Questions (and answers) from the Genkit livestream on June 17th 2024 #424
Unanswered
peterfriese
asked this question in
Q&A
Replies: 2 comments
-
Thanks a lot! ️:) Very interesting presentation. Are the examples presented available on a repository somewhere? |
Beta Was this translation helpful? Give feedback.
0 replies
-
thank you for the presentation super excited about genkit 🔥 |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hey everyone,
we had a Genkit livestream on YouTube today (see the recording).
Pavel joined Nohe and me to give an overview of Genkit and dive into some topics, such as RAG, tracing and monitoring, evaluations, and more.
Thanks to the audience for sending questions! I will post a list of questions that were asked below, including the answers we gave.
Q: Hey team! Tomas here, a pleasure as always 🙂 My question is: Do you plan to support text-to-speech and/or speech-to-text in Genkit eventually?
A: Genkit has an open plugin system, and while we don't have any plugins for TTS or STT at the moment, you can build them yourself, and share them with the community using the genkit-plugin hashtag on GitHub. Check out
Q: How does Genkit integrate with other Firebase products? Example: save a generated image from Imagen to Firebase Storage
A: Pavel showed how to implement a Genkit flow that generates a restaurant description based on a theme, then generates an image of the restaurant interior, and finally stores the image in Cloud Firestore (but you could just as easily store the image in Cloud Storage). This starts at about 17:24 into the stream.
Q: What's the difference between VertexAI and GetKit? How do I decide which one to choose?
A: VertexAI is a unified machine learning platform that allows you to build, deploy, and manage machine learning models and AI applications at scale on Google Cloud. There also are the Vertex AI SDKs for Firebase, which allow you to call Gemini from your client app. In contrast, Genkit is a framework that you can deploy to any platform that can serve an Express.js app. It is an AI integration framework that makes it easy for developers to build AI applications that connect with any APIs they might already use in their (backend) applications.
Q: How to chain flows?
A: You can invoke other flows with the
runFlow
function (see the documentation)Q: Can we use our own APIs?
A: Yes, you can call any APIs that you would also call from inside an Express.js application. These can either be local functions, or APIs that you call remotely.
Q: Can we use Gemini in our backend?
A: Yes, Gemini is supported out of the box, and the community has implemented plugins for other LLMs as well - see this community repository
Q: What about error handling inside flow if one goes wrong?
A: You can use exception handling, like you would in any other Express.js application. If you don't handle an error, it will be thrown to the caller.
Q: What's the difference between
defineFlow
andonFlow
?A:
onFlow
flows are exposed as Firebase callable functions and can be called from your client application (e.g. from your Android, iOS, or web app) using Firebase's Cloud Functions SDK.Q: What happens if you add Genkit to an existing Firebase project - will it overwrite the existing Cloud Functions?
A: This would only happen if the existing Cloud Functions have exactly the same name as your Genkit flows. The Firebase CLI will ask you if you'd like to overwrite any Cloud Functions.
Unanswered questions:
Q: Does Firebase provide us pre-built APIs for data sending and receiving?
A: (please provide an answer)
Q: Can we add Genkit in flutterflow?
A: (please provide an answer)
Q: In development it shows used token correctly not in production why?
A: (please provide an answer)
Q: What would be a proper way to test your flows? I guess we don't really test the LLM response since that's the model, but the actual logic of the flow, do you have tips on how to implement tests?
A: (please provide an answer)
Beta Was this translation helpful? Give feedback.
All reactions