Skip to main content
Using Deno in production at your company? Earn free Deno merch.
Give us feedback


The Telegram Bot Framework.
Very Popular
Go to Latest
// deno-lint-ignore-file camelcaseimport { BotError, Composer, type Middleware, type ReactionMiddleware, run,} from "./composer.ts";import { Context, type MaybeArray, type ReactionContext } from "./context.ts";import { Api } from "./core/api.ts";import { type ApiClientOptions, type WebhookReplyEnvelope,} from "./core/client.ts";import { GrammyError, HttpError } from "./core/error.ts";import { type Filter, type FilterQuery, parse, preprocess } from "./filter.ts";import { debug as d } from "./platform.deno.ts";import { type ReactionType, type ReactionTypeEmoji, type Update, type UserFromGetMe,} from "./types.ts";const debug = d("grammy:bot");const debugWarn = d("grammy:warn");const debugErr = d("grammy:error");
export const DEFAULT_UPDATE_TYPES = [ "message", "edited_message", "channel_post", "edited_channel_post", "inline_query", "chosen_inline_result", "callback_query", "shipping_query", "pre_checkout_query", "poll", "poll_answer", "my_chat_member", "chat_join_request", "chat_boost", "removed_chat_boost",] as const;
/** * Options that can be specified when running the bot via simple long polling. */export interface PollingOptions { /** * Limits the number of updates to be retrieved per `getUpdates` call. * Values between 1-100 are accepted. Defaults to 100. */ limit?: number; /** * Timeout in seconds for long polling. grammY uses 30 seconds as a default * value. */ timeout?: number; /** * A list of the update types you want your bot to receive. For example, * specify [“message”, “edited_channel_post”, “callback_query”] to only * receive updates of these types. See Update for a complete list of * available update types. Specify an empty list to receive all update types * except chat_member (default). If not specified, the previous setting will * be used. * * Please note that this parameter doesn't affect updates created before the * call to the getUpdates, so unwanted updates may be received for a short * period of time. */ allowed_updates?: ReadonlyArray<Exclude<keyof Update, "update_id">>; /** * Pass True to drop all pending updates before starting the long polling. */ drop_pending_updates?: boolean; /** * A callback function that is useful for logging (or setting up middleware * if you did not do this before). It will be executed after the setup of * the bot has completed, and immediately before the first updates are being * fetched. The bot information `bot.botInfo` will be available when the * function is run. For convenience, the callback function receives the * value of `bot.botInfo` as an argument. */ onStart?: (botInfo: UserFromGetMe) => void | Promise<void>;}
export { BotError };/** * Error handler that can be installed on a bot to catch error thrown by * middleware. */export type ErrorHandler<C extends Context = Context> = ( error: BotError<C>,) => unknown;
/** * Options to pass to the bot when creating it. */export interface BotConfig<C extends Context> { /** * You can specify a number of advanced options under the `client` property. * The options will be passed to the grammY client—this is the part of * grammY that actually connects to the Telegram Bot API server in the end * when making HTTP requests. */ client?: ApiClientOptions; /** * grammY automatically calls `getMe` when starting up to make sure that * your bot has access to the bot's own information. If you restart your bot * often, for example because it is running in a serverless environment, * then you may want to skip this initial API call. * * Set this property of the options to pre-initialize the bot with cached * values. If you use this option, grammY will not attempt to make a `getMe` * call but use the provided data instead. */ botInfo?: UserFromGetMe; /** * Pass the constructor of a custom context object that will be used when * creating the context for each incoming update. */ ContextConstructor?: new ( ...args: ConstructorParameters<typeof Context> ) => C;}
