langchain 0.3.31 → 1.0.0-alpha.1

package/dist/tools/webbrowser.d.cts

@@ -0,0 +1,69 @@
+import { BaseLanguageModelInterface } from "@langchain/core/language_models/base";
+import { CallbackManager, CallbackManagerForToolRun } from "@langchain/core/callbacks/manager";
+import { TextSplitter } from "@langchain/textsplitters";
+import { Tool, ToolParams } from "@langchain/core/tools";
+import { EmbeddingsInterface } from "@langchain/core/embeddings";
+
+//#region src/tools/webbrowser.d.ts
+declare const parseInputs: (inputs: string) => [string, string];
+declare const getText: (html: string, baseUrl: string, summary: boolean) => string;
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+type Headers = Record<string, any>;
+/**
+ * Configuration options for fetch requests, similar to axios config but for fetch
+ */
+interface RequestConfig extends Omit<RequestInit, "headers"> {
+  withCredentials?: boolean;
+}
+/**
+ * Defines the arguments that can be passed to the WebBrowser constructor.
+ * It extends the ToolParams interface and includes properties for a
+ * language model, embeddings, HTTP headers, an Axios configuration, a
+ * callback manager, and a text splitter.
+ */
+interface WebBrowserArgs extends ToolParams {
+  model: BaseLanguageModelInterface;
+  embeddings: EmbeddingsInterface;
+  headers?: Headers;
+  requestConfig?: RequestConfig;
+  /** @deprecated */
+  callbackManager?: CallbackManager;
+  textSplitter?: TextSplitter;
+}
+/**
+ * A class designed to interact with web pages, either to extract
+ * information from them or to summarize their content. It uses the native
+ * fetch API to send HTTP requests and the cheerio library to parse the
+ * returned HTML.
+ * @example
+ * ```typescript
+ * const browser = new WebBrowser({
+ *   model: new ChatOpenAI({ model: "gpt-4o-mini", temperature: 0 }),
+ *   embeddings: new OpenAIEmbeddings({}),
+ * });
+ * const result = await browser.invoke("https:exampleurl.com");
+ * ```
+ */
+declare class WebBrowser extends Tool {
+  static lc_name(): string;
+  get lc_namespace(): string[];
+  private model;
+  private embeddings;
+  private headers;
+  private requestConfig;
+  private textSplitter;
+  constructor({
+    model,
+    headers,
+    embeddings,
+    requestConfig,
+    textSplitter
+  }: WebBrowserArgs);
+  /** @ignore */
+  _call(inputs: string, runManager?: CallbackManagerForToolRun): Promise<string>;
+  name: string;
+  description: string;
+}
+//#endregion
+export { RequestConfig, WebBrowser, WebBrowserArgs, getText, parseInputs };
+//# sourceMappingURL=webbrowser.d.cts.map

package/dist/tools/webbrowser.d.cts.map

@@ -0,0 +1 @@
+{"version":3,"file":"webbrowser.d.cts","names":["BaseLanguageModelInterface","EmbeddingsInterface","CallbackManager","CallbackManagerForToolRun","Tool","ToolParams","TextSplitter","parseInputs","getText","Headers","Record","RequestConfig","RequestInit","Omit","WebBrowserArgs","WebBrowser","model","headers","embeddings","requestConfig","textSplitter","Promise"],"sources":["../../src/tools/webbrowser.d.ts"],"sourcesContent":["import type { BaseLanguageModelInterface } from \"@langchain/core/language_models/base\";\nimport type { EmbeddingsInterface } from \"@langchain/core/embeddings\";\nimport { CallbackManager, CallbackManagerForToolRun } from \"@langchain/core/callbacks/manager\";\nimport { Tool, ToolParams } from \"@langchain/core/tools\";\nimport { TextSplitter } from \"@langchain/textsplitters\";\nexport declare const parseInputs: (inputs: string) => [string, string];\nexport declare const getText: (html: string, baseUrl: string, summary: boolean) => string;\n// eslint-disable-next-line @typescript-eslint/no-explicit-any\ntype Headers = Record<string, any>;\n/**\n * Configuration options for fetch requests, similar to axios config but for fetch\n */\nexport interface RequestConfig extends Omit<RequestInit, \"headers\"> {\n    withCredentials?: boolean;\n}\n/**\n * Defines the arguments that can be passed to the WebBrowser constructor.\n * It extends the ToolParams interface and includes properties for a\n * language model, embeddings, HTTP headers, an Axios configuration, a\n * callback manager, and a text splitter.\n */\nexport interface WebBrowserArgs extends ToolParams {\n    model: BaseLanguageModelInterface;\n    embeddings: EmbeddingsInterface;\n    headers?: Headers;\n    requestConfig?: RequestConfig;\n    /** @deprecated */\n    callbackManager?: CallbackManager;\n    textSplitter?: TextSplitter;\n}\n/**\n * A class designed to interact with web pages, either to extract\n * information from them or to summarize their content. It uses the native\n * fetch API to send HTTP requests and the cheerio library to parse the\n * returned HTML.\n * @example\n * ```typescript\n * const browser = new WebBrowser({\n *   model: new ChatOpenAI({ model: \"gpt-4o-mini\", temperature: 0 }),\n *   embeddings: new OpenAIEmbeddings({}),\n * });\n * const result = await browser.invoke(\"https:exampleurl.com\");\n * ```\n */\nexport declare class WebBrowser extends Tool {\n    static lc_name(): string;\n    get lc_namespace(): string[];\n    private model;\n    private embeddings;\n    private headers;\n    private requestConfig;\n    private textSplitter;\n    constructor({ model, headers, embeddings, requestConfig, textSplitter }: WebBrowserArgs);\n    /** @ignore */\n    _call(inputs: string, runManager?: CallbackManagerForToolRun): Promise<string>;\n    name: string;\n    description: string;\n}\nexport {};\n"],"mappings":";;;;;;;cAKqBO;cACAC;AADrB;AACA,KAEKC,OAAAA,GAAUC,MAF0E,CAAA,MAAA,EAAA,GAAA,CAAA;AAAC;AAM1F;;AAA4CE,UAA3BD,aAAAA,SAAsBE,IAAKD,CAAAA,WAAAA,EAAAA,SAAAA,CAAAA,CAAAA;EAAW,eAAhBC,CAAAA,EAAAA,OAAAA;AAAI;AAS3C;;;;;;AAMsBX,UANLY,cAAAA,SAAuBT,UAMlBH,CAAAA;EAAe,KAClBI,EANRN,0BAMQM;EAAY,UAPSD,EAExBJ,mBAFwBI;EAAU,OAAA,CAAA,EAGpCI,OAHoC;EAuB7BM,aAAU,CAAA,EAnBXJ,aAmBW;EAAA;EAAA,eAQbK,CAAAA,EAzBId,eAyBJc;EAAK,YAAEC,CAAAA,EAxBNX,YAwBMW;;;;;;;;AARmB;;;;;;;;cAAvBF,UAAAA,SAAmBX,IAAAA;;;;;;;;;;;;;;KAQqCU;;qCAEtCX,4BAA4BkB"}

