Thank you for this explanation! You just saved me hours of research and figuring out the basics of this program. Something that I've been trying to look for, for a while. I'm glad your video popped up on my youtube list!!
Glad to read it is helpful! If there's anything else you'd like to see, feel free to share it - happy to see if I can incorporate into one of my videos!
Thank you Jannis for your Json Utility!!! and this videos, looking forward to next ones . Is there a way to train this assistants and fine-tune them inside Vapi as we can do with playground in OpenAI?
Hey, glad to read it is helpful. Yes, you can implement a fine-tuned version using the custom LLM module. This way you can fine tune openAI directly and leverage Vapi's environment: docs.vapi.ai/assistants/custom-llm
@@jannismoore Thanks for your reply. I am actually trying to configure it to get less latency, but when I try to get my openAI fine tuned assistant, Vapi makes an autocompletion in the URL I set, hence not allowing me to access it directly, have you had the same issue? BTW have u thought to make a live workshop for Q&A or presential event with UAE's newbie learners ? Thanks in advance🤝
@@BiaBlanco-jb2ei fine tuned models require a custom LLM setup. This means you would need to have your own backend that can handle the Vapi requests. I don't think the simple model dropdown will work for that. As for the event: I haven't heard of that one. Feel free to share some more details.🙏
I'm trying to use Custom Functions in VAPI and am seeing issues where the agent (Open AI / GPT 4 Turbo Cluster) is not always placing a call to my server. I heard you say that function calls are deprecated by Open AI at 17:41. I'm not sure how it's working sometimes and not working other times and just trying to get to the bottom of it. Thanks for the great video by the way; very helpful!
function calls itself are not deprecated, but the naming convention is. They now call it tools instead of functions, which is the only difference. The logic is still there and can be actively used within the Voice Assistants too. If it doesn't fire the function, your best bet is to work on your prompting. You can, for example, add a storyline to your prompt that tells the assistant exactly when to fire it. This is where trial and error comes from, and where I do most of consulting work. Especially with prompt engineering, you often need to tweak prompts to make sure you have the least hallucinations. I suggest looking into markdown formatting for prompts. That's a great step forward.
Certainly. I have a long form video coming up on that, but I'm happy to add one dedicated for function calling. Is there anything specific you'd like to see about it? E.g. for static assistants or for transient-based assistants?
@@jannismoore not really specific for the moment but just to understanding the functions overall, its something i don't use for the moment. For transient base yeah :)
Yes!! Please expand on VAPI for Customer Service, Appt Setting and Google Sheet input. I believe those are the top 3 type of AI Agents clients are asking to develop and use.
Sure, I can do. What of those is most important to you? The actual structure or the prompting? Each of those topics would be a separate video as there's a lot to share about.
@@jannismoore For the Customer Service to take calls on Client's phone number and yes the structure to intake information and instant deliver of information to Client in Google Sheets. Later on maybe a connect to a Zapier Automation for marketng purposes. While I am typing I am realizing that I really need to learn more about MAKE. I am about to watch right now! 👍😄
Hey Charles, I do offer consultations (By default via Google Meet). If you prefer Zoom, simply mention that during the calendar booking: integraticus.com/consultation/ Feel free to let me know if you have any questions.
@@jannismoore The biggest issue is handling voicemail, detecting voicemail and sometimes it hangs up randomly. These will all be worked out because their development team is working on improving it all the time.
Dumb question. I've made a few static Assistants and am trying to integrate Make. I have a webhook and Google Sheet that I want to update with data from the caller. I've gotten as far as updating the sheet with the CallerID and Timestamp but I haven't figured out how to get the data spoken by the caller into the sheet. What am I missing?
Not as of now, but they're working on it. It's a feature that is currently in progress. If you want to achieve if right now, you probably have to use a custom Vector Database such as Qdrant
Hey Jannis, i was just wondering when setting up your vapi like this, is it possible to have multible different builds? as i understand the phone numbers would all go to that one e.g. make webhook? or am i missing something
Lots of good content here. It would be great if you didn't have so many cuts in the video so your presentation is more concise and the flow isn't so jittery or jaring.