/** * This is the single most important class of grammY. It represents your bot. * * First, you must create a bot by talking to @BotFather, check out * Once it is ready, you obtain a secret token for your * bot. grammY will use that token to identify as your bot when talking to the * Telegram servers. Got the token? You are now ready to write some code and run * your bot! * * You should do three things to run your bot: * ```ts * // 1. Create a bot instance * const bot = new Bot('<secret-token>') * // 2. Listen for updates * bot.on('message:text', ctx => ctx.reply('You wrote: ' + ctx.message.text)) * // 3. Launch it! * bot.start() * ``` */export class Bot< C extends Context = Context, A extends Api = Api,> extends Composer<C> { private pollingRunning = false; private pollingAbortController: AbortController | undefined; private lastTriedUpdateId = 0;
/** * Gives you full access to the Telegram Bot API. * ```ts * // This is how to call the Bot API methods: * bot.api.sendMessage(chat_id, 'Hello, grammY!') * ``` * * Use this only outside of your middleware. If you have access to `ctx`, * then using `ctx.api` instead of `bot.api` is preferred. */ public readonly api: A;
private me: UserFromGetMe | undefined; private mePromise: Promise<UserFromGetMe> | undefined; private readonly clientConfig: ApiClientOptions | undefined;
private readonly ContextConstructor: new ( ...args: ConstructorParameters<typeof Context> ) => C;
/** Used to log a warning if some update types are not in allowed_updates */ private observedUpdateTypes = new Set<string>();
/** * Holds the bot's error handler that is invoked whenever middleware throws * (rejects). If you set your own error handler via `bot.catch`, all that * happens is that this variable is assigned. */ public errorHandler: ErrorHandler<C> = async (err) => { console.error( "Error in middleware while handling update", err.ctx?.update?.update_id, err.error, ); console.error("No error handler was set!"); console.error("Set your own error handler with `bot.catch = ...`"); if (this.pollingRunning) { console.error("Stopping bot"); await this.stop(); } throw err; };
/** * Creates a new Bot with the given token. * * Remember that you can listen for messages by calling * ```ts * bot.on('message', ctx => { ... }) * ``` * or similar methods. * * The simplest way to start your bot is via simple long polling: * ```ts * bot.start() * ``` * * @param token The bot's token as acquired from * @param config Optional configuration properties for the bot */ constructor(public readonly token: string, config?: BotConfig<C>) { super(); if (!token) throw new Error("Empty token!"); = config?.botInfo; this.clientConfig = config?.client; this.ContextConstructor = config?.ContextConstructor ?? (Context as unknown as new ( ...args: ConstructorParameters<typeof Context> ) => C); this.api = new Api(token, this.clientConfig) as A; }
/** * Information about the bot itself as retrieved from `api.getMe()`. Only * available after the bot has been initialized via `await bot.init()`, or * after the value has been set manually. * * Starting the bot will always perform the initialization automatically, * unless a manual value is already set. * * Note that the recommended way to set a custom bot information object is * to pass it to the configuration object of the `new Bot()` instantiation, * rather than assigning this property. */ public set botInfo(botInfo: UserFromGetMe) { = botInfo; } public get botInfo(): UserFromGetMe { if ( === undefined) { throw new Error( "Bot information unavailable! Make sure to call `await bot.init()` before accessing `bot.botInfo`!", ); } return; }
/** * @inheritdoc */ on<Q extends FilterQuery>( filter: Q | Q[], ...middleware: Array<Middleware<Filter<C, Q>>> ): Composer<Filter<C, Q>> { for (const [u] of parse(filter).flatMap(preprocess)) { this.observedUpdateTypes.add(u); } return super.on(filter, ...middleware); } /** * @inheritdoc */ reaction( reaction: MaybeArray<ReactionTypeEmoji["emoji"] | ReactionType>, ...middleware: Array<ReactionMiddleware<C>> ): Composer<ReactionContext<C>> { this.observedUpdateTypes.add("message_reaction"); return super.reaction(reaction, ...middleware); }
/** * Checks if the bot has been initialized. A bot is initialized if the bot * information is set. The bot information can either be set automatically * by calling `bot.init`, or manually through the bot constructor. Note that * usually, initialization is done automatically and you do not have to care * about this method. * * @returns true if the bot is initialized, and false otherwise */ isInited() { return !== undefined; }