package/dist/tools/webbrowser.d.ts

@@ -1,31 +1,39 @@
-import type { BaseLanguageModelInterface } from "@langchain/core/language_models/base";
-import type { EmbeddingsInterface } from "@langchain/core/embeddings";
-import { AxiosRequestConfig } from "axios";
-import { CallbackManager, CallbackManagerForToolRun } from "@langchain/core/callbacks/manager";
 import { Tool, ToolParams } from "@langchain/core/tools";
-import { TextSplitter } from "../text_splitter.js";
-export declare const parseInputs: (inputs: string) => [string, string];
-export declare const getText: (html: string, baseUrl: string, summary: boolean) => string;
+import { TextSplitter } from "@langchain/textsplitters";
+import { CallbackManager, CallbackManagerForToolRun } from "@langchain/core/callbacks/manager";
+import { BaseLanguageModelInterface } from "@langchain/core/language_models/base";
+import { EmbeddingsInterface } from "@langchain/core/embeddings";
+
+//#region src/tools/webbrowser.d.ts
+declare const parseInputs: (inputs: string) => [string, string];
+declare const getText: (html: string, baseUrl: string, summary: boolean) => string;
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
 type Headers = Record<string, any>;
+/**
+ * Configuration options for fetch requests, similar to axios config but for fetch
+ */
+interface RequestConfig extends Omit<RequestInit, "headers"> {
+  withCredentials?: boolean;
+}
 /**
  * Defines the arguments that can be passed to the WebBrowser constructor.
  * It extends the ToolParams interface and includes properties for a
  * language model, embeddings, HTTP headers, an Axios configuration, a
  * callback manager, and a text splitter.
  */
-export interface WebBrowserArgs extends ToolParams {
-    model: BaseLanguageModelInterface;
-    embeddings: EmbeddingsInterface;
-    headers?: Headers;
-    axiosConfig?: Omit<AxiosRequestConfig, "url">;
-    /** @deprecated */
-    callbackManager?: CallbackManager;
-    textSplitter?: TextSplitter;
+interface WebBrowserArgs extends ToolParams {
+  model: BaseLanguageModelInterface;
+  embeddings: EmbeddingsInterface;
+  headers?: Headers;
+  requestConfig?: RequestConfig;
+  /** @deprecated */
+  callbackManager?: CallbackManager;
+  textSplitter?: TextSplitter;
 }
 /**
  * A class designed to interact with web pages, either to extract
- * information from them or to summarize their content. It uses the axios
- * library to send HTTP requests and the cheerio library to parse the
+ * information from them or to summarize their content. It uses the native
+ * fetch API to send HTTP requests and the cheerio library to parse the
  * returned HTML.
  * @example
  * ```typescript
@@ -36,18 +44,26 @@ export interface WebBrowserArgs extends ToolParams {
  * const result = await browser.invoke("https:exampleurl.com");
  * ```
  */
-export declare class WebBrowser extends Tool {
-    static lc_name(): string;
-    get lc_namespace(): string[];
-    private model;
-    private embeddings;
-    private headers;
-    private axiosConfig;
-    private textSplitter;
-    constructor({ model, headers, embeddings, axiosConfig, textSplitter, }: WebBrowserArgs);
-    /** @ignore */
-    _call(inputs: string, runManager?: CallbackManagerForToolRun): Promise<string>;
-    name: string;
-    description: string;
+declare class WebBrowser extends Tool {
+  static lc_name(): string;
+  get lc_namespace(): string[];
+  private model;
+  private embeddings;
+  private headers;
+  private requestConfig;
+  private textSplitter;
+  constructor({
+    model,
+    headers,
+    embeddings,
+    requestConfig,
+    textSplitter
+  }: WebBrowserArgs);
+  /** @ignore */
+  _call(inputs: string, runManager?: CallbackManagerForToolRun): Promise<string>;
+  name: string;
+  description: string;
 }
-export {};
+//#endregion
+export { RequestConfig, WebBrowser, WebBrowserArgs, getText, parseInputs };
+//# sourceMappingURL=webbrowser.d.ts.map

