Feat/mulitple data channel(llm) connections by mickelr · Pull Request #4661 · webex/webex-js-sdk

mickelr · 2026-01-20T07:01:28Z

COMPLETES #< INSERT LINK TO ISSUE >

https://jira-eng-gpk2.cisco.com/jira/browse/SPARK-697743

This pull request addresses

To support the 5k webinar, client needs to keep the main LLM connection after entring practice session with practice session LLM connection. So client needs to maintain more than one LLM connections in a single meeting. Currently the llm and mercury lib only support one connection at same time. We need a way to implement multiple data channel (LLM) connections at same time.

by making the following changes

Consider the backward compatibility, I use a map to manage mulitple socket connections with different session ids. If no session id passed when connect/disconnect, we will use default session id by which the lib can be compatibility with previous caller.

Change Type

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to change)
Documentation update
Tooling change
Internal code refactor

The following scenarios were tested

Join meeting, check CC, reaction, whiteboard feature can work normally

The GAI Coding Policy And Copyright Annotation Best Practices

I certified that

I have read and followed contributing guidelines
I discussed changes with code owners prior to submitting this pull request
I have not skipped any automated checks
All existing and new tests passed
I have updated the documentation accordingly

Make sure to have followed the contributing guidelines before submitting.

…eat/mulitLLMConnections

…tions

…onsWithShutdown # Conflicts: # packages/@webex/internal-plugin-mercury/src/mercury.js

…down' into feat/mulitLLMConnectionsWithShutdown

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0045311f81

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

chatgpt-codex-connector · 2026-03-04T03:09:49Z

+    for (const sessionId of this.sockets.keys()) {
+      disconnectPromises.push(this.disconnect(options, sessionId));
+    }


Abort in-flight connects when disconnecting all sessions

disconnectAll() only iterates this.sockets.keys(), but connect() now tracks in-flight attempts in _connectPromises before any socket exists (during device.register() and early backoff setup). If logout()/disconnectAll() is called in that window, the session is skipped here and the underlying connect attempt continues, so a socket can still open after the caller believes all sessions were disconnected.

Useful? React with 👍 / 👎.

…onsWithShutdown

…down' into feat/mulitLLMConnectionsWithShutdown

rarajes2

Have reviewed the Mercury plugin. will do another round of review for llm plugin.
Let's update the README.md for both the llm and mercury plugins with old and new ways to consume the connections.
If possible please add a vidcast showcasing the working of existing features such as meeting/calling as this is affecting a lot of packages

rarajes2 · 2026-03-05T07:03:55Z

+   * @returns {boolean} True if at least one socket is connected
+   */
+  hasConnectedSockets() {
+    const socket = this.sockets.get(this.defaultSessionId);


Can we also add support for the provided sessionId and default/fallback could be defaultSessionId ?

We can simply return the socket?.connected

rarajes2 · 2026-03-06T03:03:08Z

+  },
+
+  // @oneFlight
+  connect(webSocketUrl, sessionId = this.defaultSessionId) {


Please add js-doc for this method

Does this sessionId need to be generated and maintained by Consumer application? See if we can generate the sessionId internally and return the connection object to send and listen for the messages to/from backend where the webSocketUrl will be passed manually from the Application. The default listening on the webex.internal.mercury.on() should work as expected for the usual connection

Yes we can generate the sessionId internally if Consumer application tell us which data channel it intend to connect. We can improve this in later PR

rarajes2 · 2026-03-06T03:09:26Z

+        return this._connectWithBackoff(webSocketUrl, sessionId);
+      })
+      .finally(() => {
+        this._connectPromises.delete(sessionId);