/** * Initializes the bot, i.e. fetches information about the bot itself. This * method is called automatically, you usually don't have to call it * manually. * * @param signal Optional `AbortSignal` to cancel the initialization */ async init(signal?: AbortSignal) { if (!this.isInited()) { debug("Initializing bot"); this.mePromise ??= withRetries( () => this.api.getMe(signal), signal, ); let me: UserFromGetMe; try { me = await this.mePromise; } finally { this.mePromise = undefined; } if ( === undefined) = me; else debug("Bot info was set by now, will not overwrite"); } debug(`I am ${!.username}!`); }
/** * Internal. Do not call. Handles an update batch sequentially by supplying * it one-by-one to the middleware. Handles middleware errors and stores the * last update identifier that was being tried to handle. * * @param updates An array of updates to handle */ private async handleUpdates(updates: Update[]) { // handle updates sequentially (!) for (const update of updates) { this.lastTriedUpdateId = update.update_id; try { await this.handleUpdate(update); } catch (err) { // should always be true if (err instanceof BotError) { await this.errorHandler(err); } else { console.error("FATAL: grammY unable to handle:", err); throw err; } } } }
/** * This is an internal method that you probably will not ever need to call. * It is used whenever a new update arrives from the Telegram servers that * your bot will handle. * * If you're writing a library on top of grammY, check out the * [documentation]( of the runner * plugin for an example that uses this method. * * @param update An update from the Telegram Bot API * @param webhookReplyEnvelope An optional webhook reply envelope */ async handleUpdate( update: Update, webhookReplyEnvelope?: WebhookReplyEnvelope, ) { if ( === undefined) { throw new Error( "Bot not initialized! Either call `await bot.init()`, \or directly set the `botInfo` option in the `Bot` constructor to specify \a known bot info object.", ); } debug(`Processing update ${update.update_id}`); // create API object const api = new Api( this.token, this.clientConfig, webhookReplyEnvelope, ); // configure it with the same transformers as bot.api const t = this.api.config.installedTransformers(); if (t.length > 0) api.config.use(...t); // create context object const ctx = new this.ContextConstructor(update, api,; try { // run middleware stack await run(this.middleware(), ctx); } catch (err) { debugErr(`Error in middleware for update ${update.update_id}`); throw new BotError<C>(err, ctx); } }
/** * Starts your bot using long polling. * * > This method returns a `Promise` that will never resolve except if your * > bot is stopped. **You don't need to `await` the call to `bot.start`**, * > but remember to catch potential errors by calling `bot.catch`. * > Otherwise your bot will crash (and stop) if something goes wrong in * > your code. * * This method effectively enters a loop that will repeatedly call * `getUpdates` and run your middleware for every received update, allowing * your bot to respond to messages. * * If your bot is already running, this method does nothing. * * **Note that this starts your bot using a very simple long polling * implementation.** `bot.start` should only be used for small bots. While * the rest of grammY was built to perform well even under extreme loads, * simple long polling is not capable of scaling up in a similar fashion. * You should switch over to using `@grammyjs/runner` if you are running a * bot with high load. * * What exactly _high load_ means differs from bot to bot, but as a rule of * thumb, simple long polling should not be processing more than ~5K * messages every hour. Also, if your bot has long-running operations such * as large file transfers that block the middleware from completing, this * will impact the responsiveness negatively, so it makes sense to use the * `@grammyjs/runner` package even if you receive much fewer messages. If * you worry about how much load your bot can handle, check out the grammY * [documentation]( about scaling * up. * * @param options Options to use for simple long polling */ async start(options?: PollingOptions) { // Perform setup const setup: Promise<void>[] = []; if (!this.isInited()) { setup.push(this.init(this.pollingAbortController?.signal)); } if (this.pollingRunning) { await Promise.all(setup); debug("Simple long polling already running!"); return; } else { this.pollingRunning = true; this.pollingAbortController = new AbortController(); } setup.push(withRetries(async () => { await this.api.deleteWebhook({ drop_pending_updates: options?.drop_pending_updates, }, this.pollingAbortController?.signal); }, this.pollingAbortController?.signal)); await Promise.all(setup);
