Close

Sign up to Gadget

How to: Build a chatbot that generates custom movie scenes using OpenAI

-

About

Use Gadget's built-in OpenAI connection and AI tooling to build a chatbot that helps you inject custom quotes into movies.

Problem

Solution

Result

How to: Build a chatbot that generates custom movie scenes using OpenAI

Riley Draward
August 24, 2023

Learn about Gadget's built-in AI features such as the OpenAI connection, vector databases, and cosine similarity search, and use them to build a chatbot that generates custom movie scenes.

You can fork this Gadget project and try it out yourself. After forking, you will still need to run the ingestData global action to ingest some sample data and generate embeddings for the movie quotes. This is detailed in Step 2: Data ingestion.

Fork on Gadget

Create a new Gadget app

Before we get started we need to create a new Gadget app. We can do this at gadget.new. When selecting an app template, make sure you select the AI app template.

A screenshot of the template prompt when creating a new Gadget app, with the AI app template selected

Now that we have a new Gadget app, let's start building!

Step 1: Create a movie model

The first thing we need to do is store some movie quotes in our Gadget app. We're going to make use of Gadget's data models, which are similar to tables in a Postgres database, to store this information. To fetch data, we will make use of a global action.

Start by creating a new model in Gadget:

  • Click the + button in the DATA MODELS section of the sidebar
  • Enter movie as the model's API identifier

Now add some fields to your model. Fields are similar to columns in a database table, and allow you to define what kind of data is stored in your model. For our movie model, we'll add the following fields:

  • Click + in the movie model's FIELDS section
  • Enter title as the field's API identifier
  • Click on the + Add Validations drop-down and select Required to make the title field Required

Adding a Required validation to title means that an error will be thrown if a movie is added without a title. Now let's add a field to store the movie's quotes:

  • Click + in the movie model's FIELDS section
  • Enter quote as the field's API identifier
  • Click on the + Add Validations drop-down and select Required to make the quote field Required

Now we have a place to store movie quotes! This is all the raw required for our app, but we also need a field used to store vector embeddings. Vector embeddings are a way of representing text as a vector of numbers. Gadget has built-in methods to perform similarity searches using vector embeddings - for this app, we'll use these vectors to find similar movie quotes when compared to text entered by users. To learn more about vector embeddings, check out our docs on building AI apps.

  • Click + in the movie model's FIELDS section
  • Enter embedding as the field's API identifier
  • Select vector as the field's type
A screenshot of the completed movie model, with title, quote and embedding fields added. The embedding vector field is selected.

That is all that we need to store data for our app! Now we need a way to generate embeddings. Luckily, OpenAI has an API that we can use to pass in text and get back a vector embedding.

Step 2: Data ingestion

We need some test data for our app. We're going to make a request to an open data model hosted on Hugging Face. Hugging Face is an AI community and a great resource for AI models, datasets, and great community-provided AI content. We will then use the OpenAI connection to generate embeddings for our movie quotes.

To ingest data into our app, we can make use of a Gadget global action. Unlike model actions, such as the default CRUD actions generated for our movie model, global actions serve as ways to apply updates to more than a single record.

  • Click on Global Actions in the sidebar
  • Click the + Add Action button or + next to the ACTIONS section title to create a new global action
  • Name the action's API Identifier to ingestData

Our OpenAI connection is already set up for us using Gadget-managed credentials. Each team gets $50 in OpenAI credits to use when building in development. To learn more about how to set up your own OpenAI connection, check out our OpenAI connection docs.

With our OpenAI client set up and ready to be used, we can add some code to create embeddings for our movie quotes when they are added to our Gadget database.

  1. Enter the following code in the generated code file (replace the entire file):

This code:

  • uses fetch to pull in a small sample dataset that stores movie quotes hosted on Hugging Face
  • loops through the returned data and creates a new movie record for each movie quotes
  • uses the OpenAI connection to generate embeddings for each movie quote
  • uses your Gadget app's internal API to bulk create movie records with the generated embeddings

Now we can run our global action to ingest the data:

  • Click on the Run Action button to open your action in the API Playground
  • Run the action

The action will be run and we see a success message once data has been pulled in from the dataset. We can also see the data in our Gadget database by:

  • Clicking on the movie model in the sidebar
  • Clicking on Data

You should see movie records, complete with title, quote, and embedding data!

A screenshot of the movie model's data page, with ingested data and embeddings visible

Now that we have data in our database, we are ready to build the user-facing portion of our app.

Step 3: Use a global action to find similar movie quotes

Our app will allow users to enter some text and will find movie quotes that are similar to the entered text using a similarity search on the embeddings. We will use a global action to find the top 4 most similar movie quotes, and then present these movies to the user.

We can create a new global action that we will call from our frontend:

  • Click on Global Actions in the sidebar
  • Click the + next to the ACTIONS section title to create a new global action
  • Name action's API Identifier to findSimilarMovies
  • Enter the following code in the generated code file (replace the entire file):

This code creates an embedding from the text passed in by the user and then finds the 4 most similar movies to the entered text. We then remove any duplicate movies and add the movies to the global action's response.

In our global action, we define a custom param at the bottom of the code snippet so the user's quote can be passed in as a string through the params object.


To return values from a global action in Gadget, we need to set the scope.result variable to the value we want to return. In this case, we want to return the movies that are similar to the user's entered text:


Finding similar vectors with cosine similarity

This piece of code, included in the snippet above, is the key to finding similar movies:


Gadget has built-in vector distance sorting which we use to get the most similar vectors to the user's entered text. We use the cosineSimilarityTo operator to find the cosine similarity between the user's entered text and the movie quotes in our database.

Build a UI to display the results

Now let's start building our frontend to call this global action and display the recommended movies to the user.