Q. Why are we deleting the created promise here ? Shouldn't this be deleted on disconnect ?

_connectPromises is an internal map (keyed by sessionId) that caches the in-flight “connect” Promise for each Mercury session.
Functionally it is used to:
De-duplicate concurrent connect calls: if connect() (or the internal connect path) is called multiple times for the same sessionId while a connection attempt is already running, the code can return the existing promise from _connectPromises instead of starting another socket open/backoff cycle.
Provide a single completion signal per session: all callers awaiting a connect for that sessionId resolve/reject from the same promise, so state (connected, connecting, socket refs) stays consistent.
Allow cleanup/reset: once the connect attempt finishes (success, permanent failure, or abort/disconnect), the entry should be removed/cleared so a future connect starts a fresh attempt rather than reusing a settled promise.
In short: _connectPromises is the per-session “only one connect attempt at a time” mechanism.

rarajes2 · 2026-03-06T03:24:09Z

@@ -361,8 +498,8 @@ const Mercury = WebexPlugin.extend({
      return Promise.reject(err);
    }

-    if (!isShutdownSwitchover && !this.backoffCall) {
-      const msg = `${this.namespace}: prevent socket open when backoffCall no longer defined`;
+    if (!isShutdownSwitchover && !backoffCall) {
+      const msg = `${this.namespace}: prevent socket open when backoffCall no longer defined for ${sessionId}`;
      const err = new Error(msg);



Same condition is repeated inside single function

Not same, but can simplify it

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: eef7a6d14c

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

chatgpt-codex-connector · 2026-03-08T13:11:58Z

    this.onmessage = this.onmessage.bind(this);
    this.onclose = this.onclose.bind(this);
+    // Increase max listeners to avoid memory leak warning in tests
+    this.setMaxListeners(5);


Increase Socket max listeners above connection baseline

Setting this.setMaxListeners(5) is lower than the number of listeners a normal connection adds: Mercury attaches five listeners in _attachSocketEventListeners() (close, message, pong, sequence-mismatch, ping-pong-latency), and Socket._authorize() adds another message listener with once(). That means each connection now crosses the limit and emits MaxListenersExceededWarning during normal operation, creating noisy diagnostics and masking real leak signals.

Useful? React with 👍 / 👎.

mickelr · 2026-03-09T02:22:26Z

https://app.vidcast.io/share/10b65fc9-aa88-404d-8848-275947fa6d73
https://app.vidcast.io/share/092ee802-353c-4910-b70b-54f127c88a03
vidcasts for meeting/calling

…down' into feat/mulitLLMConnectionsWithShutdown

rarajes2

For calling we should test with the web.webex.com/calling?calling url. The one shown in the vidcast uses the meeting itself

rarajes2 · 2026-03-11T04:29:08Z

+```js
+const mercury = webex.internal.mercury;
+
+// Default session
+await mercury.connect();
+
+// Additional session
+await mercury.connect(undefined, 'secondary-session');
+
+// Disconnect only one session
+await mercury.disconnect(undefined, 'secondary-session');
+
+// Disconnect everything
+await mercury.disconnectAll();
+```


Would be nice to have sample code for how to listen to multiple connections

rarajes2 · 2026-03-11T13:37:38Z

+await mercury.connect();
+
+// Additional session
+await mercury.connect(undefined, 'secondary-session');


Q. Instead of undefined we can pass a custom wss url which will be able to connect to any wss server, right?

rarajes2 · 2026-03-11T14:04:20Z

+await webex.internal.llm.disconnectLLM({code: 1000, reason: 'done'}, 'session-a');
+
+// Disconnect all sessions
+await webex.internal.llm.disconnectAllLLM({code: 1000, reason: 'shutdown'});


Same here, it will be nice to have listener logic documented as well

rarajes2 · 2026-03-11T14:06:12Z

+const SOCKET_READY_STATE = Object.freeze({
+  CONNECTING: 0,
+  OPEN: 1,
+  CLOSING: 2,
+  CLOSED: 3,
+});


Should we create a constants.ts file for consistency ?

rarajes2 · 2026-03-11T14:07:58Z

+export default // @ts-ignore
+// @ts-ignore


@ts-ignore is repeated and seems bit off

rarajes2 · 2026-03-11T14:22:59Z

+   * @returns {Promise<void>}
+   */
+  public disconnectAllLLM = (options?: {code: number; reason: string}): Promise<void> =>
+    this.disconnectAll(options).then(() => {


Q. Is this coming from Mercury plugin ?

meeting-plugin will call the interface

mickelr · 2026-03-12T02:30:59Z

https://app.vidcast.io/share/9052d53e-0a6c-4970-a21d-8ddf71fa8ff3
vidcast for calling

…onsWithShutdown

rarajes2

Approving the PR. Did a sanity test on the calling sdk. It seems to be working fine with the updated changes in the socket plugins.