Hi , love you walk-through of vapi and the difference between static and transient. Also really interested in client sdk, I could catch the name of the company you mentioned, that has the example. Could you let me know please? Thank you very much
Glad it's helpful! I have a complete tutorial on the Client SDK available here: ruclips.net/video/XOVt1qHmrlg/видео.html Is that what you're looking for?
Am I responsible for adjusting and sending the assistant back manually? or does a predefined scenario adjust the JSON construct and send it back upon receiving a call?
hi Jannis, im new to voice agents, which infrastructure provider would you recommend for a newbie like me i who wants to get a grip on ai voice agents? thanks Jannis.
Vapi :) It’s definitely not the easiest to learn, but you’ll see how much sense it makes during project deliveries, especially for clients with more tailored needs
I've talked to Vapi's support and they mentioned the limit is 10 concurrent calls at a time. Have you found this to be limiting? Also, have you tried Air Ai? another AI agent platform
Quite a few of our clients are converting from air ai as Vapi brings the flexibility and quality they look for. The concurrency limit is often a discussion, but it’s incredibly hard to reach. I’d say it’s the equivalent of having 25000 calls a month, which alone is already fairly difficult to reach. We usually chunk the amount of calls and spread it out over a given amount of time. At last, we do have some clients that exceed those limits, but for that we can directly arrange a customized package that suits their specific needs. Hope that helps.
@@jannismoore thank you. Ive been wanting to use it for my mortgage broker agency for lead appointment setting. How are you using it with your clients and how are you able to chunk the calls? I want to be able to call the leads within a couple minutes of them coming in, and am wondering if the concurrency limit will be an issue. Is there a way to structure a call queue?
@@harchitb You can do that using simple search queries with mostly any DB provider. Airtable and Google Sheets work well too. You basically bundle a given number of records per scenario
Hey man! This is pure gold. I am working on a similar project and this video just happened to pop on my feed. I actually have a few questions and projects we need help on. How can we connect?
You can do that using tools like Vocode, but Vapi is still superior as that's all they work on since founding this tool. For me it's a no brainer that Vapi is the go to choice in the AI Voice space.
@@atulboylla there's no coding necessary. JSON is a formatting language. It's one of the core concepts I've explained thoroughly on my channel. See this for example: ruclips.net/video/gNdEwr-sQg0/видео.html
@@blokmaster2952 Yes, but those are usually considered custom work. You can get directly onboarded to your own telephony server if you want to. Otherwise, if you want have more flexibility, you can look at building out your own solution using something like Vocode.
Is it possible to set up to capture emails correctly from the user and book an appointment? It always captures wrong emails and then i can't send the invatation link to a meeting
Via phone it's super hard. We've tried many ways, but if you do it via voice, you always have plenty of room for error. The best is to just create some custom logic that can send an SMS link and then use that one to request the email from the user, so they can simply fill it up during or after the call. This would then require some more automation to delay the event (whatever that may be) until you got the email.
It's similar, but with tons of features build on top that aren't really necessary in many cases. I've worked with Bland in the past, but stopped it as they haven't done a great job with their version control and backward compatibility. I've had it many times that phone numbers just stopped working in the middle of campaigns because they pushed a new release.
Yes, that's all functionality you can build on top of their infrastructure. As for Pineapple - I haven't heard of that Tool/SaaS, but if it has an API, you can connect it too.
Thanks for this - I have been trying Bland and Retell, but am interested in trying this with some of my Make Scenarios. I have no issues when there is a webhook, but in my current Make scenario with Bland, it starts with a google sheet, it runs through the list of numbers, through JSON and at the end through an http module where it calls the Bland API which then triggers the phone call. Could something similar be done with VAPI ?
Indeed. If you can stay patient until monday, I'll release a video and share some one-of-a-kind make.com templates for Vapi that makes it insanely easy to build custom agents with their platform.
@jannismoore Awesome tutorial thanks! I just found out that each agent created on the dashboard in Vapi has a 10 call concurrency limit. In the way you are creating the agent on the fly from your server, does this invalidate this restrictions because it's creating a new agent on each phone call? Additionally, does this have any limits? Can I send 1000 calls in at a time and have it create 1000 agents?