package/dist/tools/webbrowser.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"webbrowser.d.ts","names":["BaseLanguageModelInterface","EmbeddingsInterface","CallbackManager","CallbackManagerForToolRun","Tool","ToolParams","TextSplitter","parseInputs","getText","Headers","Record","RequestConfig","RequestInit","Omit","WebBrowserArgs","WebBrowser","model","headers","embeddings","requestConfig","textSplitter","Promise"],"sources":["../../src/tools/webbrowser.d.ts"],"sourcesContent":["import type { BaseLanguageModelInterface } from \"@langchain/core/language_models/base\";\nimport type { EmbeddingsInterface } from \"@langchain/core/embeddings\";\nimport { CallbackManager, CallbackManagerForToolRun } from \"@langchain/core/callbacks/manager\";\nimport { Tool, ToolParams } from \"@langchain/core/tools\";\nimport { TextSplitter } from \"@langchain/textsplitters\";\nexport declare const parseInputs: (inputs: string) => [string, string];\nexport declare const getText: (html: string, baseUrl: string, summary: boolean) => string;\n// eslint-disable-next-line @typescript-eslint/no-explicit-any\ntype Headers = Record<string, any>;\n/**\n * Configuration options for fetch requests, similar to axios config but for fetch\n */\nexport interface RequestConfig extends Omit<RequestInit, \"headers\"> {\n    withCredentials?: boolean;\n}\n/**\n * Defines the arguments that can be passed to the WebBrowser constructor.\n * It extends the ToolParams interface and includes properties for a\n * language model, embeddings, HTTP headers, an Axios configuration, a\n * callback manager, and a text splitter.\n */\nexport interface WebBrowserArgs extends ToolParams {\n    model: BaseLanguageModelInterface;\n    embeddings: EmbeddingsInterface;\n    headers?: Headers;\n    requestConfig?: RequestConfig;\n    /** @deprecated */\n    callbackManager?: CallbackManager;\n    textSplitter?: TextSplitter;\n}\n/**\n * A class designed to interact with web pages, either to extract\n * information from them or to summarize their content. It uses the native\n * fetch API to send HTTP requests and the cheerio library to parse the\n * returned HTML.\n * @example\n * ```typescript\n * const browser = new WebBrowser({\n *   model: new ChatOpenAI({ model: \"gpt-4o-mini\", temperature: 0 }),\n *   embeddings: new OpenAIEmbeddings({}),\n * });\n * const result = await browser.invoke(\"https:exampleurl.com\");\n * ```\n */\nexport declare class WebBrowser extends Tool {\n    static lc_name(): string;\n    get lc_namespace(): string[];\n    private model;\n    private embeddings;\n    private headers;\n    private requestConfig;\n    private textSplitter;\n    constructor({ model, headers, embeddings, requestConfig, textSplitter }: WebBrowserArgs);\n    /** @ignore */\n    _call(inputs: string, runManager?: CallbackManagerForToolRun): Promise<string>;\n    name: string;\n    description: string;\n}\nexport {};\n"],"mappings":";;;;;;;cAKqBO;cACAC;AADrB;AACA,KAEKC,OAAAA,GAAUC,MAF0E,CAAA,MAAA,EAAA,GAAA,CAAA;AAAC;AAM1F;;AAA4CE,UAA3BD,aAAAA,SAAsBE,IAAKD,CAAAA,WAAAA,EAAAA,SAAAA,CAAAA,CAAAA;EAAW,eAAhBC,CAAAA,EAAAA,OAAAA;AAAI;AAS3C;;;;;;AAMsBX,UANLY,cAAAA,SAAuBT,UAMlBH,CAAAA;EAAe,KAClBI,EANRN,0BAMQM;EAAY,UAPSD,EAExBJ,mBAFwBI;EAAU,OAAA,CAAA,EAGpCI,OAHoC;EAuB7BM,aAAU,CAAA,EAnBXJ,aAmBW;EAAA;EAAA,eAQbK,CAAAA,EAzBId,eAyBJc;EAAK,YAAEC,CAAAA,EAxBNX,YAwBMW;;;;;;;;AARmB;;;;;;;;cAAvBF,UAAAA,SAAmBX,IAAAA;;;;;;;;;;;;;;KAQqCU;;qCAEtCX,4BAA4BkB"}

package/dist/tools/webbrowser.js

@@ -1,223 +1,168 @@
+import { __export } from "../_virtual/rolldown_runtime.js";
+import { formatDocumentsAsString } from "../util/document.js";
+import { MemoryVectorStore } from "../vectorstores/memory.js";
 import { Document } from "@langchain/core/documents";
-import axiosMod from "axios";
-import * as cheerio from "cheerio";
-import { isNode } from "@langchain/core/utils/env";
-import { Tool } from "@langchain/core/tools";
 import { RunnableSequence } from "@langchain/core/runnables";
+import { Tool } from "@langchain/core/tools";
+import { RecursiveCharacterTextSplitter } from "@langchain/textsplitters";
 import { StringOutputParser } from "@langchain/core/output_parsers";
-import { RecursiveCharacterTextSplitter, } from "../text_splitter.js";
-import { MemoryVectorStore } from "../vectorstores/memory.js";
-import fetchAdapter from "../util/axios-fetch-adapter.js";
-import { formatDocumentsAsString } from "../util/document.js";
-export const parseInputs = (inputs) => {
-    const [baseUrl, task] = inputs.split(",").map((input) => {
-        let t = input.trim();
-        t = t.startsWith('"') ? t.slice(1) : t;
-        t = t.endsWith('"') ? t.slice(0, -1) : t;
-        // it likes to put / at the end of urls, wont matter for task
-        t = t.endsWith("/") ? t.slice(0, -1) : t;
-        return t.trim();
-    });
-    return [baseUrl, task];
+import * as cheerio from "cheerio";
+
+//#region src/tools/webbrowser.ts
+var webbrowser_exports = {};
+__export(webbrowser_exports, {
+	WebBrowser: () => WebBrowser,
+	getText: () => getText,
+	parseInputs: () => parseInputs
+});
+const parseInputs = (inputs) => {
+	const [baseUrl, task] = inputs.split(",").map((input) => {
+		let t = input.trim();
+		t = t.startsWith("\"") ? t.slice(1) : t;
+		t = t.endsWith("\"") ? t.slice(0, -1) : t;
+		t = t.endsWith("/") ? t.slice(0, -1) : t;
+		return t.trim();
+	});
+	return [baseUrl, task];
 };
-export const getText = (html, baseUrl, summary) => {
-    // scriptingEnabled so noscript elements are parsed
-    const $ = cheerio.load(html, { scriptingEnabled: true });
-    let text = "";
-    // lets only get the body if its a summary, dont need to summarize header or footer etc
-    const rootElement = summary ? "body " : "*";
-    // eslint-disable-next-line @typescript-eslint/no-explicit-any
-    $(`${rootElement}:not(style):not(script):not(svg)`).each((_i, elem) => {
-        // we dont want duplicated content as we drill down so remove children
