Power of openAI embedding in Chatbot

Power of openAI embedding in Chatbot

Partager cet article

#1 Power of  OpenAI

🔥Hot topic alert🔥

Everyone is talking about OpenAI lately!

From tech geeks to digital marketers! 🤖

Everyone is raving about the endless possibilities of this cutting-edge technology.

You can ask open AI to write converting copies, make marketing plans, write Facebook Ads, reply to emails.....

With OpenAI, you can do all sorts of cool and interesting things to make your business better! 💪

But.....

OpenAI SUCK when you want to use for your own business!

But wait..... There is a catch.

OpenAI may be powerful 💪, but it can be powerless when you want to use it for real business!

Think of OpenAI like a personal chef. 🍳

It can cook up amazing dishes:

- 🍔 Burger 🔥
- 🍕 Pizza 🍕
- 🍣 Sushi 🍱
- 🍗 Chicken Wings 🍗
- 🍝 Pasta 🍝
- 🥗 Salad 🥗
- 🍰 Desserts 🎂

Because OpenAI is learning huge amount of data from "internet".

But your restaurant, your business only servers "🍜 "Rich & Savory Ramen Bowl" 🍜 with your unique recipe.

OpenAI can't find any information about that from internet at all.

And even if it can find the data, it's not accurate, because your business have a secret ingredient 😇 that nobody knows about.

Then OpenAI is useless for your business. 😅

It's cool that OpenAI can do all those amazing stuff, but you want  OpenAI do cool stuff for "your own business"

You don't want OpenAI to ask the "internet" about anything relevant to your business.

You want OpenAI to learn and reply from the data you provided about your business.

So that it's accurate & relevant & smart.

Well, you can do it. UChat has the solution for this!

Introducing "OpenAI Embeddings"!🎉

Introducing OpenAI Embedding

What's OpenAI embedding? 🤖

I will make it easier for you to understand 😉

So you have this cool kid who knows everything(almost).

Because this kid has gone through a lot of learning & training.

Millions & billion dollars spent during this process. 💸💸

So this kid is smart 👨💻

(well, who wouldn't? with that kinda of money spent)

He can easily understand what your customer is asking.

And you hired this cool kid to work for your customer support.

You give this kid guidelines on his first day of work, and tell him,

"I know you are smart, knows what our customers are asking, but....

reply to our customer only from these guidelines,

well, you can rephrase it, change it, but should all based on these reference replies.

don't make things up, don't be a smart ass about the replies

If there is something you are not sure of, that's ok.

you can give them the most relevant & possible answers from these guidelines.

P.S: if they are asking reservation, make sure you grab their details!

So in this story, the cool kid is OpenAI(you should know it by now)

The guidelines you provided for him to reference on is called "OpenAI Embedding".

With this feature, you will be able to do below #3 important things:

#1 Generate reply from the most relevant, accurate database, don't make things up

#2 Provide most relevant, related answers if you are not sure of(Yes! you can "know" how confident OpenAI is about the matching score)

#3 Recognise the intent, provide call to action to the next step(If the user want to book reservation, you don't want to just tell them how. you want to ask: do you want to book now?" 😉)

With all the possibilities that OpenAI embedding brings, you can really harness the power of OpenAI, and use it in your own business! 🧞♂️

You might ask: "how to do all of this?"

Lucky for you. 😊

UChat is the first chatbot platform to bring this feature to the chatbot market!

OpenAI Embedding in UChat

UChat has this OpenAI embedding built right in. 👍

It's the first chatbot platform to bring this feature! 🤖

🥳 🥳 🤩 🚀

#1 Train OpenAI with your database

So the first thing we need to do is to give this "cool kid" the guideline he need to remember & stick to.

You can find the OpenAI in the "integration" tab, and update your own embedding.

You can either add it manually or import as CSV file.