Based on my knowledge, the 10 call concurrency limit counts for the standard pay-as-you-go plans. IF you need more than that, feel free to reach out to me via our agency site integraticus.com/contact I will be able to help you get a limit that suits your needs.
@@jannismoore I was told on Vapi's discord that each assistant has that limit, not the account. So you get around it making 100 agents doing 10 calls each for 1000 calls, but that is a nightmare to manage. If this is true though, I wonder if your dynamic way would work to bypass it.
@@jannismoore On the API documentation on create assistant, in the json code to send the API it says "id string required - This is the unique identifier for the assistant." If this ID is required then how is this a transient/dynamic calling system? This ID field still limits the call rate to 10 concurrent calls.
I've covered most of it in the video. If you want to read more about it, I suggest checking out this URL: docs.vapi.ai/server-url Or: In case you have a specific questions, feel free to ask it right here. :)
Can you explain all the libraires we have to use to create similar vapi for example tts we can you deepgram and stt we can use playht but how can we create an api for this ??
If you want to build it yourself, you can look at something like Vocode, which is an Open Source framework that allows you to build conversational Voice Agents: vocode.dev/ I did a video on this earlier: ruclips.net/video/034Y4PoZCbU/видео.html
@@jannismoore but i want to bulid exact like vapi for example when user speak a question other side will be a ai model soo how can i bulid it i want it bulid it which library's i can use to bulid it
@@jannismoore i already bulid a model like it detect voice to text and that text give to model that resposne will convert into voice soo then we will here the voice but its too slow i want high speed work flow how can i bulid it
@@itzmeakash9695 Choosing the right server, optimize the streaming, and internal requests. Speed mostly comes from choosing the right infrastructure and the internal handling of functions. I sugget starting with revamping your code architecture by looking what functions or processes take most time.
It's around $0.10 - The price depends quite a bit on your selection of tools within Vapi, but once you create your own static assistant, you will see a price estimation directly within Vapi.
An alternative to the platform or an alternative to the problem that caused your assistant to stop working? :) It would certainly be helpful if you can provide some more details about the actual issue you're facing.
What is with the zoomed in screen recording with motion blur on the cursor? Is this kind of editing supposed to increase youtube views at the low cost of making it unwatchable?
@@jannismoore I want to make a call, but answers should be taken feom my database. Still not clear do you show this or not :D Or how it will be done. Also what is that pink website where you put your webhooks. Not clear what you are doing))
The assistant wrapper is mostly then relevant when you want to send it along in a request. As for Vapi: I suggest following their Swagger API docs - those are the most accurate one at any given time.
Thank you for this explanation! You just saved me hours of research and figuring out the basics of this program. Something that I've been trying to look for, for a while. I'm glad your video popped up on my youtube list!!
Glad to read it is helpful! If there's anything else you'd like to see, feel free to share it - happy to see if I can incorporate into one of my videos!
Love the video man! I completely agree that conversational AI needs to be dynamic because people are very unpredictable - especially on voice!
Absolutely killer video man we’ll done
I've been experimenting with Vapi and this guide was super helpful! Always killing it with the in depth content 💪🏼
Glad to read that! There's more crazy stuff in the pipeline for early next week. Full on Vapi so you will surely enjoy this!
@@jannismoore what to do next with this webhook?
@@PubgSpeed-wl8yo You can use the webhook to exchange data between Vapi and your server or no-code service: docs.vapi.ai/server-url
You nailed it Broo. 🔥🔥🔥
moore moore moore! you're amazing at what you do keep it up
Thank you Jannis for your Json Utility!!! and this videos, looking forward to next ones . Is there a way to train this assistants and fine-tune them inside Vapi as we can do with playground in OpenAI?
