11 ай бұрын
11 ай бұрын
Жыл бұрын
2 жыл бұрын
2 жыл бұрын
2 жыл бұрын
2 жыл бұрын
2 жыл бұрын
2 жыл бұрын
3 жыл бұрын
3 жыл бұрын
3 жыл бұрын
3 жыл бұрын
3 жыл бұрын
3 жыл бұрын
3 жыл бұрын
3 жыл бұрын
3 жыл бұрын
3 жыл бұрын
Пікірлер
@FaeLLe Ай бұрын
What about summarising chat contexts and using those summarises to reduce token size? Would it be safe to assume that because he Assistants API does not allow chat history to be controlled that we cannot achieve efficiency in this manner?
@kamaleshwaran06 Ай бұрын
By far, the best tutorial on Dapr and Azure Container Apps..awesome!!
@GeertBaeke Ай бұрын
Thank you for the kind words!
@vitusyu9583 2 ай бұрын
Thanks for your illustration, I really learnt a lot in it!
@jatinarorachannel8237 3 ай бұрын
I have a question i want to preprocess the text data and i have certain defined rules it should take care so should i use normal chat completion or assistant for that
@GeertBaeke 3 ай бұрын
It's best to use the chat completions API for that. In addition, if you want to extract structured data from text, I can recommend using the completions API with the Python instructor library.
@juanacquistapace9863 4 ай бұрын
Great video. Have a question. Even if I use completions or assistant api. The tokens for ALL the conversation Will be billed?
@GeertBaeke 3 ай бұрын
Yes, every input and output token is billed depending on the model chosen. That's the same for both APIs.
@SomeInfoSecDude 5 ай бұрын
Clearly explained ! All the other videos are filled with useless babbling and go directly in Yaml without context. Thanks.
@iGuide_net 5 ай бұрын
are you available on fiverr for custom work?
@GeertBaeke 5 ай бұрын
no, sadly not... 🙂
@iGuide_net 5 ай бұрын
clearly and simply explained A+ 🎉
@chevens 6 ай бұрын
Great video - very helpful! I'm intrigued to know if the CLU (conversational language understanding) fits into the assistant model, I can't piece together in my head how this could be utilized when looking at the playground UI.. I guess it's probably something you'd need to code for (fire the prompt to CLU to extract intent, then pass to assistants completion API?) appreciate your thoughts :)
@boris- 6 ай бұрын
Underrated by KZbin :) Thank you
@karishmasharma3203 6 ай бұрын
Apart from file_ids could we get data from other data source like SQL Server, BQ.... to assistants? Please could you do code video on it.
@GeertBaeke 6 ай бұрын
The only way to do that is via function calling and custom code that gets the data from SQL etc.. and feeds it to the assistant. There's no built-in support for that just yet.
@PandaUruguayo 6 ай бұрын
Amazing video!
@pallavkan 6 ай бұрын
its so difficult to find the repos which you have mentioned in all of your videos.. could have been more better if the repos were maintained properly..
@GeertBaeke 6 ай бұрын
It’s all in my free time and I try to do my best. These videos are old already so much of the code will not be very relevant anyway. Tnx for the tip…
@anuradhasharma2464 7 ай бұрын
superb explaination. Thanks alot
@gautam5692 7 ай бұрын
I was struggling with the service account concept and now that has gone, awesome content. thanks!!!
@hiandhra9971 7 ай бұрын
Hi can you show how we add two domains to the same origin group and two different routes
@AyushSharma-gs8sl 7 ай бұрын
great video
@jaiderariza1292 7 ай бұрын
hi, what tool do you use to create the diagram?, can you share the file? thanks.
@luisfelipebaptista 7 ай бұрын
Thank you for the demo! Really appreciate it. Do you know if there's a way to integrate an APIM with this kind of setup? Would the APIM go in between the load balancer and the cluster, or in front of the load balancer?
@GeertBaeke 7 ай бұрын
You can put APIM in front of the load balancer but APIM does not support private link to connect to backends. You can deploy premium or the new standard v2 which allows connection to internal backends.
@Ajmal_Yazdani 8 ай бұрын
It's fantastic to see you back on KZbin after a long time. Please continue sharing, and I truly appreciate your blog and videos.
@GeertBaeke 8 ай бұрын
Thanks for the kind words. More to come soon I hope…
@seeyditti 8 ай бұрын
What is the point of all this...
@GeertBaeke 8 ай бұрын
I am in no position to answer existential questions.
@glibmar 8 ай бұрын
Very nice. Thanks for sharing ❤
@TANVEER991164 8 ай бұрын
clear and concise , thank you :)
@shantipahari 9 ай бұрын
you didn't talk about the helm charts?
@AdnanKhan01 9 ай бұрын
You made this look so simple!! greatly appreciate it!! Thank you
@davidpratr 9 ай бұрын
Hi, I was thinking about the blue/green and canary with Argo rollouts and they are an effective way to test being able to quickly rollback in case of issue. But what happens if one wants to just test online but exclusively in a private way? That would be a staging environment. An option could be to just have a secondary domain and deploy the website in that secondary domain temporarily and perhaps with a modified react component or any equivalent layer only giving access to one staging user? How do big companies make that staging environment to be possible to be online but without having it exposed to public in production? In data analytics applications it's easier because usually all is in an intranet but what happens with webs that are usually B2C applications?.
@egger6311 9 ай бұрын
Thanks for the clear explanation and the architecture drawing. Helped me to finally find the mapping issue in my code. Cheers and have a nice day😊
@GeertBaeke 9 ай бұрын
I am glad it was helpful!