-        let content = $(elem).clone().children().remove().end().text().trim();
-        const $el = $(elem);
-        // if its an ahref, print the content and url
-        let href = $el.attr("href");
-        if ($el.prop("tagName")?.toLowerCase() === "a" && href) {
-            if (!href.startsWith("http")) {
-                try {
-                    href = new URL(href, baseUrl).toString();
-                }
-                catch {
-                    // if this fails thats fine, just no url for this
-                    href = "";
-                }
-            }
-            const imgAlt = $el.find("img[alt]").attr("alt")?.trim();
-            if (imgAlt) {
-                content += ` ${imgAlt}`;
-            }
-            text += ` [${content}](${href})`;
-        }
-        // otherwise just print the content
-        else if (content !== "") {
-            text += ` ${content}`;
-        }
-    });
-    return text.trim().replace(/\n+/g, " ");
+const getText = (html, baseUrl, summary) => {
+	const $ = cheerio.load(html, { scriptingEnabled: true });
+	let text = "";
+	const rootElement = summary ? "body " : "*";
+	$(`${rootElement}:not(style):not(script):not(svg)`).each((_i, elem) => {
+		let content = $(elem).clone().children().remove().end().text().trim();
+		const $el = $(elem);
+		let href = $el.attr("href");
+		if ($el.prop("tagName")?.toLowerCase() === "a" && href) {
+			if (!href.startsWith("http")) try {
+				href = new URL(href, baseUrl).toString();
+			} catch {
+				href = "";
+			}
+			const imgAlt = $el.find("img[alt]").attr("alt")?.trim();
+			if (imgAlt) content += ` ${imgAlt}`;
+			text += ` [${content}](${href})`;
+		} else if (content !== "") text += ` ${content}`;
+	});
+	return text.trim().replace(/\n+/g, " ");
 };
 const getHtml = async (baseUrl, h, config) => {
-    const axios = ("default" in axiosMod ? axiosMod.default : axiosMod);
-    const domain = new URL(baseUrl).hostname;
-    const headers = { ...h };
-    // these appear to be positional, which means they have to exist in the headers passed in
-    headers.Host = domain;
-    headers["Alt-Used"] = domain;
-    let htmlResponse;
-    try {
-        htmlResponse = await axios.get(baseUrl, {
-            ...config,
-            headers,
-        });
-    }
-    catch (e) {
-        if (axios.isAxiosError(e) && e.response && e.response.status) {
-            throw new Error(`http response ${e.response.status}`);
-        }
-        throw e;
-    }
-    const allowedContentTypes = [
-        "text/html",
-        "application/json",
-        "application/xml",
-        "application/javascript",
-        "text/plain",
-    ];
-    const contentType = htmlResponse.headers["content-type"];
-    const contentTypeArray = contentType.split(";");
-    if (contentTypeArray[0] &&
-        !allowedContentTypes.includes(contentTypeArray[0])) {
-        throw new Error("returned page was not utf8");
-    }
-    return htmlResponse.data;
+	const domain = new URL(baseUrl).hostname;
+	const headers = { ...h };
+	headers.Host = domain;
+	headers["Alt-Used"] = domain;
+	let htmlResponse;
+	try {
+		const fetchOptions = {
+			method: "GET",
+			headers,
+			credentials: config.withCredentials ? "include" : "same-origin",
+			...config
+		};
+		htmlResponse = await fetch(baseUrl, fetchOptions);
+		if (!htmlResponse.ok) throw new Error(`http response ${htmlResponse.status}`);
+	} catch (e) {
+		if (e && typeof e === "object" && "message" in e && typeof e.message === "string" && e.message.includes("http response")) throw e;
+		throw e;
+	}
+	const allowedContentTypes = [
+		"text/html",
+		"application/json",
+		"application/xml",
+		"application/javascript",
+		"text/plain"
+	];
+	const contentType = htmlResponse.headers.get("content-type") || "";
+	const contentTypeArray = contentType.split(";");
+	if (contentTypeArray[0] && !allowedContentTypes.includes(contentTypeArray[0])) throw new Error("returned page was not utf8");
+	return htmlResponse.text();
 };
 const DEFAULT_HEADERS = {
-    Accept: "text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,*/*;q=0.8",
-    "Accept-Encoding": "gzip, deflate",
-    "Accept-Language": "en-US,en;q=0.5",
-    "Alt-Used": "LEAVE-THIS-KEY-SET-BY-TOOL",
-    Connection: "keep-alive",
-    Host: "LEAVE-THIS-KEY-SET-BY-TOOL",
-    Referer: "https://www.google.com/",
-    "Sec-Fetch-Dest": "document",
-    "Sec-Fetch-Mode": "navigate",
-    "Sec-Fetch-Site": "cross-site",
-    "Upgrade-Insecure-Requests": "1",
-    "User-Agent": "Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/111.0",
+	Accept: "text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,*/*;q=0.8",
+	"Accept-Encoding": "gzip, deflate",
+	"Accept-Language": "en-US,en;q=0.5",
+	"Alt-Used": "LEAVE-THIS-KEY-SET-BY-TOOL",
+	Connection: "keep-alive",
+	Host: "LEAVE-THIS-KEY-SET-BY-TOOL",
+	Referer: "https://www.google.com/",
+	"Sec-Fetch-Dest": "document",
+	"Sec-Fetch-Mode": "navigate",
+	"Sec-Fetch-Site": "cross-site",
+	"Upgrade-Insecure-Requests": "1",
+	"User-Agent": "Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/111.0"
 };
 /**
- * A class designed to interact with web pages, either to extract
- * information from them or to summarize their content. It uses the axios
- * library to send HTTP requests and the cheerio library to parse the
- * returned HTML.
