Speeding Up AI Development With AI-Digest


Lately, I’ve been building projects with LLMs such as Claude and ChatGPT.

I use these AI tools, especially for evaluating ideas quickly, working with new languages and frameworks. And today, I figured out a way to make this process even faster.

My wish was always, that LLMs know about my code base. There are tools for this, but mostly paid. If you already have a premium subscription for ChatGPT or Claude, then this way is absolutely free.

Introducing ai-digest

ai-digestis a NPM library, that turns your code base into a single markdown file. Then, you can use this file to pass it into the LLM of your Choice. For ChatGPT, one can create custom GPTs and upload a knowledge base. For Claude, you can create projects, and also upload relevant files the model should know about.

Run npx ai-digest to generate the file, by default named codebase.md. Alternatively, you can also use bunx ai-digest.

By default, this tool will ignore some files, such as node_modules. But you can also specify the files you want to be ignored, by creating a .aidigestignore. This file can be filed as usual, here is an example from my latest SvelteKit project:

.gitignore 
/static 
/build 
.npmrc 
.prettierrc 
.prettierignore 
local.db 
.env.example 
.env 

Have fun building with this tool!