Hey, glad to read it is helpful. Yes, you can implement a fine-tuned version using the custom LLM module. This way you can fine tune openAI directly and leverage Vapi's environment: docs.vapi.ai/assistants/custom-llm
@@jannismoore Thanks for your reply. I am actually trying to configure it to get less latency, but when I try to get my openAI fine tuned assistant, Vapi makes an autocompletion in the URL I set, hence not allowing me to access it directly, have you had the same issue?
BTW have u thought to make a live workshop for Q&A or presential event with UAE's newbie learners ?
Thanks in advance🤝
@@BiaBlanco-jb2ei fine tuned models require a custom LLM setup. This means you would need to have your own backend that can handle the Vapi requests. I don't think the simple model dropdown will work for that.
As for the event: I haven't heard of that one. Feel free to share some more details.🙏
I'm trying to use Custom Functions in VAPI and am seeing issues where the agent (Open AI / GPT 4 Turbo Cluster) is not always placing a call to my server. I heard you say that function calls are deprecated by Open AI at 17:41. I'm not sure how it's working sometimes and not working other times and just trying to get to the bottom of it. Thanks for the great video by the way; very helpful!
function calls itself are not deprecated, but the naming convention is. They now call it tools instead of functions, which is the only difference. The logic is still there and can be actively used within the Voice Assistants too.
If it doesn't fire the function, your best bet is to work on your prompting. You can, for example, add a storyline to your prompt that tells the assistant exactly when to fire it.
This is where trial and error comes from, and where I do most of consulting work.
Especially with prompt engineering, you often need to tweak prompts to make sure you have the least hallucinations. I suggest looking into markdown formatting for prompts. That's a great step forward.
Finally someone is good at vapi, can you explain the concept of functions section in video pls ?
Certainly. I have a long form video coming up on that, but I'm happy to add one dedicated for function calling. Is there anything specific you'd like to see about it? E.g. for static assistants or for transient-based assistants?
@@jannismoore not really specific for the moment but just to understanding the functions overall, its something i don't use for the moment. For transient base yeah :)
@@Emilie-974 Just added this to my timeline. Will release a in-depth video on function calling soon :)
Yes!! Please expand on VAPI for Customer Service, Appt Setting and Google Sheet input. I believe those are the top 3 type of AI Agents clients are asking to develop and use.
Sure, I can do. What of those is most important to you? The actual structure or the prompting? Each of those topics would be a separate video as there's a lot to share about.
@@jannismoore For the Customer Service to take calls on Client's phone number and yes the structure to intake information and instant deliver of information to Client in Google Sheets. Later on maybe a connect to a Zapier Automation for marketng purposes. While I am typing I am realizing that I really need to learn more about MAKE. I am about to watch right now! 👍😄
@@jannismoore Thank you for your helpful videos and you’re a wonderful person for sharing your knowledge and resources!
@@ascension-spiritualawakeni7865 Thanks for the kind words, I'm happy to see it's helpful to you!
@Jannis Do you offer paid private Zoom consulting sessions, for those who have specific questions about Vapi, if yes, How do one sign up?
Hey Charles, I do offer consultations (By default via Google Meet). If you prefer Zoom, simply mention that during the calendar booking: integraticus.com/consultation/
Feel free to let me know if you have any questions.
Agreed. I love Vapi it's not perfect but none are right now. But Vapi is definitely worth trying.
Indeed it is! What are the problems you see with it?
@@jannismoore The biggest issue is handling voicemail, detecting voicemail and sometimes it hangs up randomly. These will all be worked out because their development team is working on improving it all the time.
@@stefano94103 Yup I see they are actively on it. I think it already improved compared to last month though.
bro thx for your stuff. Really great.
Konn des sein dass Du von Österreich bist?^^
Not quite, but close. I'm from Germany originally
Dumb question. I've made a few static Assistants and am trying to integrate Make. I have a webhook and Google Sheet that I want to update with data from the caller. I've gotten as far as updating the sheet with the CallerID and Timestamp but I haven't figured out how to get the data spoken by the caller into the sheet. What am I missing?
There's a key called "transcript" within the end-of-call-report. This one contains a readable version of the transcript
cool video men. but how can i connect some type of calender to my assistent so he knows when im available and when not ?