// All async ops of setup complete, run callback await options?.onStart?.(this.botInfo);
// Bot was stopped during `onStart` if (!this.pollingRunning) return;
// Prevent common misuse that leads to missing updates validateAllowedUpdates( this.observedUpdateTypes, options?.allowed_updates, ); // Prevent common misuse that causes memory leak this.use = noUseFunction;
// Start polling debug("Starting simple long polling"); await this.loop(options); debug("Middleware is done running"); }
/** * Stops the bot from long polling. * * All middleware that is currently being executed may complete, but no * further `getUpdates` calls will be performed. The current `getUpdates` * request will be cancelled. * * In addition, this method will _confirm_ the last received update to the * Telegram servers by calling `getUpdates` one last time with the latest * offset value. If any updates are received in this call, they are * discarded and will be fetched again when the bot starts up the next time. * Confer the official documentation on confirming updates if you want to * know more: * * > Note that this method will not wait for the middleware stack to finish. * > If you need to run code after all middleware is done, consider waiting * > for the promise returned by `bot.start()` to resolve. */ async stop() { if (this.pollingRunning) { debug("Stopping bot, saving update offset"); this.pollingRunning = false; this.pollingAbortController?.abort(); const offset = this.lastTriedUpdateId + 1; await this.api.getUpdates({ offset, limit: 1 }) .finally(() => this.pollingAbortController = undefined); } else { debug("Bot is not running!"); } }
/** * Sets the bots error handler that is used during long polling. * * You should call this method to set an error handler if you are using long * polling, no matter whether you use `bot.start` or the `@grammyjs/runner` * package to run your bot. * * Calling `bot.catch` when using other means of running your bot (or * webhooks) has no effect. * * @param errorHandler A function that handles potential middleware errors */ catch(errorHandler: ErrorHandler<C>) { this.errorHandler = errorHandler; }
/** * Internal. Do not call. Enters a loop that will perform long polling until * the bot is stopped. */ private async loop(options?: PollingOptions) { const limit = options?.limit; const timeout = options?.timeout ?? 30; // seconds let allowed_updates: PollingOptions["allowed_updates"] = options?.allowed_updates ?? []; // reset to default if unspecified
while (this.pollingRunning) { // fetch updates const updates = await this.fetchUpdates( { limit, timeout, allowed_updates }, ); // check if polling stopped if (updates === undefined) break; // handle updates await this.handleUpdates(updates); // Telegram uses the last setting if `allowed_updates` is omitted so // we can save some traffic by only sending it in the first request allowed_updates = undefined; } }
/** * Internal. Do not call. Reliably fetches an update batch via `getUpdates`. * Handles all known errors. Returns `undefined` if the bot is stopped and * the call gets cancelled. * * @param options Polling options * @returns An array of updates, or `undefined` if the bot is stopped. */ private async fetchUpdates( { limit, timeout, allowed_updates }: PollingOptions, ) { const offset = this.lastTriedUpdateId + 1; let updates: Update[] | undefined = undefined; do { try { updates = await this.api.getUpdates( { offset, limit, timeout, allowed_updates }, this.pollingAbortController?.signal, ); } catch (error) { await this.handlePollingError(error); } } while (updates === undefined && this.pollingRunning); return updates; }
/** * Internal. Do not call. Handles an error that occurred during long * polling. */ private async handlePollingError(error: unknown) { if (!this.pollingRunning) { debug("Pending getUpdates request cancelled"); return; } let sleepSeconds = 3; if (error instanceof GrammyError) { debugErr(error.message); if (error.error_code === 401) { debugErr( "Make sure you are using the bot token you obtained from @BotFather (", ); throw error; } else if (error.error_code === 409) { debugErr( "Consider revoking the bot token if you believe that no other instance is running.", ); throw error; } else if (error.error_code === 429) { debugErr("Bot API server is closing."); sleepSeconds = error.parameters.retry_after ?? sleepSeconds; } } else debugErr(error); debugErr( `Call to getUpdates failed, retrying in ${sleepSeconds} seconds ...`, ); await sleep(sleepSeconds); }}
