8 SECRET Flowise AI Features You DIDN'T Know Existed (Supercharge Your Workflow!)

8.92k views1411 WordsCopy TextShare

Leon van Zyl

In this Flowise AI video we will have a look at 8 Secret Features in Flowise that will greatly impro...

Video Transcript:

In this video, I will show you 8 hidden features of Flowwise that will greatly improve your chat flows. For instance, did you know that you can allow users to upload images by enabling this "Allow images upload" toggle on some of the chat models? This will allow us to upload an image to the chat by clicking on this "Add image" button.

And for this example, let's upload this menu for this fake restaurant and let's ask the model what the name of the restaurant is. So let's enter what is the name of the restaurant and we do get the correct response back. Let's try another example.

Let's upload this meme and let's say "explain the meme" and this explanation is 100% correct that you know that you can enable speech to text recognition as well. This will enable the user to ask questions using their voice. We can enable this by going to settings, configuration, we can click on the speech to text menu.

Under providers, let's select open AI, select your open AI credentials. Optionally, we can set the language which will help improve the accuracy of the model. We can also set the prompt as well as the temperature.

I'll simply leave it as is. Let's click on save and now we can chat to the model using our voice by clicking on this microphone button. Tell me a joke about horses.

And first we can see that it's converted my voice into text and the model is responding. Nice. Another useful feature for larger chat flows is sticky notes.

We can add sticky notes to our canvas by clicking on add notes, utilities and then sticky note. We can move the sticky notes around on the canvas and we can then leave notes for ourselves. So let's say we wanted to replace this buffer memory note with a redus note instead.

We can simply change the text and leave a note for ourselves. These can be extremely useful for keeping track of certain decisions or to-dos in the chat flow. Another useful feature is starter prompts.

These prompts will show up just above the chat box when a user starts a new conversation. And this can be very useful for getting the conversation started. We can add starter prompts for going to settings, configuration and we can click on the starter prompts menu.

From here we can add multiple starter prompts. For instance, schedule an appointment. Let's add get my order status and maybe what services do you offer?

It's saved as it's close to pop up. And now when we open the chat, we can see those initial prompts showing up. Another user has the option of typing a message or simply clicking on one of these to get started.

Next, I'm going to show you a few more important features like getting feedback from the user during the conversation as well as preventing duplicates in our vector stores. But first, if you like this video, then please hit the like button and subscribe to my channel. Also, let me know down in the comments if you like this type of video where I show you quick tips about flow wise.

And also let me know if I missed any features in the comments below. The following two features are extremely important for protecting your chat flows in production. The first is rate limiting.

We can use rate limiting to prevent users from spamming our chatbot by sending multiple messages and driving up your costs. We can set rate limiting by going to settings, configuration, and then go to the rate limiting tab. From here, you can set how many messages should be allowed within a certain period of time.

For example, let's say that we only allowed 10 messages within a one minute period. If we exceed this limit, we will show the following message to the user like, "Sly your horses, try again in a few seconds. " Another great option of preventing people from embedding your chat flow into their websites is to set an allowed domain.

You can access that from the settings by going to allowed domains. And as per this description, your chatbot will only work when used from the following domains. So here, you can simply enter the name of your website and this chatbot will only work from the specific domain.

Another incredibly useful feature is chat feedback. When users send messages to your chatbot, these conversations are recorded and you can view them by going to settings and view messages. View messages is extremely helpful for figuring out what's going well and what's not working with your chat flows.

And it's very useful for priming your models. But did you know that we can also receive feedback from the users, which will further help us in improving our chat flows? On the screen, we can see these blocks for the total amount of feedback received as well as a positive feedback percentage.

But how does this actually work? This is extremely easy. We can enable chat feedback by going to settings, configuration, then click on chat feedback, enable this, save.

Now let's have a look at our chat window. We now get these buttons to like or dislike this response. Let's go ahead and like this response.

We can also leave feedback like this was cool. Let's submit this. Let's continue the conversation.

What is your name? Let's actually dislike this comment. Let's say I don't like the name.

Let's submit this. Now let's go back to our view messages. And now we can see that we receive feedback from our users with a 50% positive feedback rate.

And in the conversation, we can also see that the user liked this message as well as the feedback provided. And we can also see that dislike with the reasons for the dislike. This is so useful for improving our chatbots.

Now lastly is a very important feature that some of you have been asking for for quite a while. Here I have a rack chatbot that upsets data into a pinecone database. And if I go to the pinecone database, I can see my one record over here.

Now some of you might be familiar with this very common issue. When we change data in our source file and we click on absurd again, this is simply going to create a duplicate record in the vector database instead of deleting the old record. Now this is slightly more technical to set up, but I do think it's worth mentioning.

In flow wise, we have something called the record manager. And we can see that this keeps tracks of records to prevent duplication. So what we can do is add a record manager to this flow and that will ensure that records are not duplicated.

And when we make changes to our source data, only the updates are inserted. So we can simply click on add notes, then go to record manager. And at the moment, we have the options of adding a MySQL record manager, Postgres or SQLite.

So if you do have access to one of these databases, you can simply add it to your canvas, then connect it to this pinecone node and this will help ensure that we do not insert duplicates. Setting up these databases can take some time. So let me know in the comments if you would like to see a dedicated video on record manager.

But what I do want to show you is if you click on additional parameters, we can select from different cleanup options. And these options are very well documented in the flow wise documentation. So we will use incremental when you want to upset multiple documents without deleting any of the existing documents that is not part of the current upsetting process.

With full mode, when you upset multiple documents, it will automatically delete any vector embeddings that are not part of the current upsetting process. Or you can just leave it on none, which will not perform any cleanup process. This video was a lot of fun to make and it is different to my usual format.

Please let me know in the comments if you would like to see more of these quick tip flow wise videos. If you like this video, then please hit the like button and subscribe to my channel.