- * @example
- * ```typescript
- * const browser = new WebBrowser({
- *   model: new ChatOpenAI({ model: "gpt-4o-mini", temperature: 0 }),
- *   embeddings: new OpenAIEmbeddings({}),
- * });
- * const result = await browser.invoke("https:exampleurl.com");
- * ```
- */
-export class WebBrowser extends Tool {
-    static lc_name() {
-        return "WebBrowser";
-    }
-    get lc_namespace() {
-        return [...super.lc_namespace, "webbrowser"];
-    }
-    constructor({ model, headers, embeddings, axiosConfig, textSplitter, }) {
-        super(...arguments);
-        Object.defineProperty(this, "model", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "embeddings", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "headers", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "axiosConfig", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "textSplitter", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: void 0
-        });
-        Object.defineProperty(this, "name", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: "web-browser"
-        });
-        Object.defineProperty(this, "description", {
-            enumerable: true,
-            configurable: true,
-            writable: true,
-            value: `useful for when you need to find something on or summarize a webpage. input should be a comma separated list of "ONE valid http URL including protocol","what you want to find on the page or empty string for a summary".`
-        });
-        this.model = model;
-        this.embeddings = embeddings;
-        this.headers = headers ?? DEFAULT_HEADERS;
-        this.axiosConfig = {
-            withCredentials: true,
-            adapter: isNode() ? undefined : fetchAdapter,
-            ...axiosConfig,
-        };
-        this.textSplitter =
-            textSplitter ??
-                new RecursiveCharacterTextSplitter({
-                    chunkSize: 2000,
-                    chunkOverlap: 200,
-                });
-    }
-    /** @ignore */
-    async _call(inputs, runManager) {
-        const [baseUrl, task] = parseInputs(inputs);
-        const doSummary = !task;
-        let text;
-        try {
-            const html = await getHtml(baseUrl, this.headers, this.axiosConfig);
-            text = getText(html, baseUrl, doSummary);
-        }
-        catch (e) {
-            if (e) {
-                return e.toString();
-            }
-            return "There was a problem connecting to the site";
-        }
-        const texts = await this.textSplitter.splitText(text);
-        let context;
-        // if we want a summary grab first 4
-        if (doSummary) {
-            context = texts.slice(0, 4).join("\n");
-        }
-        // search term well embed and grab top 4
-        else {
-            const docs = texts.map((pageContent) => new Document({
-                pageContent,
-                metadata: [],
-            }));
-            const vectorStore = await MemoryVectorStore.fromDocuments(docs, this.embeddings);
-            const results = await vectorStore.similaritySearch(task, 4, undefined, runManager?.getChild("vectorstore"));
-            context = formatDocumentsAsString(results);
-        }
-        const input = `Text:${context}\n\nI need ${doSummary ? "a summary" : task} from the above text, also provide up to 5 markdown links from within that would be of interest (always including URL and text). Links should be provided, if present, in markdown syntax as a list under the heading "Relevant Links:".`;
-        const chain = RunnableSequence.from([this.model, new StringOutputParser()]);
-        return chain.invoke(input, runManager?.getChild());
-    }
-}
+* A class designed to interact with web pages, either to extract
+* information from them or to summarize their content. It uses the native
+* fetch API to send HTTP requests and the cheerio library to parse the
+* returned HTML.
+* @example
+* ```typescript
+* const browser = new WebBrowser({
+*   model: new ChatOpenAI({ model: "gpt-4o-mini", temperature: 0 }),
+*   embeddings: new OpenAIEmbeddings({}),
+* });
+* const result = await browser.invoke("https:exampleurl.com");
+* ```
+*/
+var WebBrowser = class extends Tool {
+	static lc_name() {
+		return "WebBrowser";
+	}
+	get lc_namespace() {
+		return [...super.lc_namespace, "webbrowser"];
+	}
+	model;
+	embeddings;
+	headers;
+	requestConfig;
+	textSplitter;
+	constructor({ model, headers, embeddings, requestConfig, textSplitter }) {
+		super(...arguments);
+		this.model = model;
+		this.embeddings = embeddings;
+		this.headers = headers ?? DEFAULT_HEADERS;
+		this.requestConfig = {
+			withCredentials: true,
+			...requestConfig
+		};
+		this.textSplitter = textSplitter ?? new RecursiveCharacterTextSplitter({
+			chunkSize: 2e3,