/** * Performs a network call task, retrying upon known errors until success. * * If the task errors and a retry_after value can be used, a subsequent retry * will be delayed by the specified period of time. * * Otherwise, if the first attempt at running the task fails, the task is * retried immediately. If second attempt fails, too, waits for 100 ms, and then * doubles this delay for every subsequent attemt. Never waits longer than 1 * hour before retrying. * * @param task Async task to perform * @param signal Optional `AbortSignal` to prevent further retries */async function withRetries<T>( task: () => Promise<T>, signal?: AbortSignal,): Promise<T> { // Set up delays between retries const INITIAL_DELAY = 50; // ms let lastDelay = INITIAL_DELAY;
// Define error handler /** * Determines the error handling strategy based on various error types. * Sleeps if necessary, and returns whether to retry or rethrow an error. */ async function handleError(error: unknown) { let delay = false; let strategy: "retry" | "rethrow" = "rethrow";
if (error instanceof HttpError) { delay = true; strategy = "retry"; } else if (error instanceof GrammyError) { if (error.error_code >= 500) { delay = true; strategy = "retry"; } else if (error.error_code === 429) { const retryAfter = error.parameters.retry_after; if (typeof retryAfter === "number") { // ignore the backoff for sleep, then reset it await sleep(retryAfter, signal); lastDelay = INITIAL_DELAY; } else { delay = true; } strategy = "retry"; } }
if (delay) { // Do not sleep for the first retry if (lastDelay !== INITIAL_DELAY) { await sleep(lastDelay, signal); } const TWENTY_MINUTES = 20 * 60 * 1000; // ms lastDelay = Math.min(TWENTY_MINUTES, 2 * lastDelay); }
return strategy; }
// Perform the actual task with retries let result: { ok: false } | { ok: true; value: T } = { ok: false }; while (!result.ok) { try { result = { ok: true, value: await task() }; } catch (error) { debugErr(error); const strategy = await handleError(error); switch (strategy) { case "retry": continue; case "rethrow": throw error; } } } return result.value;}
/** * Returns a new promise that resolves after the specified number of seconds, or * rejects as soon as the given signal is aborted. */async function sleep(seconds: number, signal?: AbortSignal) { let handle: number | undefined; let reject: ((err: Error) => void) | undefined; function abort() { reject?.(new Error("Aborted delay")); if (handle !== undefined) clearTimeout(handle); } try { await new Promise<void>((res, rej) => { reject = rej; if (signal?.aborted) { abort(); return; } signal?.addEventListener("abort", abort); handle = setTimeout(res, 1000 * seconds); }); } finally { signal?.removeEventListener("abort", abort); }}
/** * Takes a set of observed update types and a list of allowed updates and logs a * warning in debug mode if some update types were observed that have not been * allowed. */function validateAllowedUpdates( updates: Set<string>, allowed: readonly string[] = DEFAULT_UPDATE_TYPES,) { const impossible = Array.from(updates).filter((u) => !allowed.includes(u)); if (impossible.length > 0) { debugWarn( `You registered listeners for the following update types, \but you did not specify them in \`allowed_updates\` \so they may not be received: ${ => `'${u}'`).join(", ")}`, ); }}function noUseFunction(): never { throw new Error(`It looks like you are registering more listeners \on your bot from within other listeners! This means that every time your bot \handles a message like this one, new listeners will be added. This list grows until \your machine crashes, so grammY throws this error to tell you that you should \probably do things a bit differently. If you're unsure how to resolve this problem, \you can ask in the group chat:
On the other hand, if you actually know what you're doing and you do need to install \further middleware while your bot is running, consider installing a composer \instance on your bot, and in turn augment the composer after the fact. This way, \you can circumvent this protection against memory leaks.`);}