This depends on your calendar. You can achieve that using tool calling. I'll release a video on that soon
@@jannismoore would love to see it men !#
Is there a waitlist for your software? I'd love to learn more about this.
yes sir! onu.ai/
amazing video man
Nice video men! Is it possible to import a document with data into vapi? Thanks.
Not as of now, but they're working on it. It's a feature that is currently in progress. If you want to achieve if right now, you probably have to use a custom Vector Database such as Qdrant
Hey man. I have done all you said, but when i call, it says connect Vapi assistantId. But the server url is provided. What am I missing?
is there any way that we can do more for Speech Recognition part?currently my voice agent is having hard time to capture all name and email correctly.
Can vapi use data from GHL customer fields during conversation?
Yes, absolutely
Hey Jannis, i was just wondering when setting up your vapi like this, is it possible to have multible different builds?
as i understand the phone numbers would all go to that one e.g. make webhook? or am i missing something
Nope that’s right, but you can filter against the phone number within the scenario, or route them to a different one
Amazing content!🚀
VoiceOS and Sindarin are also very good. I’d love to see a comparison video!
Noted!
Do you need to pay for additional tiers for your API? My calls keep crashing
No, they have only one plan, except if you go enterprise, which isn't necessary for the cases I show in my videos.
Nice video 🎉 may I ask what do you use for your screen recording to be dynamic like this ? Thanks !
It's called Screen Studio
Lots of good content here. It would be great if you didn't have so many cuts in the video so your presentation is more concise and the flow isn't so jittery or jaring.
Appreciate the feedback
Hi , love you walk-through of vapi and the difference between static and transient. Also really interested in client sdk, I could catch the name of the company you mentioned, that has the example. Could you let me know please? Thank you very much
Glad it's helpful! I have a complete tutorial on the Client SDK available here: ruclips.net/video/XOVt1qHmrlg/видео.html
Is that what you're looking for?
Great video but that description...
> delve spotted 😆
Am I responsible for adjusting and sending the assistant back manually? or does a predefined scenario adjust the JSON construct and send it back upon receiving a call?
If you're using any of my templates, it works out of the box, but if you want to build it yourself, you're responsible returning the assistant JSON
hi Jannis, im new to voice agents, which infrastructure provider would you recommend for a newbie like me i who wants to get a grip on ai voice agents? thanks Jannis.
Vapi :)
It’s definitely not the easiest to learn, but you’ll see how much sense it makes during project deliveries, especially for clients with more tailored needs
Great video. Do you charge to set this up for others? I have a small business that makes outbound calls.
Hey Jimmy, yes we do. Feel free to send me a message with the details using the following link: integraticus.com/contact/
what is the link to the resources for this project please?
I've tons of resources for Vapi inside of my resource hub. There's also a dedicated playlist for this: ruclips.net/video/0Oew-dgehfM/видео.html
I've talked to Vapi's support and they mentioned the limit is 10 concurrent calls at a time. Have you found this to be limiting?
Also, have you tried Air Ai? another AI agent platform
Quite a few of our clients are converting from air ai as Vapi brings the flexibility and quality they look for.
The concurrency limit is often a discussion, but it’s incredibly hard to reach.
I’d say it’s the equivalent of having 25000 calls a month, which alone is already fairly difficult to reach.
We usually chunk the amount of calls and spread it out over a given amount of time.
At last, we do have some clients that exceed those limits, but for that we can directly arrange a customized package that suits their specific needs.
Hope that helps.
@@jannismoore thank you.
Ive been wanting to use it for my mortgage broker agency for lead appointment setting. How are you using it with your clients and how are you able to chunk the calls? I want to be able to call the leads within a couple minutes of them coming in, and am wondering if the concurrency limit will be an issue. Is there a way to structure a call queue?
@@harchitb You can do that using simple search queries with mostly any DB provider. Airtable and Google Sheets work well too.
You basically bundle a given number of records per scenario
Hey man! This is pure gold. I am working on a similar project and this video just happened to pop on my feed. I actually have a few questions and projects we need help on. How can we connect?