+			chunkOverlap: 200
+		});
+	}
+	/** @ignore */
+	async _call(inputs, runManager) {
+		const [baseUrl, task] = parseInputs(inputs);
+		const doSummary = !task;
+		let text;
+		try {
+			const html = await getHtml(baseUrl, this.headers, this.requestConfig);
+			text = getText(html, baseUrl, doSummary);
+		} catch (e) {
+			if (e) return e.toString();
+			return "There was a problem connecting to the site";
+		}
+		const texts = await this.textSplitter.splitText(text);
+		let context;
+		if (doSummary) context = texts.slice(0, 4).join("\n");
+		else {
+			const docs = texts.map((pageContent) => new Document({
+				pageContent,
+				metadata: []
+			}));
+			const vectorStore = await MemoryVectorStore.fromDocuments(docs, this.embeddings);
+			const results = await vectorStore.similaritySearch(task, 4, void 0, runManager?.getChild("vectorstore"));
+			context = formatDocumentsAsString(results);
+		}
+		const input = `Text:${context}\n\nI need ${doSummary ? "a summary" : task} from the above text, also provide up to 5 markdown links from within that would be of interest (always including URL and text). Links should be provided, if present, in markdown syntax as a list under the heading "Relevant Links:".`;
+		const chain = RunnableSequence.from([this.model, new StringOutputParser()]);
+		return chain.invoke(input, runManager?.getChild());
+	}
+	name = "web-browser";
+	description = `useful for when you need to find something on or summarize a webpage. input should be a comma separated list of "ONE valid http URL including protocol","what you want to find on the page or empty string for a summary".`;
+};
+
+//#endregion
+export { WebBrowser, getText, parseInputs, webbrowser_exports };
+//# sourceMappingURL=webbrowser.js.map

package/dist/tools/webbrowser.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"webbrowser.js","names":["inputs: string","html: string","baseUrl: string","summary: boolean","elem: any","h: Headers","config: RequestConfig","fetchOptions: RequestInit","runManager?: CallbackManagerForToolRun"],"sources":["../../src/tools/webbrowser.ts"],"sourcesContent":["import type { BaseLanguageModelInterface } from \"@langchain/core/language_models/base\";\nimport { Document } from \"@langchain/core/documents\";\nimport type { EmbeddingsInterface } from \"@langchain/core/embeddings\";\n\nimport * as cheerio from \"cheerio\";\nimport {\n  CallbackManager,\n  CallbackManagerForToolRun,\n} from \"@langchain/core/callbacks/manager\";\n\nimport { Tool, ToolParams } from \"@langchain/core/tools\";\nimport { RunnableSequence } from \"@langchain/core/runnables\";\nimport { StringOutputParser } from \"@langchain/core/output_parsers\";\nimport {\n  RecursiveCharacterTextSplitter,\n  TextSplitter,\n} from \"@langchain/textsplitters\";\nimport { MemoryVectorStore } from \"../vectorstores/memory.js\";\n\nimport { formatDocumentsAsString } from \"../util/document.js\";\n\nexport const parseInputs = (inputs: string): [string, string] => {\n  const [baseUrl, task] = inputs.split(\",\").map((input) => {\n    let t = input.trim();\n    t = t.startsWith('\"') ? t.slice(1) : t;\n    t = t.endsWith('\"') ? t.slice(0, -1) : t;\n    // it likes to put / at the end of urls, wont matter for task\n    t = t.endsWith(\"/\") ? t.slice(0, -1) : t;\n    return t.trim();\n  });\n\n  return [baseUrl, task];\n};\n\nexport const getText = (\n  html: string,\n  baseUrl: string,\n  summary: boolean\n): string => {\n  // scriptingEnabled so noscript elements are parsed\n  const $ = cheerio.load(html, { scriptingEnabled: true });\n\n  let text = \"\";\n\n  // lets only get the body if its a summary, dont need to summarize header or footer etc\n  const rootElement = summary ? \"body \" : \"*\";\n\n  // eslint-disable-next-line @typescript-eslint/no-explicit-any\n  $(`${rootElement}:not(style):not(script):not(svg)`).each((_i, elem: any) => {\n    // we dont want duplicated content as we drill down so remove children\n    let content = $(elem).clone().children().remove().end().text().trim();\n    const $el = $(elem);\n\n    // if its an ahref, print the content and url\n    let href = $el.attr(\"href\");\n    if ($el.prop(\"tagName\")?.toLowerCase() === \"a\" && href) {\n      if (!href.startsWith(\"http\")) {\n        try {\n          href = new URL(href, baseUrl).toString();\n        } catch {\n          // if this fails thats fine, just no url for this\n          href = \"\";\n        }\n      }\n\n      const imgAlt = $el.find(\"img[alt]\").attr(\"alt\")?.trim();\n      if (imgAlt) {\n        content += ` ${imgAlt}`;\n      }\n\n      text += ` [${content}](${href})`;\n    }\n    // otherwise just print the content\n    else if (content !