langchain-parallel-web

LangChain.js tool integration for Parallel’s web research API. It provides a StructuredTool that researches current events and returns ranked, compressed excerpts from multiple sites based on your objective and search queries.

Installation

npm install langchain-parallel-web @langchain/core

This package, along with the main LangChain packages, depends on @langchain/core. If you are using this package with other LangChain packages, ensure they all depend on the same version of @langchain/core. One way is to add the following to your project’s package.json:

{
  "dependencies": {
    "@langchain/core": "^0.3.0",
    "langchain-parallel-web": "^0.0.1"
  },
  "resolutions": {
    "@langchain/core": "^0.3.0"
  },
  "overrides": {
    "@langchain/core": "^0.3.0"
  },
  "pnpm": {
    "overrides": {
      "@langchain/core": "^0.3.0"
    }
  }
}

Setup

export PARALLEL_API_KEY=your-api-key

Node.js 18+ is required.

Usage

Direct invocation

import { ParallelWebTool } from "langchain-parallel-web";

const tool = new ParallelWebTool({
  // You can also omit this and rely on PARALLEL_API_KEY env var
  apiKey: process.env.PARALLEL_API_KEY,
  // Optional overrides
  // apiUrl: "https://api.parallel.ai/v1beta",
  // config: { max_results: 10, processor: "base", max_chars_per_result: 1500 },
});

const output = await tool.invoke({
  objective: "Summarize the latest impacts of Hurricane Beryl on air travel.",
  search_queries: [
    "Hurricane Beryl flight cancellations",
    "airlines response Beryl"
  ],
});

// Tool returns a JSON string; parse it as needed
const results = JSON.parse(output);
console.log(results);

CommonJS usage:

const { ParallelWebTool } = require("langchain-parallel-web");

With agents (conceptual)

ParallelWebTool is a LangChain StructuredTool, so you can register it with your agent/tool-using chain as you would other tools. Provide clear objectives and a small set of focused queries for best results.

API

Class: ParallelWebTool

• name: parallel_web_search
• description: Research current events; returns ranked, compressed excerpts from multiple sites.

Constructor fields:

• apiKey: string (optional). Defaults to process.env.PARALLEL_API_KEY. Required if env var not set.
• apiUrl: string (optional). Defaults to https://api.parallel.ai/v1beta.
• config (optional):
  • max_results?: number (default 10)
  • processor?: "base" | "pro" (default "base")
  • max_chars_per_result?: number (default 1500)

Input schema (validated):

• objective: string, up to 5000 characters
• search_queries: string[], between 1 and 5 items

Return value:

• A JSON string of the Parallel API response. Parse with JSON.parse to access structured results.

Development

Install dependencies

yarn install
# or: npm install

Build the package

yarn build
# or: npm run build

Run tests

yarn test
# or: npm test

Lint & Format

yarn lint && yarn format
# or: npm run lint && npm run format

Adding new entrypoints

If you add a new file to be exported, either import & re-export it from src/index.ts, or add it to the entrypoints field in config inside langchain.config.js, then rebuild.