After you upload the content, click "Regenerate"

That will make sure this "cool kid" remembers & memorises all the data in his smart brains 🧠... in seconds ⚡

#2 Trigger OpenAI action - Let the "cool kid" think about it

I am not going into too much detail(which is going to cover in lesson #2, #3)

But you can easily find OpenAI action in the action -> Integrations, it will look like this:

The input field will be whatever the user is asking, and if you are using our built-in system field, it will make things much easier.

Using the "Embedding Match" action of OpenAI, will let the "cool kid" to understand the user's question, and think hard from the guidelines he just learned.....

#3 Did openAI find any high relevant match?

After you run this action, OpenAI will pop up one most relevant entry from your guidelines - embeddings!

And it will have a score for you to measure relevance.

Generally speaking, if more than 0.8, it will be high relevant.

So, you can reply back to the user from this text or provide this text as the context for OpenAI to generate new response.

You will need to use the "Embedding Match and Completion" action to do so.

#4 What if none of the results are relevant?

It's possible that the "cool Kid" doesn't find anything that is relevant to the user's question.

Even in that case, we don't want to OpenAI to ask "internet" for a reply.

All the relevant data is saved in this "item" field.

You can actually display all these relevant reply in cards to your customer for reference.

#5 How to trigger Call to Action/Follow up intent?

There is one very important things that your support team needs to do.

You don't want to stop there when the user wants help or book a reservation, etc

You want to move along the conversation.

After the OpenAI detected the "intent" and replied according to the guidelines.

You might want to offer different ways to contact you when the user is looking for help.

And, you want to ask "do you want to book an appointment today?" when the user is asking "how can we make a reservation?"

That's very easy to do in the UChat OpenAI integration.

Simply, Save the OpenAI recognised intent, and then followed by a condition check, and redirect the user to the relevant flow.

#2 Get started in 5 mins

Hey there! 🤗

Excited to build your own restaurant demo in the previous video(quick demo here) with UChat in 5 mins? 💻🤖

That's true! We get everything ready for you! 😇

It will be a smooth sailing 🚤 ⛵, I promise

Just to make sure you're ready, here's what you'll need:

- Download the restaurant Open AI embedding 🍔🍟
-Open AI API Key 🔑

You can download the open AI embedding here, and make the personalised changes if you want.

The chatbot will reply to your customers using the data from this embedding! 🙃

Your chabot will be accurate, relevant & smart.

Easy deployment + Smart chatbot = 🤑 Happy customer + easy money to make! 🤑

You will also need your own open API key, you can start the free account here 👈

And the best part?

You can get a 14-day free trial of UChat, no credit card required and access to all the pro features! 🎉

Ready to rock and roll?

Let's get started! Click the "get started" button below and let's create your UChat account.

After collecting your name & email, we will create new account with restaurant open AI template installed.

Now you can click the button below to get started.


#3 Template Walkthrough

Great work so far! 😋

You should already have your UChat free trial workspace with template installed! 🥳

Your openAI should also ready to work(just put your API key in the openAI integration)

You should already feed & customerize the embedding data to the openAI(do remember to click "regenerate" button to activate it)

So in this lesson, we will walkthrough how to do the below 3 important feature:

1️⃣ How to use openAI to reply from your own database

2️⃣ How to decide if the reply is good enough or not? What to do if openAI can't find good match from your database

3️⃣ How to ask follow up questions

Ready?

Let's get started.

OpenAI integration in UChat

First of all, let's find openAI integration in UChat

In the flow builder, you can right click anywhere, and then find the action -> integration -> openAI action

The actions UChat support:

1) Create Text Completion: This can be used for openAI to reply any open question.

2) Image Generation: Yes, you can use openAI to generate dynamic images as well

3) Embedding Match: You will be able to use this action to let openAI looks through your database, and find the best match.

4) Embedding Match and completion: So you can use the best match as the context in the prompt, and it will provide the necessary background information for openAI to generate a dynamic & accurate response.