Please check one last comment regarding connected state and take a call on how do we want to handle it.

rarajes2 · 2026-03-13T03:56:25Z

      resolve();
+
+      // Update overall connected status
+      this.connected = this.hasConnectedSockets();


Suppose we have multiple connections alive and if the default connection is closed but other connections are still intact, in that case we are marking here that overall connection is lost. This feels bit misleading.

Should we maintain a map of session IDs including the default one with the connected property and based on that every connection can be decided to be connected or disconnected ?

This change will need to be updated with for the derived ampersand property as well. Shown below:

derived: { listening: { deps: ['connected'], fn() { return this.connected; }, }, },

Please discuss this with your team and take a call on this.

github-actions · 2026-03-13T07:14:56Z

🎉 Your changes are now available!

Released in: v3.11.0-next.52
Packages updated: @webex/internal-plugin-llm, @webex/internal-plugin-mercury

📖 View full changelog →

Thank you for your contribution!

_{🤖 This is an automated message. For questions, please refer to the documentation.}

Co-authored-by: Shreyas Sharma <72344404+Shreyas281299@users.noreply.github.com>

mickelr and others added 30 commits September 10, 2025 16:52

feat(webinar): refactor mercury/llm to support multiple connections

3c96269

feat(webinar): refactor mercury/llm to support multiple connections

6406e7c

Merge remote-tracking branch 'origin/next' into feat/mulitLLMConnections

af94e46

feat(webinar): refactor mercury/llm to support multiple connections

17e7e26

Merge remote-tracking branch 'origin/next' into feat/mulitLLMConnections

d78f99a

feat(webinar): modify unit test

6ca77ca

Merge remote-tracking branch 'origin/next' into feat/mulitLLMConnections

3ae1629

feat(multipleLLM): make unit test work for mercury and llm

694a6e0

feat(multipleLLM): for deploy feature branch

08a72ce

Merge branch 'next' into feat/mulitLLMConnections

7f5e301

feat(multipleLLM): update log

58598b0

Merge remote-tracking branch 'webex/next' into feat/mulitLLMConnections

ea420cf

Merge remote-tracking branch 'origin/feat/mulitLLMConnections' into f…

91197c8

…eat/mulitLLMConnections

feat(multipleLLM): force publish

69548b5

feat(multipleLLM): force publish

f70e3ef

feat(multipleLLM): force publish (webex#4539)

51ec0b8

feat(legacy-tools): publish all packages (webex#4543)

62cece6

Merge remote-tracking branch 'upstream/next' into multi-llms

556ceee

feat(multipleLLM): merge multiple LLM channel logic

46d11cb

Merge remote-tracking branch 'upstream/multi-llms' into multi-llms

1dd547e

Merge remote-tracking branch 'upstream/next' into feat/mulitLLMConnec…

439177c

…tions

feat(multiLLM): modify shutdown logic

65a98ac

feat(multiLLM): modify shutdown logic

09e4e58

feat(multiLLM): modify shutdown logic

b096015

Merge remote-tracking branch 'origin/next' into feat/mulitLLMConnecti…

955ebd9

…onsWithShutdown # Conflicts: # packages/@webex/internal-plugin-mercury/src/mercury.js

