions of images. Capabilities:
  • Generate comprehensive image descriptions
  • Identify elements and objects
  • Describe scenes and activities
  • Extract text from images
Parameters:
  • images (required) - List of image paths or URLs
Returns:
  • Detailed textual description
  • Key elements identified
  • Context and setting
Limit: Maximum 10 images per call Requires: Vision model (multimodal LLM)

​
detect_objects

Detect and identify objects in images. Capabilities:
  • Find objects in images
  • Count occurrences
  • Locate positions
  • Object categories
Parameters:
  • images (required) - List of image paths or URLs
  • objects (optional) - Specific objects to detect
Returns:
  • Detected objects list
  • Bounding boxes
  • Confidence scores
  • Object counts
Limit: Maximum 10 images per call Requires: Vision model (multimodal LLM)

​
Image Tools

​
generate_image

Generate images from text descriptions. Capabilities:
  • Create images from text
  • Various styles and formats
  • Automatic prompt optimization
  • Text handling in images
Parameters:
  • prompt (required) - Description of desired image
  • size (optional) - Image size (default: “1024*1024”)
  • model_id (optional) - Specific model to use
Returns:
  • Generated image URL
  • Image file path (saved to workspace)
  • Generation metadata
Requires: Image generation model (OpenAI, DashScope, Xinference)

​
edit_image

Edit and modify existing images. Capabilities:
  • Modify image content
  • Change styles
  • Add/remove elements
  • Combine multiple images
Parameters:
  • image_url (required) - Source image(s) to edit
  • prompt (required) - Description of desired changes
  • model_id (optional) - Specific model to use
Returns:
  • Edited image URL
  • Image file path (saved to workspace)
  • Edit metadata
Requires: Image generation model with edit capability

​
Browser Tools

​
browser_automation

Automate web browser interactions using Playwright. Capabilities:
  • Navigate to websites
  • Click buttons and links
  • Fill forms
  • Extract data
  • Take screenshots
  • Manage multiple tabs
Parameters:
  • actions (required) - List of browser actions
    • action - “goto”, “click”, “fill”, “screenshot”, etc.
    • selector - CSS selector for element
    • value - Value to input or text to find
Returns:
  • Screenshot images
  • Extracted data
  • Action results
  • Page information
Features:
  • Anti-detection settings
  • Multi-tab support
  • Session persistence
  • Error recovery

​
Special Image Tools

​
search_images

Search for images on the web. Capabilities:
  • Find images by query
  • Get image URLs
  • Browse image collections
Parameters:
  • query (required) - Search query for images
  • num_results (optional) - Number of results (default: 10)
Returns:
  • Image URLs
  • Image sources
  • Thumbnail links

​
logo_overlay

Add logo overlay to images. Capabilities:
  • Place logo on images
  • Position control
  • Size adjustment
  • Opacity settings
Parameters:
  • image_url (required) - Base image
  • logo_url (required) - Logo image
  • position (optional) - Logo position (default: “top-right”)
  • opacity (optional) - Logo opacity (default: 0.8)
Returns:
  • Composite image URL
  • Saved file path

​
Tool Requirements

Some tools require specific models or configurations:
ToolRequirement
understand_images, describe_images, detect_objectsVision model (multimodal LLM)
generate_image, edit_imageImage generation model
search_knowledge_baseEmbedding model + knowledge base
web_searchSearch API credentials
browser_automationBrowser automation enabled

​
Best Practices

​
Choosing Tools

  • Web search - For current events and online information
  • Knowledge base - For domain-specific documentation
  • File operations - For working with uploaded files
  • Vision tools - For image analysis and OCR
  • Image generation - For creating visual content
  • Python executor - For data analysis and computation
  • Browser automation - For web scraping and automation

​
Tool Chaining

Xagent can use multiple tools in sequence: 
1. Search web for information
2. Save results to file
3. Analyze data with Python
4. Generate report

​
Error Handling

Tools automatically handle errors:
  • Retry failed operations
  • Provide clear error messages
  • Suggest alternatives
  • Fall back to safe defaults

​
Next Steps