Docs Under Construction

Don't panic! Our AI agents are working hard to build something amazing.

We're putting the finishing touches on our documentation. Check back soon!

Back to Website
Building in progress...
AgentFlare
GitHubCommunityGet Started

API Reference

Complete reference documentation for the AgentFlare API.

Agent Class

The Agent class is the core of AgentFlare. It represents an autonomous AI entity that can execute tasks.

Constructor

new Agent(config: AgentConfig)

Parameters

  • config.name (string): Unique identifier for the agent
  • config.model (string): LLM model to use (e.g., 'gpt-4', 'claude-3')
  • config.tools (string[]): Array of tool names to enable
  • config.systemPrompt (string): System prompt defining agent behavior
  • config.memory (MemoryConfig): Optional memory configuration

Methods

run()

Execute a task with the agent.

async run(input: string): Promise<AgentResponse>

Example:

const response = await agent.run('Analyze this dataset')
console.log(response.output)

addTool()

Add a custom tool to the agent.

addTool(tool: Tool): void

Example:

agent.addTool(customWeatherTool)

Tool Class

Create custom tools for your agents.

Constructor

new Tool(config: ToolConfig)

Parameters

  • config.name (string): Tool identifier
  • config.description (string): What the tool does
  • config.parameters (ParameterSchema): Input parameter schema
  • config.execute (Function): Tool execution function

Example

const searchTool = new Tool({
  name: 'web-search',
  description: 'Search the web for information',
  parameters: {
    query: { type: 'string', required: true }
  },
  execute: async ({ query }) => {
    const results = await searchAPI(query)
    return results
  }
})

Memory

Configure agent memory for context retention.

Types

type MemoryConfig = {
  type: 'conversation' | 'vector' | 'hybrid'
  maxTokens?: number
  vectorStore?: VectorStoreConfig
}

Example

const agent = new Agent({
  name: 'Assistant',
  model: 'gpt-4',
  memory: {
    type: 'hybrid',
    maxTokens: 8000,
    vectorStore: {
      provider: 'pinecone',
      index: 'agent-memory'
    }
  }
})

Workflows

Orchestrate multiple agents in complex workflows.

Creating a Workflow

import { Workflow } from 'agentflare'
 
const workflow = new Workflow({
  name: 'DataPipeline',
  agents: [dataAgent, analysisAgent, reportAgent],
  steps: [
    { agent: 'dataAgent', input: 'fetch data' },
    { agent: 'analysisAgent', input: '{{dataAgent.output}}' },
    { agent: 'reportAgent', input: '{{analysisAgent.output}}' }
  ]
})
 
const result = await workflow.execute()

Error Handling

AgentFlare provides comprehensive error handling.

try {
  const response = await agent.run(input)
} catch (error) {
  if (error instanceof AgentError) {
    console.error('Agent error:', error.message)
  } else if (error instanceof ToolError) {
    console.error('Tool error:', error.toolName, error.message)
  }
}