agsamantha/node_modules/@langchain/community/dist/document_loaders/web/serpapi.js

117 lines
4.1 KiB
JavaScript
Raw Permalink Normal View History

2024-10-02 15:15:21 -05:00
import { Document } from "@langchain/core/documents";
import { getEnvironmentVariable } from "@langchain/core/utils/env";
import { BaseDocumentLoader } from "@langchain/core/document_loaders/base";
/**
* Class representing a document loader for loading search results from
* the SerpAPI. It extends the BaseDocumentLoader class.
* @example
* ```typescript
* const loader = new SerpAPILoader({ q: "{query}", apiKey: "{apiKey}" });
* const docs = await loader.load();
* ```
*/
export class SerpAPILoader extends BaseDocumentLoader {
constructor(params) {
super();
Object.defineProperty(this, "apiKey", {
enumerable: true,
configurable: true,
writable: true,
value: void 0
});
Object.defineProperty(this, "searchQuery", {
enumerable: true,
configurable: true,
writable: true,
value: void 0
});
const { apiKey = getEnvironmentVariable("SERPAPI_API_KEY"), q } = params;
if (!apiKey) {
throw new Error("SerpAPI API key not set. You can set it as SERPAPI_API_KEY in your .env file, or pass it to SerpAPI.");
}
this.apiKey = apiKey;
this.searchQuery = q;
}
/**
* Builds the URL for the SerpAPI search request.
* @returns The URL for the search request.
*/
buildUrl() {
const params = new URLSearchParams();
params.append("api_key", this.apiKey);
params.append("q", this.searchQuery);
return `https://serpapi.com/search?${params.toString()}`;
}
/**
* Extracts documents from the provided output.
* @param output - The output to extract documents from.
* @param responseType - The type of the response to extract documents from.
* @returns An array of Documents.
*/
extractDocuments(output, responseType) {
const documents = [];
const results = Array.isArray(output) ? output : [output];
for (const result of results) {
const pageContent = JSON.stringify(result);
const metadata = {
source: "SerpAPI",
responseType,
};
documents.push(new Document({ pageContent, metadata }));
}
return documents;
}
/**
* Processes the response data from the SerpAPI search request and converts it into an array of Documents.
* @param data - The response data from the SerpAPI search request.
* @returns An array of Documents.
*/
processResponseData(data) {
const documents = [];
const responseTypes = [
"answer_box",
"sports_results",
"shopping_results",
"knowledge_graph",
"organic_results",
];
for (const responseType of responseTypes) {
if (responseType in data) {
documents.push(...this.extractDocuments(data[responseType], responseType));
}
}
return documents;
}
/**
* Fetches the data from the provided URL and returns it as a JSON object.
* If an error occurs during the fetch operation, an exception is thrown with the error message.
* @param url - The URL to fetch data from.
* @returns A promise that resolves to the fetched data as a JSON object.
* @throws An error if the fetch operation fails.
*/
async fetchData(url) {
const response = await fetch(url);
const data = await response.json();
if (data.error) {
throw new Error(`Failed to load search results from SerpAPI due to: ${data.error}`);
}
return data;
}
/**
* Loads the search results from the SerpAPI.
* @returns An array of Documents representing the search results.
* @throws An error if the search results could not be loaded.
*/
async load() {
const url = this.buildUrl();
const data = await this.fetchData(url);
try {
return this.processResponseData(data);
}
catch (error) {
console.error(error);
throw new Error(`Failed to process search results from SerpAPI: ${error}`);
}
}
}