mirror of
https://github.com/simstudioai/sim.git
synced 2026-09-24 15:45:35 +08:00
Waleed Latif
eb5d01956e
* feat[sdk]: added scaffolding for sdk pkg * feat[sdk]: simplified usage of tools for agent block * fix[sdk]: added api key directly into AgentOptions * fix[sdk]: simplified the conditional block by removing the code input since we don't need it * feat[sdk]: added account-level api keys & db table to manage them
99 lines
3.2 KiB
TypeScript
99 lines
3.2 KiB
TypeScript
import { SimStudio } from '../src'
|
|
import { AgentBlock } from '../src/blocks/agent'
|
|
|
|
/**
|
|
* Example showing how to use tools with agent blocks
|
|
*/
|
|
async function usingToolsExample() {
|
|
const simStudio = new SimStudio({
|
|
apiKey: 'your-api-key', // Replace with your actual API key
|
|
})
|
|
|
|
// Create a workflow
|
|
const workflow = simStudio.createWorkflow(
|
|
'Research Assistant',
|
|
'Research a topic and summarize findings'
|
|
)
|
|
|
|
// Create the agent block with tools
|
|
const researchAgentBlock = new AgentBlock({
|
|
model: 'claude-3-opus',
|
|
prompt: 'Research the topic "{{input.topic}}" and provide a comprehensive summary. Use the search tools to gather information.',
|
|
systemPrompt: 'You are a research assistant that can search the web for information and compile comprehensive reports.',
|
|
temperature: 0.5,
|
|
tools: [], // Will be populated with tool references
|
|
apiKey: 'your-agent-api-key' // Now included directly in options
|
|
}).setName('Research Agent')
|
|
|
|
// Reference built-in tools by their IDs
|
|
// The system will automatically transform these references into tool definitions
|
|
researchAgentBlock.data.toolReferences = ['tavily_search', 'serper_search']
|
|
|
|
// Configure tool settings with required parameters
|
|
researchAgentBlock.data.toolSettings = {
|
|
// Each tool gets its required parameters
|
|
tavily_search: {
|
|
apiKey: 'your-tavily-api-key'
|
|
},
|
|
serper_search: {
|
|
apiKey: 'your-serper-api-key'
|
|
}
|
|
}
|
|
|
|
// Add the agent to the workflow
|
|
workflow.addBlock(researchAgentBlock)
|
|
|
|
// For custom tools, users can still define them directly if needed
|
|
// This is equivalent to when a user creates a custom tool in the UI
|
|
const customTool = {
|
|
name: 'findCompany',
|
|
description: 'Find information about a company by name',
|
|
parameters: {
|
|
type: 'object',
|
|
properties: {
|
|
companyName: {
|
|
type: 'string',
|
|
description: 'The name of the company to find information about'
|
|
},
|
|
detailLevel: {
|
|
type: 'string',
|
|
enum: ['basic', 'detailed'],
|
|
description: 'The level of detail to return',
|
|
default: 'basic'
|
|
}
|
|
},
|
|
required: ['companyName']
|
|
}
|
|
}
|
|
|
|
// Add the custom tool to the agent
|
|
if (!researchAgentBlock.data.tools) {
|
|
researchAgentBlock.data.tools = []
|
|
}
|
|
researchAgentBlock.data.tools.push(customTool)
|
|
|
|
console.log('Added tools to the agent')
|
|
|
|
// Connect the starter block to the agent
|
|
const starterBlock = workflow.getStarterBlock()
|
|
workflow.connect(starterBlock.id, researchAgentBlock.id)
|
|
|
|
// Build the workflow
|
|
const builtWorkflow = workflow.build()
|
|
console.log('Workflow built successfully with search tools')
|
|
|
|
// In a real scenario, you would save and deploy the workflow
|
|
console.log('To execute this workflow, you would:')
|
|
console.log('1. Save the workflow: simStudio.saveWorkflow(workflow)')
|
|
console.log('2. Deploy the workflow: simStudio.deployWorkflow(workflowId)')
|
|
console.log('3. Execute the workflow: simStudio.executeWorkflow(workflowId, { topic: "AI ethics" })')
|
|
|
|
return builtWorkflow
|
|
}
|
|
|
|
// Run the example
|
|
if (require.main === module) {
|
|
usingToolsExample().catch(console.error)
|
|
}
|
|
|
|
export default usingToolsExample
|