That's all the integration we have.

Now, Let's get openAI to work. 👨💻

How to use openAI to reply from your own database

Let's first create a simple question node to take the user's input/questions.

Whatever the user types, it will save into {{question}} user custom field.

We will use that for the openAI input/question.

Now, let's create an openAI action "Embedding Match"

Only this action,openAI will actually look our database(embeddings) to find the relevant answer!

You can see that we are providing {{question}} to the openAI.

Also, the left side, you can provide the test data to test the response.

It's pretty simple & straight forward.

Now, how do we make use of the response returned?

Keep reading.....

How to decide if the reply is good enough or not?

Let's take a look of a typical openAI embedding match response below 👇

There will be always a highest matching/ranking results, with a score.

That's is best match or relevant results(what openAI thinks).

The score is how you can decide if the reply is good to use or not good to use.

Generally speaking, if the score is >0.8, it's high relevant, you can use the text directly.

Also, you can use this text(which is highly relevant) as the prompt in the "Embedding Match and Completion" action, and make it more powerful.

If the score is <0.8, then openAI can find results from the embedding, just all results is not very high relevant.

All those relevant data will be saved in that {{item}} JSON.

You can save & display these data in card in UChat, or

You can offer them extra ways to get help for you.

No matter which way you choose,

You already "harness" the power of openAI to use for your business.

You should really proud of yourself by now. 👍

How to ask follow up questions?

This is the last but important feature we need to cover.

In most cases, you don't simply want to use openAI to reply the question.

You want to understand the user's intent, and then push for the next step, add call to action based on the intent.

For example, if the user is looking for "how to make reservation in your restaurant".

You don't want to end the conversation by tell them how. 😓

You want to ask immediately the question: "Do you want to book a reservation now?"

So, in this typical openAI embedding match response.

You can save the best match "heading" into a user custom field.

And then after the openAI response, you can use a simple condition to check if the user's intent is "book a reservation"

If it is, and then you can redirect the user to the next step.

That's how easy it is to move the conversation along with the user.

That's it.

You already quickly learned 3 important user case of openAI embedding inside of UChat.

Below is a very detailed, long template walkthrough video, if you need to deep dive into UChat openAI embedding feature.

What's Next

So, what else can we use openAI embedding in UChat?

Well,

How about let the "cool kid" who understand your business reply to your Facebook post comments, Instagram post comment, Instagram Story mentions, Instagram Story replies, Instagram Live comments?

That's right!

You can set up automation easily in UChat to reply to those messages....

It will be highly relevant to your business.

It will be accurate & smart.

Imaging, you are running ads to an ecommerce products.

The post go viral, there are a lot of comments asking for a lot of questions.

You don't have enough time or team member to handle all those comments.

It's like leaving money on the table!

But if you set up openAI embedding chatbot 🤖

It will work around the clock to reply to the comments, and it's replying according to the guidelines(embedding) you set up.

You can have a good rest, and relax, and watch the order rolling in. 🤑🤑

We will bring the tutorial to do this very soon.

Make sure you join our Facebook Group, Subscribe to our Youtube Channel!

We will post tutorial how to do it step by step in our social channels!

Join UChat Facebook Group

Want to join a community of chatbot enthusiasts and experts?🔥🤖

Our Facebook group has 3k members who love to share chatbot tips, updates, and tutorials!

🤗🤖 See you in the group! 🔥

Just smack the link here to join our Facebook Group 🙂

Subscribe to UChat Youtube Channel

If you want to stay up-to-date on all the latest chatbot related content,Then you won't want to miss our YouTube channel! 💻🤖

We post chatbot relevant content regularly to keep you informed and entertained! 🤗👨💻

So hit that subscribe button and join us on this chatbot journey! 🎥🤖

Just smack the link here to subscribe to our Youtube Channel 🙂

Partager cet article

Inscrivez-vous à notre newsletter