Transcript

Hey everyone, today I want to show you a project I've been working on. It's a command line database agent built with TypeScript and the Versal AI SDK. Think of it like a Gemini CLI or Claude code assistant but specialized for database and back end development all within your terminal. The real magic comes from the custom tools I've built for it. The agent can use file system tools to read and edit code, a whole suite of database tools for interacting with Drizzle ORM, and even tools for API and front end integration like creating API routes or entire React. Components. OK, let's see it in action. Here's my projects database running on Neon. You can see we have a few tables here, but nothing for products yet. We're going to build that from scratch. Alright, I'll just run the agent with NPM run agent. It has a neat little interactive mode for guided tasks, but for this demo, we're going to give it a direct one shot prompt to show its full automation power. Here's the prompt, just plain English. Create a products table and its API's. That's it. It's initializing the AI and tools. And right away it's starting the workflow by creating the database schema. You can see in the source control that it is already modified my schema TTS file, intelligently adding a new products table without messing up my existing schemas. Back in the terminal, it's automatically generating the migration with Drizzle Kit, and now it's running that migration to apply the changes to the actual database. If we refresh our Neon console, there it is a brand new products table with all the columns we'd expect. The prompt also asked for API's. The agent is now creating the API directory and root handlers. You can see it created a products folder under API including a dynamic ID route for fetching single products. The entire execution is complete. Here's the code it generated. We have full CRUD API endpoints for our products table, complete with pagination, filtering, and zod validation. Let's start the dev server to test this out using Thunder Client. I'll first hit the API products endpoint and perfect, we get back the list of all our products from the database. Now let's try getting a single product with ID 4 and that works too. Let's give it a follow up prompt. I'm telling it the products table is already there create. Front end components and hooks to fetch that data and add it into the existing UI. It's going to use the context of our previous work. Here we go. Notice how it's skipping all the database steps. This time it understands the table already exists and is jumping straight to creating the front end integration. Now I'll just quickly import that new component into my main page. Alright, let's refresh the browser. And there it is, a full data table rendered in my UI, pulling live data from the database we just created. Using the API and the front end components that the agent generated all starting from a couple of simple text prompts. This is the power of AI driven development tools.