== \"\") {\n      text += ` ${content}`;\n    }\n  });\n\n  return text.trim().replace(/\\n+/g, \" \");\n};\n\nconst getHtml = async (baseUrl: string, h: Headers, config: RequestConfig) => {\n  const domain = new URL(baseUrl).hostname;\n\n  const headers = { ...h };\n  // these appear to be positional, which means they have to exist in the headers passed in\n  headers.Host = domain;\n  headers[\"Alt-Used\"] = domain;\n\n  let htmlResponse;\n  try {\n    const fetchOptions: RequestInit = {\n      method: \"GET\",\n      headers,\n      credentials: config.withCredentials ? \"include\" : \"same-origin\",\n      ...config,\n    };\n\n    htmlResponse = await fetch(baseUrl, fetchOptions);\n\n    if (!htmlResponse.ok) {\n      throw new Error(`http response ${htmlResponse.status}`);\n    }\n  } catch (e) {\n    if (\n      e &&\n      typeof e === \"object\" &&\n      \"message\" in e &&\n      typeof e.message === \"string\" &&\n      e.message.includes(\"http response\")\n    ) {\n      throw e;\n    }\n    throw e;\n  }\n\n  const allowedContentTypes = [\n    \"text/html\",\n    \"application/json\",\n    \"application/xml\",\n    \"application/javascript\",\n    \"text/plain\",\n  ];\n\n  const contentType = htmlResponse.headers.get(\"content-type\") || \"\";\n  const contentTypeArray = contentType.split(\";\");\n  if (\n    contentTypeArray[0] &&\n    !allowedContentTypes.includes(contentTypeArray[0])\n  ) {\n    throw new Error(\"returned page was not utf8\");\n  }\n  return htmlResponse.text();\n};\n\nconst DEFAULT_HEADERS = {\n  Accept:\n    \"text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,*/*;q=0.8\",\n  \"Accept-Encoding\": \"gzip, deflate\",\n  \"Accept-Language\": \"en-US,en;q=0.5\",\n  \"Alt-Used\": \"LEAVE-THIS-KEY-SET-BY-TOOL\",\n  Connection: \"keep-alive\",\n  Host: \"LEAVE-THIS-KEY-SET-BY-TOOL\",\n  Referer: \"https://www.google.com/\",\n  \"Sec-Fetch-Dest\": \"document\",\n  \"Sec-Fetch-Mode\": \"navigate\",\n  \"Sec-Fetch-Site\": \"cross-site\",\n  \"Upgrade-Insecure-Requests\": \"1\",\n  \"User-Agent\":\n    \"Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/111.0\",\n};\n\n// eslint-disable-next-line @typescript-eslint/no-explicit-any\ntype Headers = Record<string, any>;\n\n/**\n * Configuration options for fetch requests, similar to axios config but for fetch\n */\nexport interface RequestConfig extends Omit<RequestInit, \"headers\"> {\n  withCredentials?: boolean;\n}\n\n/**\n * Defines the arguments that can be passed to the WebBrowser constructor.\n * It extends the ToolParams interface and includes properties for a\n * language model, embeddings, HTTP headers, an Axios configuration, a\n * callback manager, and a text splitter.\n */\nexport interface WebBrowserArgs extends ToolParams {\n  model: BaseLanguageModelInterface;\n\n  embeddings: EmbeddingsInterface;\n\n  headers?: Headers;\n\n  requestConfig?: RequestConfig;\n\n  /** @deprecated */\n  callbackManager?: CallbackManager;\n\n  textSplitter?: TextSplitter;\n}\n\n/**\n * A class designed to interact with web pages, either to extract\n * information from them or to summarize their content. It uses the native\n * fetch API to send HTTP requests and the cheerio library to parse the\n * returned HTML.\n * @example\n * ```typescript\n * const browser = new WebBrowser({\n *   model: new ChatOpenAI({ model: \"gpt-4o-mini\", temperature: 0 }),\n *   embeddings: new OpenAIEmbeddings({}),\n * });\n * const result = await browser.invoke(\"https:exampleurl.com\");\n * ```\n */\nexport class WebBrowser extends Tool {\n  static lc_name() {\n    return \"WebBrowser\";\n  }\n\n  get lc_namespace() {\n    return [...super.lc_namespace, \"webbrowser\"];\n  }\n\n  private model: BaseLanguageModelInterface;\n\n  private embeddings: EmbeddingsInterface;\n\n  private headers: Headers;\n\n  private requestConfig: RequestConfig;\n\n  private textSplitter: TextSplitter;\n\n  constructor({\n    model,\n    headers,\n    embeddings,\n    requestConfig,\n    textSplitter,\n  }: WebBrowserArgs) {\n    super(...arguments);\n\n    this.model = model;\n    this.embeddings = embeddings;\n    this.headers = headers ?? DEFAULT_HEADERS;\n    this.requestConfig = {\n      withCredentials: true,\n      ...requestConfig,\n    };\n    this.textSplitter =\n      textSplitter ??