TalkJS Core

The @talkjs/core package lets you connect to your TalkJS chat as a user and read, subscribe to, and update your chat data.

See our docs for more information.

Prerequisites

To use this package, you will need a TalkJS account. TalkJS provides a ready-to-use chat client for your application. Your account gives you access to TalkJS's free development environment.

Example

This example demonstrates how to create a TalkJS session and then create two users and conversation between them.

Install the @talkjs/core package:

``sh
npm install @talkjs/core


or


yarn add @talkjs/core
`

Import it into the component where you want to use it:

`js
import { getTalkSession } from "@talkjs/core";
`

Then add the following code to create the session, users and conversation:

`js
// Replace with your own app ID
const appId = "";
const userId = "alice";

const session = getTalkSession({ appId, userId });
session.currentUser.createIfNotExists({ name: "Alice" });

const conversation = session.conversation("my_conversation");
conversation.createIfNotExists();

conversation.subscribeMessages((messages) => {
console.log(messages);
});
`

For example, if you are using React, add this code inside a useEffect hook.

Node.js compatibility

This package is primarily intended to be used in the browser because it only connects as a single user. However, it also works in Node.js.

The @talkjs/core package uses WebSockets to create a connection to the TalkJS servers. If you use @talkjs/core in the browser or in Node.js version 22, it will use the built-in WebSocket implementation automatically. Node.js version 21 will also use the built-in implementation if the --experimental-websocket flag is enabled.

If you are using an older version of Node.js without built-in support for WebSockets, you will need to add support with a library. We recommend installing the ws package. Then tell @talkjs/core to use WebSocket from the library:

`js
import { getTalkSession, registerPolyfills } from "@talkjs/core";
import { WebSocket } from "ws";

registerPolyfills({ WebSocket: WebSocket });

const session = getTalkSession(...);
`

Support

If you encounter any problems with @talkjs/core, please open a chat with support. TalkJS support is staffed by engineers.

Changelog

$3

- Calling loadMore on one subscription no longer affects other subscriptions
- While a loadMore request is in progress, you can now call loadMore again with a higher count to fetch a bigger page

$3

- Added ParticipantRef.subscribe which lets you subscribe to a specific participant in a conversation
- Optimised TalkSession.uploadFile (and other upload methods) to check that a file will be accepted before uploading it. Previously it fully uploaded the file, then received the error.

$3

- Added support for global message and conversation searching:
- Added SearchMessageSnapshot, MessageSearch, MessageSearchState, ConversationSearch, and ConversationSearchState
- Added TalkSession.searchMessages(onResults: (results: SearchMessageSnapshot[], loadedAll: boolean) => void): MessageSearch
- Added TalkSession.searchConversations(onResults: (results: ConversationSnapshot[], loadedAll: boolean) => void): ConversationSearch

$3

- "Ref" objects now maintain referential integrity. If you call Session.conversation("123") twice, in two different parts of your app, the same object will be reused both times.

$3

- Made locale in SetUserParams nullable. Setting locale to null means the user will fall-back to the default locale set on the dashboard.
- Improved performance when repeatedly calling ConversationRef.get().

$3

- Fixed a bug where two authentication methods could be active at the same time, causing race conditions.
- Prevented switching to JWT authentication once connected to TalkJS servers. Your initial setToken or onNeedToken configuration must happen before the first requests / before mounting any UI components.
- Optimised initial WebSocket connection to happen in parallel with fetching token, to improve startup speed.

To use JWTs, call setToken or set onNeedToken before triggering any requests or mounting any components. We recommend setting up authentication in your app's entry point (App.jsx, main.js, or similar) so it runs first.

$3

- Added TalkSession.setToken(token: string) which reauthenticates the session with a new token.
- Added TalkSession.onNeedToken: () => void which gets called whenever a new auth token is needed. To use refreshable auth tokens, set a callback that fetches a new auth token and passes it to setToken.
- Deprecated token and tokenFetcher options in getTalkSession in favour of those two properties.
- Removed some duplicate logging on auth-related errors.
- Sessions no longer get garbage collected under any circumstances (to enable persistence of .setToken).
- Initial token fetch is now delayed until you actually use the session to send a request, rather than happening as soon as you call getTalkSession.

$3

- Fixed a bug that would make the library crash under certain conditions.

$3

- Bugfix for React Native.

$3

- Fixed a bug that would cause a crash in React Native.

$3

- Bugfix for on-prem deployments, no impact on the SAAS offering.

$3

- Marked snapshot properties as readonly and applied Object.freeze to snapshots. These were never intended to be editable, as the snapshots are reused and shared between sessions.
- Improved docs for ConversationRef to better explain the reason why ConversationRef.set adds you to the conversation.

$3

- Fixed a bug that prevented the websocket from disconnecting when idle after unsubscribing.

$3

- Fixed a bug where subscribing to a conversation's messages after previously subscribing to that conversation, could cause an error related to sequence numbers being out-of-order.

$3

- Added browser field to package.json.

$3

- When passing markup text to ConversationRef.send, it now ignores formatted links and action buttons/links. Previously it parsed those nodes and tried to send them, which caused an error because users do not have permission to send those nodes.

$3

- Fixed a bug where MessageRef.get() would always set MessageSnapshot.referencedMessage to null
- Added support for getting and setting emoji reactions on messages.
- Added ReactionRef which is created via MessageRef.reaction.
- Added ReactionSnapshot which is available in MessageSnapshot.reactions.
- Session.user, Session.conversation, ConversationRef.message, ConversationRef.participant, and MessageRef.reaction now all perform type validation on the argument and will throw an error if you use an invalid type or an empty string.

$3

- Changed GenericFileBlock from subtype?: undefined to subtype?: never so that type narrowing works consistently when you check subtype === "".

$3

- Fixed an error when subscribing to a conversation while already subscribed to all conversations, or vice-versa.

$3

- Added UserRef.subscribeOnline which subscribes to a user's online status.
- Added UserOnlineSnapshot, UserOnlineSubscription, and UserOnlineActiveState.

$3

- Added TalkSession.subscribeConversations which returns a windowed subscription to your most recently active conversations.
- Added ConversationListSubscription and ConversationListActiveState.

$3

- Added ConversationRef.subscribeParticipants which returns a windowed subscription to the participants in the conversation.
- Added ParticipantSubscription and ParticipantActiveState.
- Improved startup time by sending queued requests faster when you first open the connection.

$3

- Added ConversationSnapshot.everyoneReadUntil which is the minimum of all the participants' readUntil values.

$3

- Added ConversationRef.subscribeTyping which lets you subscribe to typing indicators in a conversation.
- Added TypingSubscription, TypingSnapshot, ManyTypingSnapshot, FewTypingSnapshot

$3

- Fixed a bug with ConversationRef.subscribe where it would emit the same snapshot multiple times in a row, when old messages were edited or deleted.
- Exported types for GenericFileMetadata, AudioFileMetadata, ImageFileMetadata, VideoFileMetadata, and VoiceRecordingFileMetadata`.

$3

- Initial Release