V20 Update 10 Alpha: MCP, AI Server & More Admin Control

[Voiceking.be]

"We are actually planning to bring this visibility directly into the subscription list so that you can proactively filter or find these subscriptions without having to attempt placing an order first."

That would be great!

 

[SAV.IB]

Hello,

Thank you for reintroducing support for Yealink Android phones.

However, it is unfortunate that the Yealink T88W Pro phones are still not supported. I hope they will be included in a future beta release.

Kind regards,
Alain

 

[KyriacosS_3CX]

Im falling behind with the AI side of 3CX having touched none of it. Whats best for me to do as a starting point? I dont want to be stuck at the port waving the ship of as it leaves

Actually, the new MCP functionality is the simplest thing to get to grips with. 3 clicks (number of clicks may vary ) and you can start talking to your PBX.

 

[MichaelB_3CX]

Same issue here. Is it possible that this is the same issue as the one with the AI Agents integration in U8 Alpha that was caused by the system "language/locale" settings?

Ensure please you have the new models (gpt-realtime-2 or gpt-realtime-1.5 and gpt-realtime-whisper) allowed and not blocked on the Open AI platform.

 

[Sonal-MA]

Hi,


Very nice update. Can't wait to test the MCP features. You guys are killing it!

I have one question about the Caller ID selection. I read this is only for users that have multiple DDI's assigned to them. We have customers with muli-queue agents who need to change their caller ID depending on the departpment they use for that call. The DDI's are assigned to their queues and not the users themselves. It would be awsome if they could select Queue assigned DDI's as well. At this moment we use multiple outbound rules and prefixes to handle this.

Thanks.

 

[Evolute IT]

Getting this error when calling a DID linked to an AI agent (receptionist.) U10, standard prompt, no KBs. It just fails and reverts back to the normal IVR (my fallback destination.) Yes, models are authorized in OpenAI.

 

[Alejandro_3CX]

Getting this error when calling a DID linked to an AI agent (receptionist.) U10, standard prompt, no KBs. It just fails and reverts back to the normal IVR (my fallback destination.) Yes, models are authorized in OpenAI.
View attachment 52145

is this just with the external calls, right? what is the sip trunk? please check in this sip trunk that you are using only G711u (PCMU), G711a (PCMA), then make a new test.

 

[Evolute IT]

is this just with the external calls, right? what is the sip trunk? please check in this sip trunk that you are using only G711u (PCMU), G711a (PCMA), then make a new test.

Even internal calls. It just plays MoH when calling from internal, it's trying to send the call back to me somehow, but none of the config tells it to do that.

SIP trunk is locked to PCMU ever since the first AI agents.

 

[Alejandro_3CX]

Even internal calls. It just plays MoH when calling from internal, it's trying to send the call back to me somehow, but none of the config tells it to do that.

SIP trunk is locked to PCMU ever since the first AI agents.

ok if it is not working for you even with internal calls, then please set verbose logging, restart the AI service, perform a test and check the 3cxai.log, you may open a support ticket for further troubleshooting.

 

[MichaelB_3CX]

Please ensure you have it

Getting this error when calling a DID linked to an AI agent (receptionist.) U10, standard prompt, no KBs. It just fails and reverts back to the normal IVR (my fallback destination.) Yes, models are authorized in OpenAI.
View attachment 52145

See comment above - ensure you have the following allowed models on the AI Project i,e gpt-realtime-1.5, gpt-realtime-2, gpt-realtime-whisper.

 

[Ryan_Fisher_Techmode]

What is the provider you have? We need to see the logs.
If you call the AI from a local extension does it work?
You can contact me on pm on this case - I'm reserving some time today for the community..

I have some new info on this that may explain what the user above is experiencing as well.

I have now tested this on two seperate systems. Worked perfect on U9. Upgrade to U10 Alpha and it breaks. They both behave exactly the same. At first I thought this was for external calls only but what I have found in testing as that the AI agent will only answer from the internal extension for its boss. If I call internally from any other extension on the system its the same behavior as the external calls. Hang up / Fallback extension.

Error code 4096 "AI Bridge" 500 Server Internal Error

 

[Alejandro_3CX]

I have some new info on this that may explain what the user above is experiencing as well.

I have now tested this on two seperate systems. Worked perfect on U9. Upgrade to U10 Alpha and it breaks. They both behave exactly the same. At first I thought this was for external calls only but what I have found in testing as that the AI agent will only answer from the internal extension for its boss. If I call internally from any other extension on the system its the same behavior as the external calls. Hang up / Fallback extension.

Error code 4096 "AI Bridge" 500 Server Internal Error

@Ryan_Fisher_Techmode please check Mike's response above

ensure you have the following allowed models on the AI Project i,e gpt-realtime-1.5, gpt-realtime-2, gpt-realtime-whisper.

 

[Ryan_Fisher_Techmode]

Alejando-

Thanks. I have verified the set up.I am positive the api is working because i get full functionality out of it when i call internally as its boss.

 

[Evolute IT]

I have some new info on this that may explain what the user above is experiencing as well.

I have now tested this on two seperate systems. Worked perfect on U9. Upgrade to U10 Alpha and it breaks. They both behave exactly the same. At first I thought this was for external calls only but what I have found in testing as that the AI agent will only answer from the internal extension for its boss. If I call internally from any other extension on the system its the same behavior as the external calls. Hang up / Fallback extension.

Error code 4096 "AI Bridge" 500 Server Internal Error

@Ryan_Fisher_Techmode please check Mike's response above

ensure you have the following allowed models on the AI Project i,e gpt-realtime-1.5, gpt-realtime-2, gpt-realtime-whisper.

Exact same issue as me. Models are allowed and there are plenty of credits. It just fails for some reason (setup is configured for realtime-2 and gpt-5.6-luna for chat). Internal or external, neither work.

 

[Alejandro_3CX]

@Ryan_Fisher_Techmode please open a support ticket for further troubleshooting as I have mentioned in my other messages. Thanks!

 

[nikolascx]

Hello Dears
I confirmed and double checked.
From my side everything is working. However one of our clients had the problem and now it suddenly started to work.
Can you try again?

Strange because https://status.openai.com/ always showed up.

No changes are required to 3CX. Please everyone try again now.

 

[Evolute IT]

Hello Dears
I confirmed and double checked.
From my side everything is working. However one of our clients had the problem and now it suddenly started to work.
Can you try again?

Strange because https://status.openai.com/ always showed up.

No changes are required to 3CX. Please everyone try again now.

Confirmed. It seems it started working but is not super stable. Probably OpenAI issues! I'll keep testing and report back if I have any new clues.

 

[nikolascx]

It's stable dont worry. And it was openai Sending 500 (with empty message)
We also contacted openai in the meantimem.
Sometimes it happens. I think having 1 billion users it's okay if sometimes they give a 500 for some cases.
For us it never failed. So it wasnt a global downtime.

 

[Nick - SPN Co]

Yes music on hold only, if a call is parked it uses the music on hold that you set within System —- Call Parking —- Music on hold file for parked calls.

It would be awesome if the park keys could be set to play the default department music so they could operate in the same way as the hold does per department, hopefully that is an easy feature add on for you guys to build.

 

[Nick - SPN Co]

So many great new features in this update! The pop out dialler is a game changer for the PWA app, our customers are going to love this! Switching phone templates is going to save us so much time with custom handset changes too!