We're going to use the Chakra UI library to style our frontend. We can install it by running the following commands:

  • Open the Gadget command palette with P or Ctrl P
  • Type > to enable terminal mode
  • Run the following command:

Once installation is complete, we can start building our UI:

  • Open the frontend/main.jsx file
  • Import the ChakraProvider component at the top of the file:

  • Wrap the <App /> component with the ChakraProvider:

  • Replace the contents of frontend/routes/index.jsx with the following code (replace the entire file):

Most of this code is just normal React! The only Gadget-specific code is the useGlobalAction hook, which we use to call our global action. Let's take a closer look at how it works:


We provide our findSimilarMovies global action as input to the useGlobalAction hook. The imported api object, which is your automatically-created Gadget app's API client, is how we can access our Gadget app API from the frontend.

The useGlobalAction hook returns a response object containing:

  • data - the data returned from running the action,
  • fetching - a boolean that is true when the action is running
  • error - an object that details any errors that occurred while running the action

These objects can be used to display returned data, a loading spinner, error banners or messages, when building your UI.

The findSimilar function returned by the hook is what we call to run the action. We pass in the quote as input to the action, and the data returned from the useGlobalAction hook is an array of movies with similar quotes.


Update permissions

By default, newly added actions are not callable by just anyone. We need to grant unauthenticated users permission to call findSimilarMovies.

  • Click on Settings in the left sidebar
  • Click on Roles & Permissions
  • Check the box next to findSimilarMovies under the unauthenticated role
A screenshot of the permissions page, with the Unauthenticated permission for the findSimilarMovies global action checked and highlighted

You should now be able to enter a quote and see similar movies! Let's test it out.

Test out the frontend

To view our development app:

  • Click on the app name at the top of the left sidebar
  • Hover over Go to app and click Development

Try it out! Enter a (fake) movie quote and movies with 'similar' quotes (as decided by our vector similarity search) should be returned.

We're at the last step: let's use OpenAI to generate fake movie scenes using our quote and the selected movie. We can also add some chatbot capabilities to our app so that we can interact with the generated scene.

Step 4: Add a route to generate a scene

Now for the final step in our development process: adding an HTTP route to our Gadget app that will be called by the frontend to generate a scene. We'll use the Vercel AI SDK to make it easy to handle the array of chat responses in our frontend.

Why not use a global action?

We previously used a global action to find similar movies, but we're using a route to generate a scene. You might be asking yourself why? There are two main reasons:

- Global actions do not support streaming responses, and we want to stream the text returned from OpenAI to the frontend
- The openAIResponseStream helper we are using integrates seamlessly with HTTP routes

In general, we suggest you use global actions over HTTP routes whenever possible. But when streaming or integrating with external systems or packages, HTTP routes can be a better choice. To read more about when to use each, see the Actions guide.
  • Start by modifying the routes/POST-chat.js HTTP route file in your Gadget app (replace the entire file):

This modifies the sample POST route for our app at /chat.

A custom queryparam is defined in the schema object. This queryparam is used to pass the messages array to the route. The messages array contains the prompt and the generated scene.

We reply using Gadget's openAiResponseStream helper which assists with streaming responses that are returned from the OpenAI connection.

Update the UI to add chatbot functionality

Now that we have defined our route, we can add support to call this route from the frontend. First, install the Vercel AI SDK npm package:

  • Open the Gadget command palette with P or Ctrl P
  • Type > to enable terminal mode
  • Run the following command:

  • Once again, paste the following code into frontend/routes/index.jsx:

The useChat hook is imported from Vercel's AI SDK (ai/react) and sends a request to our /chat route when handleSubmit is called.


The useChat hook also provides us with:

  • a messages array which we can use to display the chat history
  • a setMessages function which we can use to send messages to the chatbot
  • a setInput function we use to seed the initial prompt passed into our route

For more details on Vercel's AI SDK, check out the documentation.

None of this added code is Gadget-specific! It is, once again, plain old React, making use of different libraries and packages (Chakra UI and Vercel's AI SDK) to build a simple chatbot.

Test your chatbot

We are done building! Let's test out the AI screenwriter. Enter some text, select a movie, and watch as the AI screenwriter generates a new scene! You can also enter some editor's notes to give the AI screenwriter some feedback.

The final step is deploying to production.

Step 5 (Optional): Deploy toProduction

If you want to deploy a Production version of your app, you can do so in just a couple of clicks!

First, you need to use your own OpenAI API key in the OpenAI connection:

  • Click on the Connections tab in the left sidebar
  • Click on the OpenAI connection
  • Edit the connection and use your API key for the Production environment

Now, deploy your app to Production:

  • Click on the Deploy button in the bottom right corner of the Gadget UI
  • Click Deploy Changes

That's it! Your app will be built, optimized, and deployed!

You can preview your Production app:

  • Click on the app name at the top of the left sidebar
  • Hover over Go to app and click Production

Alternatively, you can remove --development from the domain of the window you were using to preview your frontend changes while developing.

Next steps

Congrats! You've built a fullstack web app that makes use of generative AI and vector embeddings! 🎉

In this tutorial, we learned:

  • How to create and store vector fields in Gadget
  • How to stream chat responses from OpenAI to a Gadget frontend using Vercel's AI SDK
  • When to use global actions vs routes in Gadget

This app also contains Gadget's built-in authentication helpers, which we did not cover in this tutorial. If you want to learn more about auth in Gadget, check out the tutorial: Build a blog with authentication

Questions?

If you have any questions, feel free to reach out to us on Discord to ask Gadget employees or the Gadget developer community!

Keep reading

No items found.
Keep reading to learn about how it's built

Under the hood

We're on Discord, drop in and say hi!
Join Discord
Bouncing Arrow