Glad it helps! Feel free to text me on LinkedIn: www.linkedin.com/in/thuemmig
@@jannismoore just sent you a request there. Looking forward to connecting!
what is the software used ( for recording the video with zooming in effects )
It's called Screen Studio: screen.studio/
Hi, thanks for this video, great ! Did you try any other languages than English? What is your experience with other languages?
Other languages work well too, especially with Azure. Do you have any specific example in mind?
with this logic, would it be better just to create your own ai calling software like vapi?
You can do that using tools like Vocode, but Vapi is still superior as that's all they work on since founding this tool. For me it's a no brainer that Vapi is the go to choice in the AI Voice space.
All the code you show what should we do with them copy past them somewhere for transient-based? Or do we just need maker to get transient call?
Hey, what kind of assistant are you looking for? If you can share some more details, I'm happy to give you some better guidance.
@@jannismoore just told you transient based
@@kasforai that doesn’t tell me if it’s inbound or outbound
@@jannismoore inbound sir
@@kasforai In that case you might want to look into this video: ruclips.net/video/RMOHpWAPan8/видео.html
this video is not for newbie
Where did you get stuck? :)
@@jannismoorecoding part
@@atulboylla there's no coding necessary. JSON is a formatting language. It's one of the core concepts I've explained thoroughly on my channel. See this for example: ruclips.net/video/gNdEwr-sQg0/видео.html
Hi. Is it possible to make phone calls through other providers?
What kind of providers are you referring to? Do you mean apart from Twilio and Vonage?
Yeah, except for these.
@@blokmaster2952 Yes, but those are usually considered custom work. You can get directly onboarded to your own telephony server if you want to. Otherwise, if you want have more flexibility, you can look at building out your own solution using something like Vocode.
I have my own sip number and I want to use it. what are the solutions?
Is it possible to set up to capture emails correctly from the user and book an appointment?
It always captures wrong emails and then i can't send the invatation link to a meeting
Via phone it's super hard. We've tried many ways, but if you do it via voice, you always have plenty of room for error.
The best is to just create some custom logic that can send an SMS link and then use that one to request the email from the user, so they can simply fill it up during or after the call.
This would then require some more automation to delay the event (whatever that may be) until you got the email.
Doesnt Bland Ai do the same thing?
It's similar, but with tons of features build on top that aren't really necessary in many cases. I've worked with Bland in the past, but stopped it as they haven't done a great job with their version control and backward compatibility. I've had it many times that phone numbers just stopped working in the middle of campaigns because they pushed a new release.
Can it be connected to pineapple to store previous customer info and continue speaking with the customers based on past calls
Yes, that's all functionality you can build on top of their infrastructure. As for Pineapple - I haven't heard of that Tool/SaaS, but if it has an API, you can connect it too.
Thanks for this - I have been trying Bland and Retell, but am interested in trying this with some of my Make Scenarios. I have no issues when there is a webhook, but in my current Make scenario with Bland, it starts with a google sheet, it runs through the list of numbers, through JSON and at the end through an http module where it calls the Bland API which then triggers the phone call. Could something similar be done with VAPI ?
Indeed. If you can stay patient until monday, I'll release a video and share some one-of-a-kind make.com templates for Vapi that makes it insanely easy to build custom agents with their platform.
@@jannismoore Absolutely. Looking forward to it.
@jannismoore Awesome tutorial thanks! I just found out that each agent created on the dashboard in Vapi has a 10 call concurrency limit. In the way you are creating the agent on the fly from your server, does this invalidate this restrictions because it's creating a new agent on each phone call? Additionally, does this have any limits? Can I send 1000 calls in at a time and have it create 1000 agents?
Based on my knowledge, the 10 call concurrency limit counts for the standard pay-as-you-go plans. IF you need more than that, feel free to reach out to me via our agency site integraticus.com/contact
I will be able to help you get a limit that suits your needs.
@@jannismoore I was told on Vapi's discord that each assistant has that limit, not the account. So you get around it making 100 agents doing 10 calls each for 1000 calls, but that is a nightmare to manage. If this is true though, I wonder if your dynamic way would work to bypass it.