feat(multiLLM): modify shutdown logic

b32c452

feat(multiLLM): rewrite _emit function

b15d82b

Merge branch 'multi-llms' into feat/mulitLLMConnectionsWithShutdown

db4d5c3

feat(multiLLM): make this.socket as default session one

abeeabd

Merge remote-tracking branch 'origin/feat/mulitLLMConnectionsWithShut…

c5c5fde

…down' into feat/mulitLLMConnectionsWithShutdown

feat(multi-llm): roll back

0045311

chatgpt-codex-connector Bot reviewed Mar 4, 2026

View reviewed changes

mickelr and others added 3 commits March 4, 2026 17:33

Merge branch 'next' into feat/mulitLLMConnectionsWithShutdown

347a269

Merge remote-tracking branch 'origin/next' into feat/mulitLLMConnecti…

646b3d4

…onsWithShutdown

Merge remote-tracking branch 'origin/feat/mulitLLMConnectionsWithShut…

be29181

…down' into feat/mulitLLMConnectionsWithShutdown

rarajes2 reviewed Mar 6, 2026

View reviewed changes

mickelr and others added 3 commits March 8, 2026 20:51

feat(multiple data channel): update doc

02770db

feat(multiple data channel): update doc

237860a

Merge branch 'next' into feat/mulitLLMConnectionsWithShutdown

eef7a6d

mickelr requested a review from rarajes2 March 8, 2026 13:08

chatgpt-codex-connector Bot reviewed Mar 8, 2026

View reviewed changes

mickelr and others added 3 commits March 10, 2026 09:19

Merge branch 'next' into feat/mulitLLMConnectionsWithShutdown

a25f531

feat(multiple data channel): increase max listeners

d44237c

Merge remote-tracking branch 'origin/feat/mulitLLMConnectionsWithShut…

f81de23

…down' into feat/mulitLLMConnectionsWithShutdown

rarajes2 reviewed Mar 11, 2026

View reviewed changes

feat(multiple data channel): add new constants file

5407b9f

mickelr added 2 commits March 12, 2026 10:36

Merge remote-tracking branch 'origin/next' into feat/mulitLLMConnecti…

05b7f22

…onsWithShutdown

feat(multiple data channel): update doc

5da4716

mickelr requested a review from rarajes2 March 12, 2026 05:52

rarajes2 approved these changes Mar 13, 2026

View reviewed changes

mickelr and others added 2 commits March 13, 2026 13:35

feat(multiple data channel): fix query parameters format error

c0288f8

Merge branch 'next' into feat/mulitLLMConnectionsWithShutdown

2b676a8

JudyZhuHz approved these changes Mar 13, 2026

View reviewed changes

JudyZhuHz merged commit 12d4fd5 into webex:next Mar 13, 2026
12 checks passed

mickelr mentioned this pull request Mar 13, 2026

Feat(webinar 5k) - keep both datachannels for main and practice session when in PS #4772

Merged

23 tasks

fnowakow pushed a commit to fnowakow/webex-js-sdk that referenced this pull request Mar 20, 2026

Feat/mulitple data channel(llm) connections (webex#4661)

ca69fc5

Co-authored-by: Shreyas Sharma <72344404+Shreyas281299@users.noreply.github.com>

vamshigovardhana pushed a commit to vamshigovardhana/webex-js-sdk that referenced this pull request Apr 1, 2026

Feat/mulitple data channel(llm) connections (webex#4661)

adbb683

Co-authored-by: Shreyas Sharma <72344404+Shreyas281299@users.noreply.github.com>

Conversation

mickelr commented Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

COMPLETES #< INSERT LINK TO ISSUE >

This pull request addresses

by making the following changes

Change Type

The following scenarios were tested

The GAI Coding Policy And Copyright Annotation Best Practices

I certified that

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 4, 2026

Choose a reason for hiding this comment

Uh oh!

rarajes2 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mickelr Mar 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mickelr Mar 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

mickelr commented Mar 9, 2026

Uh oh!

rarajes2 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mickelr commented Mar 12, 2026

Uh oh!

rarajes2 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions Bot commented Mar 13, 2026

🎉 Your changes are now available!

mickelr commented Jan 20, 2026 •

edited

Loading

mickelr Mar 8, 2026 •

edited

Loading

mickelr Mar 8, 2026 •

edited

Loading