Tavily

Quick Start

Install the plugin:

pnpm install @corsair-dev/tavily

Add the Tavily plugin to your Corsair instance:

corsair.ts

import { createCorsair } from "corsair";
import { tavily } from "@corsair-dev/tavily";

export const corsair = createCorsair({
    plugins: [
        tavily(),
    ],
});

Once configured, you can search the web:

// Search the web
const results = await corsair.tavily.api.search({
    query: "latest AI research papers 2024",
});

Authentication

Supported Auth Types

The Tavily plugin supports:

api_key (default) - Use a Tavily API key

Default Auth Type

If no authType is specified, the plugin defaults to api_key.

Configuring API Key Authentication

Store credentials with the CLI:

pnpm corsair setup --tavily api_key=your-api-key

Alternatively, provide credentials directly in the config:

corsair.ts

tavily({
    key: process.env.TAVILY_API_KEY,
})

See Authentication for details on managing credentials.

Options

Option	Type	Description
`authType`	`'api_key'`	Authentication method (defaults to `'api_key'`)
`key`	`string`	API key (optional, uses database if not provided)
`hooks`	`object`	Endpoint hooks for custom logic
`errorHandlers`	`object`	Custom error handlers
`permissions`	`object`	Permission configuration for AI agent access

Hooks

tavily({
    hooks: {
        search: {
            before: async (ctx, input) => {
                console.log("Searching for:", input.query);
                return { ctx, input };
            },
        },
    },
})

See Hooks for complete documentation.

Error Handling

The plugin includes built-in error handlers for common scenarios. For complete documentation, see the Error Handlers reference.

Usage

Accessing the API

// Basic search
const results = await corsair.tavily.api.search({
    query: "TypeScript best practices 2024",
});

// Advanced search with options
const detailed = await corsair.tavily.api.search({
    query: "React server components tutorial",
    search_depth: "advanced", 
    max_results: 10,
    include_answer: true,
});

See API Endpoints for the complete reference.

Multi-Tenancy

const tenant = corsair.withTenant("user-123");

const results = await tenant.tavily.api.search({
    query: "machine learning news",
});

Examples

Example 1: AI Research Assistant

inngest/functions.ts

export const researchTopic = inngest.createFunction(
    { id: "research-topic" },
    { event: "app/research-requested" },
    async ({ event }) => {
        const tenant = corsair.withTenant(event.data.tenantId);

        const results = await tenant.tavily.api.search({
            query: event.data.topic, 
            search_depth: "advanced",
            include_answer: true, 
            max_results: 5,
        });

        return {
            answer: results.data.answer,
            sources: results.data.results.map(r => r.url),
        };
    }
);

Example 2: Real-time News Aggregation

inngest/functions.ts

export const fetchNewsDigest = inngest.createFunction(
    { id: "fetch-news-digest" },
    { cron: "0 9 * * *" },
    async () => {
        const topics = ["AI", "TypeScript", "web development"];

        const searches = await Promise.all(
            topics.map(topic =>
                corsair.tavily.api.search({
                    query: `${topic} news today`, 
                    topic: "news",
                    time_range: "day",
                    max_results: 3,
                })
            )
        );

        return searches.map((s, i) => ({
            topic: topics[i],
            results: s.data.results,
        }));
    }
);

Tavily

On this page