multitools

Multi-Tool Intelligent Assistant Example

This example demonstrates how to create an intelligent assistant that integrates multiple tools, including calculator, time tool, text processing tool, file operation tool, and web search tool.

Features

🧮 Calculator Tool (calculator)

• Basic Operations: Addition, subtraction, multiplication, division
• Scientific Functions: sqrt(square root), sin(sine), cos(cosine), abs(absolute value)
• Mathematical Constants: pi(pi), e(natural logarithm base)
• Examples:
  • Calculate 123 + 456 * 789
  • Calculate sqrt(16)
  • Calculate sin(30*pi/180)

⏰ Time Tool (time_tool)

• Current Time: Get current date and time
• Date Information: Get current date
• Day Information: Get current day of the week
• Timestamp: Get Unix timestamp
• Examples:
  • What time is it now?
  • What day of the week is today?
  • Get current timestamp

📝 Text Processing Tool (text_tool)

• Case Conversion: Convert to uppercase or lowercase
• Character Statistics: Count text length and word count
• Text Reversal: Reverse text content
• Examples:
  • Convert 'Hello World' to uppercase
  • Count characters in 'Hello World'
  • Reverse text 'Hello World'

📁 File Operation Tool (file_tool)

• Read Files: Read file contents
• Write Files: Create or write files
• List Directory: View directory contents
• Check Existence: Check if file exists
• Examples:
  • Read README.md file
  • Create a test file in current directory
  • List all files in current directory

🔍 Web Search Tool (duckduckgo_search)

• Entity Search: Search for information about people, companies, locations, etc.
• Definition Queries: Find definitions of concepts and terms
• Historical Information: Find historical facts and data
• Examples:
  • Search for information about Steve Jobs
  • Find information about Tesla company
  • What is photosynthesis?

Usage

1. Environment Setup

Make sure you have set up the relevant environment variables and dependencies:

# Set environment variables like API keys
export OPENAI_API_KEY="your-api-key"

2. Run the Example

cd trpc.group/trpc-go/trpc-agent-go/examples/multi_tools
go run main.go

3. Specify Model

go run main.go -model="gpt-4"

4. Interactive Usage

After the program starts, you can:

• Enter various questions and requests
• Observe how the assistant selects and uses different tools
• Enter exit to quit the program

Example Conversation

🚀 Multi-Tool Intelligent Assistant Demo
Model: deepseek-chat
Enter 'exit' to end the conversation
Available tools: calculator, time_tool, text_tool, file_tool, duckduckgo_search
============================================================
✅ Multi-tool intelligent assistant is ready! Session ID: multi-tool-session-1703123456

💡 Try asking these questions:
   [Calculator] Calculate 123 + 456 * 789
   [Calculator] Calculate the square root of pi
   [Time] What time is it now?
   [Time] What day of the week is today?
   [Text] Convert 'Hello World' to uppercase
   [Text] Count characters in 'Hello World'
   [File] Read the README.md file
   [File] Create a test file in the current directory
   [Search] Search for information about Steve Jobs
   [Search] Find information about Tesla company

👤 User: Calculate 100 + 200 * 3
🔧 Tool calls:
   🧮 calculator (ID: call_abc123)
     Arguments: {"expression":"100 + 200 * 3"}

⚡ Executing...
✅ Tool result (ID: call_abc123): {"expression":"100 + 200 * 3","result":700,"message":"Calculation result: 700"}

🤖 Assistant: According to mathematical operation rules, multiplication is performed first, then addition:
100 + 200 * 3 = 100 + 600 = 700

The calculation result is 700.

Tool Design Principles

1. Security

• File operation tools are restricted to current directory and subdirectories
• Prevent path traversal attacks
• Limit file reading content length

2. User Experience

• English interface and prompts
• Clear tool call visualization
• Rich usage examples and help information

3. Extensibility

• Modular tool design
• Unified tool interface
• Easy to add new tools

Extending New Tools

To add new tools, follow these steps:

1. Define Request and Response Structures

type myToolRequest struct {
    Input string `json:"input" jsonschema:"description=Input description"`
}

type myToolResponse struct {
    Output string `json:"output"`
    Status string `json:"status"`
}

2. Implement Tool Function

func myToolHandler(req myToolRequest) myToolResponse {
    // Implement tool logic
    return myToolResponse{
        Output: "Processing result",
        Status: "Success",
    }
}

3. Create Tool Instance

func createMyTool() tool.CallableTool {
    return function.NewFunctionTool(
        myToolHandler,
        function.WithName("my_tool"),
        function.WithDescription("Tool description"),
    )
}

4. Register Tool

tools := []tool.Tool{
    createMyTool(),
    // Other tools...
}

Notes

1. API Limitations: Some tools may require network access or API keys
2. Performance Considerations: Large file operations may affect performance
3. Error Handling: Tool call failures will return error messages
4. Security: File operations are protected by path restrictions

Technical Architecture

• Framework: trpc-agent-go
• Models: Supports various OpenAI-compatible models
• Tool System: Function calling-based tool invocation
• Streaming: Supports streaming responses and real-time interaction

License

This example follows the project's license terms.