Product

Tutorials are a great way to learn new tools and APIs but many times developers just want to go right into the code and see real examples that work. This week @ilja and I worked on a feature that generates demo repositories when creating tutorials with Mordisco 🧵

Product @crossmint | prev @manifoldxyz @spotify

Our approach in this area has been to leave LLMs out of most decisions we can make ourselves.

For instance, if we know you want to build a Discord bot, we will allow you to select a template for that type of project and make the LLM just generate the pieces of code that make your bot different and unique.

We won’t cover every type of demo out there, but we can probably cover a good chunk of what the platforms we are working with want to generate. Are you building a script, a webhook, a bot, a webapp, a frame, a game? Each of these will have their own templates.

We will use LLMs just for the actual piece of the code that makes your app different (a bot that translates casts, a webhook that notifies me every time someone mentions my brand, etc.)

Having a set of templates for different tutorial types helps set expectations on the user side: you know the file structure, framework, and tools you’ll be using as well as how to run and deploy the code.

You can even automatically host the demo to play with when generating the tutorial.

Here’s an example of a repo we generated with Mordisco:

You can even automatically host the demo to play with when generating the tutorial.

Here’s an example of a repo we generated with Mordisco: https://github.com/mordiscoxyz/optimism-nft-minter

Once we have the project created we use that as context to generate a tutorial around it. Using a different set of prompts we create a setup section that checks the tech you are using and gives you the required steps to run the tutorial.

For instance if you need to get an API key to use Crossmint’s API, it will give you the steps to do so without you having to write it.

We then create a “steps” section that walks you through the code and explains it. Here’s the example from the previous repo:

We then create a “steps” section that walks you through the code and explains it. Here’s the example from the previous repo: https://blog.mordisco.xyz/how-to-create-a-script-that-mints-nfts-from-a-csv-file/

Are you looking to create demos and tutorials that showcase the breadth of your platform? Feel free to ping us!

Are you looking to create demos and tutorials that showcase the breadth of your platform? Feel free to ping us!

https://www.mordisco.xyz

LLMs are pretty good at generating code, but when it comes to generating a full working project, things get a bit more complicated. Vague prompts will push the LLM to make assumptions of what you are trying to build, what framework, tools, file structure, etc. you want to use.