@@envolvetechnicalsolutions1315 There's nothing better than a limit test. :)
Appreciate you sharing this information.
@@jannismoore Please let me know if it works!
@@jannismoore On the API documentation on create assistant, in the json code to send the API it says "id string required - This is the unique identifier for the assistant." If this ID is required then how is this a transient/dynamic calling system? This ID field still limits the call rate to 10 concurrent calls.
how to set up the server url?
I've covered most of it in the video. If you want to read more about it, I suggest checking out this URL: docs.vapi.ai/server-url
Or: In case you have a specific questions, feel free to ask it right here. :)
Can you explain all the libraires we have to use to create similar vapi for example tts we can you deepgram and stt we can use playht but how can we create an api for this ??
If you want to build it yourself, you can look at something like Vocode, which is an Open Source framework that allows you to build conversational Voice Agents: vocode.dev/
I did a video on this earlier: ruclips.net/video/034Y4PoZCbU/видео.html
@@jannismoore but i want to bulid exact like vapi for example when user speak a question other side will be a ai model soo how can i bulid it i want it bulid it which library's i can use to bulid it
@@jannismoore i already bulid a model like it detect voice to text and that text give to model that resposne will convert into voice soo then we will here the voice but its too slow i want high speed work flow how can i bulid it
@@itzmeakash9695 Choosing the right server, optimize the streaming, and internal requests. Speed mostly comes from choosing the right infrastructure and the internal handling of functions. I sugget starting with revamping your code architecture by looking what functions or processes take most time.
@@jannismoore but how do i convert that into a api to integrate with my frontend?
new subsciber: Definitely NOT a "Newbie" ...
How much is it per min
It's around $0.10 - The price depends quite a bit on your selection of tools within Vapi, but once you create your own static assistant, you will see a price estimation directly within Vapi.
It has stopped working. 😢
Any alternative?😢😢😢
An alternative to the platform or an alternative to the problem that caused your assistant to stop working? :)
It would certainly be helpful if you can provide some more details about the actual issue you're facing.
What is with the zoomed in screen recording with motion blur on the cursor? Is this kind of editing supposed to increase youtube views at the low cost of making it unwatchable?
Just for you
I'm assuming it might be a side effect of an automated AI video editor? Not a criticism. Just curious. Great video!
@@stephenbollich Yep, it's a bad sideeffect of Screen Studio. :)
Hy, it's free ?
You get some credits upfront so technically it is free, at least in the beginning
vapirtisment(mix of vapi and advertisment) 😅😅😅
Haha - It’s literally the only voice AI platform I really enjoy using
why are you moving the screen so fast? It makes hard to watch
That changed already in future videos - it was a “feature” of the recording app
@@jannismoore I want to make a call, but answers should be taken feom my database. Still not clear do you show this or not :D Or how it will be done. Also what is that pink website where you put your webhooks. Not clear what you are doing))
I am from India you
Welcome!
Your country is name
I am youtuber
Interesting but too fast for Beginners...... thank you
I'm working on new content to make the entry easier
This looks commplicated AF
Wait until you see my other tutorials :)
moving with the zoom in is very annoying !
It is!
If anyone is advanced with Vapi, send me a message
sanity check. this...
{
"assistant": {
"firstMessage": "Hey Ryan, how are you?",
"model": {
"provider": "openai",
"model": "gpt-3.5-turbo",
"messages": [
{
"role": "system",
"content": "You're Ryan's assistant..."
}
]
}
}
}
will give you misery...
{
"message": [
"property assistant should not exist"
],
"error": "Bad Request",
"statusCode": 400
}
this...
{
"firstMessage": "Hey Ryan, how are you?",
"model": {
"provider": "openai",
"model": "gpt-3.5-turbo",
"messages": [
{
"role": "system",
"content": "You're Ryan's assistant..."
}
]
}
}
will bring you the 201 peace you seek. docs often lag behind the tech.
The assistant wrapper is mostly then relevant when you want to send it along in a request.
As for Vapi: I suggest following their Swagger API docs - those are the most accurate one at any given time.