V20 Update 8 Alpha - The AI Chapter

[Nik_si]

Hello, Regarding Department Phonebook I have only this options, what goes wrong? (16SC pro)

View attachment 50016

Hello, can someone explain me what to do to reach department based phonebook?

 

[N_G]

I just want to confirm, is it currently possible to use the same GPU server for multiple 3cx instances, or at this time only a 1:1 configuration? If possible, do we just run the same install again on the GPU server that already has an existing 3cx tied to it?

Also - for feature improvement, could you add down alert similar to when a trunk or SBC is unreachable, and for additional clarification, would "failover" be possible, if local GPU is not reachable/down, could say 3CX Cloud (if configured), be able to failover to ?

Hi, This will be possible in future updates but not right now.... But its in the plan

 

[Xiente Admin]

We already spoke with 11 Labs, and after running the numbers, it turns out a large segment of our users would face a standard bill ranging between $1,000–$3,000 per month. The pricing is simply mind-blowing.

You can reach out to them yourself -- I already tried. I was actually waiting for this thread to come up.

As an AI enthusiast and academic, I genuinely admire their software. However, the way they handle partnerships is incredibly one-sided.

They expected us to develop the integration, attach it to our MCP, embed it in our product, promote it across 190 countries, and expose them to our entire install base and user network…

And in return? We were supposed to be satisfied with $300 worth of development credits. They genuinely believed that was a fair exchange.

Their inability to distinguish between fair collaboration and exploitation - or as I call it "Software Slavery" was truly astonishing.

Hi,

New here.
Have you guys considered Chatterbox that is an Opensource and more powerful that Elevenlabs?

 

[Xiente Admin]

I also have the issue of the AI agent not saving the call routing destinations. Has a fix been found yet?

I have the same problem.

 

[noord]

I have the same problem.

This has been confirmed as a bug and will be fixed in the next alpha (or beta).

 

[cmp1]

Another Item for the FUTURE for the OpenAI API configuration - currently asking for the Company Name and Description .... this will be problematic for MT installations.

 

[N_G]

Another Item for the FUTURE for the OpenAI API configuration - currently asking for the Company Name and Description .... this will be problematic for MT installations.

Yes good point

 

[KyriacosS_3CX]

Hello, can someone explain me what to do to reach department based phonebook?

When you try to add a new contact you should see the option to save it as a Department contact - see below. If this is not their, first clear your browser cache & cookies and check again. If not, let us know the specifics of your install and we can advice further.

 

[KyriacosS_3CX]

Another Item for the FUTURE for the OpenAI API configuration - currently asking for the Company Name and Description .... this will be problematic for MT installations.

For now:

 

[nikolascx]

I just want to confirm, is it currently possible to use the same GPU server for multiple 3cx instances, or at this time only a 1:1 configuration? If possible, do we just run the same install again on the GPU server that already has an existing 3cx tied to it?

Also - for feature improvement, could you add down alert similar to when a trunk or SBC is unreachable, and for additional clarification, would "failover" be possible, if local GPU is not reachable/down, could say 3CX Cloud (if configured), be able to failover to ?

Hi @cmp1
Appreciate the question and the way you are fast forwarding this.

For now 1:1.
So if you have onboard AI, it will take a pipeline of transcription jobs from that same machine.

As for your second request, lets keep 1 gpu busy first.. Moving from local to cloud is not the right way. You will stall and starve the gpu's until the resource moves from local domain to the cloud. If you want failover, you need standby GPU nodes with an orchestration that does traditional failover when one GPU fails. You cannot expect to use 3CX Cloud for this.

Lets make this perfect onboard first, with a machine with one gpu or at most multiple gpu's stacked in one server. Then we can expand later.

Maybe we can consider MIG or GPU virtualization later - but this is datacenter level.

thanks so much!

 

[nikolascx]

Hello, can someone explain me what to do to reach department based phonebook?

Good question dear @Nik_si
You are right to question this. It's coming soon. For now you can save and populate trusting it will be in the department phonebook "bucket"
Soon you will be able to see this..

 

[nikolascx]

Hi,

New here.
Have you guys considered Chatterbox that is an Opensource and more powerful that Elevenlabs?

Thanks for the heads up. I never heard of it and I will definitely check it out.
is it by resemble.ai? I did hear of resemble AI..
https://www.resemble.ai/chatterbox/
thanks for sharing.

 

[nikolascx]

I have the same problem.

Contact me privately and send me your email and I will get one of our engineers to install an updated build for you.
I would need a PM with your email, Problem and OS (windows or Debian)

 

[nikolascx]

Another Item for the FUTURE for the OpenAI API configuration - currently asking for the Company Name and Description .... this will be problematic for MT installations.

Of course..

This is on our roadmap. But if we thought of MT First, we would have released next year. Anyway you can achieve this right now if you want.

If you want to create a swarm of agents each with a different company.
Idea is to set the company name and description as empty, and tell the AI that for this environment, it should get name and desc from the main system prompt.
Example:

Company name put: <defined in main prompt>
Description: <defined in main prompt>

then In the main prompt start with this:

IMPORTANT:
{{ company_name }} Variable use this company name instead "INDIGO"
Do not say that this is an internal reference used by this environment.

Company description: Metropolis Healthcare, also known as Metropolis Labs, is an Indian multinational chain of diagnostic labs, with its central laboratory in Mumbai, Maharashtra. Metropolis Healthcare has a chain of 124 clinical laboratories[4] and 2400 collection centers etc etc

Let me know how this works for you.

——

Also the billing for now is 1 API Key for 1 PBX but we are also going to change it.

To sum this up, in the future, we will take out Key, Name, Company, Description and move it per Swarm (or Department / team)

Like this you can have autonomous agents representing different companies and all being billed separately.

Great work team!

 

[nikolascx]

For now:
View attachment 50064


View attachment 50065

Another Item for the FUTURE for the OpenAI API configuration - currently asking for the Company Name and Description .... this will be problematic for MT installations.

Actually this approach is cleaner but the idea is the same.

Sweet @KyriacosS_3CX

 

[Evolute IT]

Contact me privately and send me your email and I will get one of our engineers to install an updated build for you.
I would need a PM with your email, Problem and OS (windows or Debian)

Same here please lol

 

[robinsonconsulting]

Hi @noord , I wonder where are you routing the call? if it goes to an user extension, then he will be screened with that info in the web client when receiving the call.

Hi Alejandro I'm experiencing the same results. The AI Agent confirms the conversation however the message is not delivered. I've also turned on Transcription and call recording using Google & GCP so I can see the conversation. I might have missed your follow up message regarding this.Very exciting update!

 

[Colorado VoIP]

Been messing around with the AI agent this morning and it's extremely cool. We've already got a multi-tenant customer frothing to implement this lol!

I've noticed when I call the extension, it's supposed to say "Welcome, you've reached {{ company_name }} - this is {{ agent_name }}. How can I help you today?" but instead it's saying "Hello, this is test agent at reception. how many i help you reach the right person or department?"

maybe this is because i'm just calling internally right now and it knows that? just curious. awesome either way! can't wait to see what the updates hold for this!

 

[Colorado VoIP]

correction to my previous post! it was because I was calling internally. assigned a DID to the system and bam! working as script indicated.

 

[Nik_si]

When you try to add a new contact you should see the option to save it as a Department contact - see below. If this is not their, first clear your browser cache & cookies and check again. If not, let us know the specifics of your install and we can advice further.
View attachment 50063

Hello, I tried cache, incognito also but it's still not visible




it's 16SC, self-hosted, latest update, Multi Tenant mode is enabled.