Exa Search Tool
Description
Exa is an advanced search API that combines neural search with content processing capabilities. It's particularly effective for in-depth research, academic content, and comprehensive data gathering.
Try it in the Kaiban Board!
Want to see this tool in action? Check out our interactive Kaiban Board! Try it now!
Enhance your agents with:
- Neural Search: Advanced semantic understanding of search queries
- Content Processing: Get full text, summaries, and highlights
- Auto-prompt Enhancement: Automatic query improvement
- Structured Results: Well-organized content with metadata
Installation
First, install the KaibanJS tools package:
npm install @kaibanjs/tools
API Key
Before using the tool, ensure that you have created an API Key at Exa to enable search functionality.
Example
Here's how to use the Exa tool to create a research and writing team:
import { ExaSearch } from '@kaibanjs/tools';
// Configure Exa tool
const exaSearch = new ExaSearch({
apiKey: 'your-exa-api-key',
type: 'neural',
contents: {
text: true,
summary: true,
highlights: true
},
useAutoprompt: true,
limit: 10
});
// Create a research agent
const researcher = new Agent({
name: 'DataMiner',
role: 'Research Specialist',
goal: 'Gather comprehensive information from reliable sources',
background: 'Expert in data collection and research',
tools: [exaSearch]
});
// Create a team
const team = new Team({
name: 'Research Team',
agents: [researcher],
tasks: [/* your tasks */],
inputs: {
topic: 'Your research topic'
}
});
Parameters
apiKey
Required. The API key generated from Exa. Provide your API key directly as a string.type
Optional. The type of search to perform. Options:'neural'
: Semantic search using AI'keyword'
: Traditional keyword-based search Defaults to'neural'
.
contents
Optional. Configure what content to retrieve:text
: Get full text contentsummary
: Get AI-generated summarieshighlights
: Get relevant text highlights
useAutoprompt
Optional. Enable AI query enhancementlimit
Optional. Number of results to return. Default is 10.
We Love Feedback!
Is there something unclear or quirky in the docs? Maybe you have a suggestion or spotted an issue? Help us refine and enhance our documentation by submitting an issue on GitHub. We're all ears!