\n      new RecursiveCharacterTextSplitter({\n        chunkSize: 2000,\n        chunkOverlap: 200,\n      });\n  }\n\n  /** @ignore */\n  async _call(inputs: string, runManager?: CallbackManagerForToolRun) {\n    const [baseUrl, task] = parseInputs(inputs);\n    const doSummary = !task;\n\n    let text;\n    try {\n      const html = await getHtml(baseUrl, this.headers, this.requestConfig);\n      text = getText(html, baseUrl, doSummary);\n    } catch (e) {\n      if (e) {\n        return e.toString();\n      }\n      return \"There was a problem connecting to the site\";\n    }\n\n    const texts = await this.textSplitter.splitText(text);\n\n    let context;\n    // if we want a summary grab first 4\n    if (doSummary) {\n      context = texts.slice(0, 4).join(\"\\n\");\n    }\n    // search term well embed and grab top 4\n    else {\n      const docs = texts.map(\n        (pageContent) =>\n          new Document({\n            pageContent,\n            metadata: [],\n          })\n      );\n\n      const vectorStore = await MemoryVectorStore.fromDocuments(\n        docs,\n        this.embeddings\n      );\n      const results = await vectorStore.similaritySearch(\n        task,\n        4,\n        undefined,\n        runManager?.getChild(\"vectorstore\")\n      );\n      context = formatDocumentsAsString(results);\n    }\n\n    const input = `Text:${context}\\n\\nI need ${\n      doSummary ? \"a summary\" : task\n    } from the above text, also provide up to 5 markdown links from within that would be of interest (always including URL and text). Links should be provided, if present, in markdown syntax as a list under the heading \"Relevant Links:\".`;\n\n    const chain = RunnableSequence.from([this.model, new StringOutputParser()]);\n    return chain.invoke(input, runManager?.getChild());\n  }\n\n  name = \"web-browser\";\n\n  description = `useful for when you need to find something on or summarize a webpage. input should be a comma separated list of \"ONE valid http URL including protocol\",\"what you want to find on the page or empty string for a summary\".`;\n}\n"],"mappings":";;;;;;;;;;;;;;;;;AAqBA,MAAa,cAAc,CAACA,WAAqC;CAC/D,MAAM,CAAC,SAAS,KAAK,GAAG,OAAO,MAAM,IAAI,CAAC,IAAI,CAAC,UAAU;EACvD,IAAI,IAAI,MAAM,MAAM;EACpB,IAAI,EAAE,WAAW,KAAI,GAAG,EAAE,MAAM,EAAE,GAAG;EACrC,IAAI,EAAE,SAAS,KAAI,GAAG,EAAE,MAAM,GAAG,GAAG,GAAG;EAEvC,IAAI,EAAE,SAAS,IAAI,GAAG,EAAE,MAAM,GAAG,GAAG,GAAG;AACvC,SAAO,EAAE,MAAM;CAChB,EAAC;AAEF,QAAO,CAAC,SAAS,IAAK;AACvB;AAED,MAAa,UAAU,CACrBC,MACAC,SACAC,YACW;CAEX,MAAM,IAAI,QAAQ,KAAK,MAAM,EAAE,kBAAkB,KAAM,EAAC;CAExD,IAAI,OAAO;CAGX,MAAM,cAAc,UAAU,UAAU;CAGxC,EAAE,GAAG,YAAY,gCAAgC,CAAC,CAAC,CAAC,KAAK,CAAC,IAAIC,SAAc;EAE1E,IAAI,UAAU,EAAE,KAAK,CAAC,OAAO,CAAC,UAAU,CAAC,QAAQ,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM;EACrE,MAAM,MAAM,EAAE,KAAK;EAGnB,IAAI,OAAO,IAAI,KAAK,OAAO;AAC3B,MAAI,IAAI,KAAK,UAAU,EAAE,aAAa,KAAK,OAAO,MAAM;AACtD,OAAI,CAAC,KAAK,WAAW,OAAO,CAC1B,KAAI;IACF,OAAO,IAAI,IAAI,MAAM,SAAS,UAAU;GACzC,QAAO;IAEN,OAAO;GACR;GAGH,MAAM,SAAS,IAAI,KAAK,WAAW,CAAC,KAAK,MAAM,EAAE,MAAM;AACvD,OAAI,QACF,WAAW,CAAC,CAAC,EAAE,QAAQ;GAGzB,QAAQ,CAAC,EAAE,EAAE,QAAQ,EAAE,EAAE,KAAK,CAAC,CAAC;EACjC,WAEQ,YAAY,IACnB,QAAQ,CAAC,CAAC,EAAE,SAAS;CAExB,EAAC;AAEF,QAAO,KAAK,MAAM,CAAC,QAAQ,QAAQ,IAAI;AACxC;AAED,MAAM,UAAU,OAAOF,SAAiBG,GAAYC,WAA0B;CAC5E,MAAM,SAAS,IAAI,IAAI,SAAS;CAEhC,MAAM,UAAU,EAAE,GAAG,EAAG;CAExB,QAAQ,OAAO;CACf,QAAQ,cAAc;CAEtB,IAAI;AACJ,KAAI;EACF,MAAMC,eAA4B;GAChC,QAAQ;GACR;GACA,aAAa,OAAO,kBAAkB,YAAY;GAClD,GAAG;EACJ;EAED,eAAe,MAAM,MAAM,SAAS,aAAa;AAEjD,MAAI,CAAC,aAAa,GAChB,OAAM,IAAI,MAAM,CAAC,cAAc,EAAE,aAAa,QAAQ;CAEzD,SAAQ,GAAG;AACV,MACE,KACA,OAAO,MAAM,YACb,aAAa,KACb,OAAO,EAAE,YAAY,YACrB,EAAE,QAAQ,SAAS,gBAAgB,CAEnC,OAAM;AAER,QAAM;CACP;CAED,MAAM,sBAAsB;EAC1B;EACA;EACA;EACA;EACA;CACD;CAED,MAAM,cAAc,aAAa,QAAQ,IAAI,eAAe,IAAI;CAChE,MAAM,mBAAmB,YAAY,MAAM,IAAI;AAC/C,KACE,iBAAiB,MACjB,CAAC,oBAAoB,SAAS,iBAAiB,GAAG,CAElD,OAAM,IAAI,MAAM;AAElB,QAAO,aAAa,MAAM;AAC3B;AAED,MAAM,kBAAkB;CACtB,QACE;CACF,mBAAmB;CACnB,mBAAmB;CACnB,YAAY;CACZ,YAAY;CACZ,MAAM;CACN,SAAS;CACT,kBAAkB;CAClB,kBAAkB;CAClB,kBAAkB;CAClB,6BAA6B;CAC7B,cACE;AACH;;;;;;;;;;;;;;;AA+CD,IAAa,aAAb,cAAgC,KAAK;CACnC,OAAO,UAAU;AACf,SAAO;CACR;CAED,IAAI,eAAe;AACjB,SAAO,CAAC,GAAG,MAAM,cAAc,YAAa;CAC7C;CAED,AAAQ;CAER,AAAQ;CAER,AAAQ;CAER,AAAQ;CAER,AAAQ;CAER,YAAY,EACV,OACA,SACA,YACA,eACA,cACe,EAAE;EACjB,MAAM,GAAG,UAAU;EAEnB,KAAK,QAAQ;EACb,KAAK,aAAa;EAClB,KAAK,UAAU,WAAW;EAC1B,KAAK,gBAAgB;GACnB,iBAAiB;GACjB,GAAG;EACJ;EACD,KAAK,eACH,gBACA,IAAI,+BAA+B;GACjC,WAAW;GACX,cAAc;EACf;CACJ;;CAGD,MAAM,MAAMP,QAAgBQ,YAAwC;EAClE,MAAM,CAAC,SAAS,KAAK,GAAG,YAAY,OAAO;EAC3C,MAAM,YAAY,CAAC;EAEnB,IAAI;AACJ,MAAI;GACF,MAAM,OAAO,MAAM,QAAQ,SAAS,KAAK,SAAS,KAAK,cAAc;GACrE,OAAO,QAAQ,MAAM,SAAS,UAAU;EACzC,SAAQ,GAAG;AACV,OAAI,EACF,QAAO,EAAE,UAAU;AAErB,UAAO;EACR;EAED,MAAM,QAAQ,MAAM,KAAK,aAAa,UAAU,KAAK;EAErD,IAAI;AAEJ,MAAI,WACF,UAAU,MAAM,MAAM,GAAG,EAAE,CAAC,KAAK,KAAK;OAGnC;GACH,MAAM,OAAO,MAAM,IACjB,CAAC,gBACC,IAAI,SAAS;IACX;IACA,UAAU,CAAE;GACb,GACJ;GAED,MAAM,cAAc,MAAM,kBAAkB,cAC1C,MACA,KAAK,WACN;GACD,MAAM,UAAU,MAAM,YAAY,iBAChC,MACA,GACA,QACA,YAAY,SAAS,cAAc,CACpC;GACD,UAAU,wBAAwB,QAAQ;EAC3C;EAED,MAAM,QAAQ,CAAC,KAAK,EAAE,QAAQ,WAAW,EACvC,YAAY,cAAc,KAC3B,wOAAwO,CAAC;EAE1O,MAAM,QAAQ,iBAAiB,KAAK,CAAC,KAAK,OAAO,IAAI,oBAAqB,EAAC;AAC3E,SAAO,MAAM,OAAO,OAAO,YAAY,UAAU,CAAC;CACnD;CAED,OAAO;CAEP,cAAc,CAAC,0NAA0N,CAAC;AAC3O"}