@davidpratr 10 ай бұрын
I was wondering if an alternative to using Argo rollouts would be to use Istio. It would add more complexity to the deployment in general but Istio already has the implementation for blue/green and canary that kubernetes basic networking lacks.
@GeertBaeke 10 ай бұрын
Argo Rollouts works together with Istio and other service meshes to manipulate the traffic as needed. Although you could use Istio on its own and configure it all yourself, using it in combination with Argo Rollouts or Flagger is easier.
@davidpratr 10 ай бұрын
Hi@@GeertBaeke , I thought if one is using Argo rollouts or Flagger, and then one had it configure it with Istio, the blue/green and canary features of Istio are not needed any more, one is only going to use the service mesh part (and others) of Istio.
@GeertBaeke 10 ай бұрын
@@davidpratrThat was indeed what I was pointing at. Flagger and Rollouts using the service mesh features to orchestrate the canary or blue/green deployment. I must admit that I do not use Istio that often and tend to use Linkerd. So I am not that familiar with Istio's standalone features. With Linkerd, you have to use it with Rollouts or Flagger because it has no such standalone functions. Last time I checked 🙂
@davidpratr 10 ай бұрын
@@GeertBaeke I see Flagger is more related to Flux. Well, I think it's clear that if one starts a project from scratch and one knows one will use ArgoCD, Argo rollouts is the way to go. Another thing would be if the project was already started with it's Blue/Green or canary with Istio, then it could be more challenging to adapt the whole thing to make the rollouts with both things Argo rollouts and the Istio rollouts capability.
@tzahik1 10 ай бұрын
Great video! Thank you very much! 🎓🥇🥉🥈🏆
@TomerShaiman 10 ай бұрын
Great Tutorial. when trying this is SSL Cert on the server side ( and of course the right configuration as you shown on the last part of the video , aiming for Https forward route , and even disabling health checks) , it seems the the command is returning <Service Not Available> [even though its there] . would be interesting to see how this is configured with HTTPS and SSL certificates
@josejayesh 10 ай бұрын
Great video!. Thank you!
@jayasuriyap8748 11 ай бұрын
I.e mean in the secrets
@jayasuriyap8748 11 ай бұрын
Azure credentials what are the things you have given
@GeertBaeke 11 ай бұрын
Can you clarify somewhat more? It's been a while since I made this video...
@jayasuriyap8748 11 ай бұрын
@@GeertBaeke in the secrets in the GitHub side what we need to give what are the things I am asking . Workflows getting failed.
@Ajmal_Yazdani 11 ай бұрын
Thanks and appreciate 🎉🎉
@GeertBaeke 11 ай бұрын
You’re welcome 😊
@gadwal 11 ай бұрын
Good one Geert. Only vid on internet that deals with home automn and openai API with fn calling. Thanks!
@GeertBaeke 11 ай бұрын
Thank you and glad you liked it.
@MrTrigsy1 Жыл бұрын
Thankyou Geert, found this extemely useful 🙂
@lyubomyrvyhovskyy2668 Жыл бұрын
Thanks for the video. Where can I find that online tool to generate patch from base YAMLs?
@yuriykutsiy3781 Жыл бұрын
Nice demo)
@guillermoquiros2402 Жыл бұрын
This is really well done!!
@sfincione2000 Жыл бұрын
FYI: Canary is pronounced "can-ay-ry". I hear it a lot from non-native English speakers. Hopefully it doesn't annoy you to mention this. Other than that, thanks for taking the time to give the world a very nice overview. Nice job.
@GeertBaeke Жыл бұрын
Oh no not at all. The problem will be to remember it for next time. 😀
@amjads8971 Жыл бұрын
This is great
@johnraj5274 Жыл бұрын
Why we need redis in this setup what is the purpose
@GeertBaeke Жыл бұрын
It’s not required for Flux. It’s just an example of installing some infra components an app might need. Could be anything really…
@shanmugasundaram8828 Жыл бұрын
As I want to add the service to configure the path for downstream request using kong ingress controller. I followed this approach but in the kong documentation says it is deprecated . can you please help in writing kind: Service type of resource in helm chart
@marcusmartinsleite1891 Жыл бұрын
Hello! Excelent explain. I Have a question about environment. How to create a way to not depend on the number of possible IP's in the VNET. Imagine, in Kubernetes we have our own technology taking care of the IP's, I understand that each app in the ACA takes an IP that I have available in the subnets. If my set of APP's grows I can reach a point of not having more IP's available
@GeertBaeke Жыл бұрын
With a VNET-integrated Container Apps environment you will have to take the number of available IP addresses into account and make the subnet large enough. That is because each replica for a container app in the consumption plan takes up an IP address. There is of course a maximum number of replicas per environment so you can take that into account for the subnet size. However, when you use workload profiles, there's one IP per host not per container app. Of course, in that case, you do not use the consumption plan and you need to provision hosts to run your apps. In larger environments, we see customers having a benefit from using workload profiles so in that case, the IP limits are not at play.
@jogurt4605 Жыл бұрын
Awesome demo, thank you!
@GeertBaeke Жыл бұрын
Thank you! Much appreciated!
@Devopsforyou Жыл бұрын
Geert You are rock star in explaining with easy examples!!!
@vikashplayer Жыл бұрын
Hi Geert, Thanks for this demo. Can we have Application gateway (Not AGIC) instead of Azure Front Door? Any demo will be highly appreciable.
@amjads8971 Жыл бұрын
You can have it, but i had tons of issues implementing it in a hub n spoke model. Front door is good if you have client residing in diff regions accessing your application. Otherwise go for appGateway
@wesleygyger1210 Жыл бұрын
Very helpful, thank you
@cipcip4986 Жыл бұрын
Thanks for your content. Do you have a git repo with the codebase?