@dillon_mulroy@eastdakota Any chance to have access to dynamic worker in advance? Registrered to the beta but not been accepted yet. I would like to deploy a Shopify admin mcp server that use codemode to serve the huge 3k plus entities of the Shopify admin graphql api. Thanks 🙏
https://t.co/tYNcTEIETR
@haydenbleasel@vercel Thanks the work @haydenbleasel 🙏. I've created a port of Streamdown for Svelte and implemented some of the latest release.
https://t.co/JRy2aNnSyv
Nice stack ! I've just released svelte-streamdown
https://t.co/ZLX2IovzBY
A lightweight way to handle Markdown in Svelte. Port of Vercel’s Streamdown but powered by marked.js only, with some extras like GitHub-style alerts and footnotes.
It simplifies markdown handling soooo much, would be glad to have your thoughts.
day 5 and 6 of vibecoding an ORM. The query parser is merged ⁉
The parser now generates clean, efficient SQL that matches Prisma queries and allow complex filtering and nested relations selection.
✅ Generate one big query to minimize database roundtrips
✅ Uses subqueries instead of joins to maximize database provider compatibility (Hello planetscale)
✅ CTE-Wrapped Mutations to allow complex returns.
✅ Handle implicit many to many relationships gracefully
🧹 We are now trying to cleanup the project after this huge update, reducing code duplication, splitting files to improve code maintainability, writing tests that make sense etc.
🧠 Day 4 of Viborm, the vibe-coded ORM. The query parser is 60% complete, along with the database-adapter interface. Treating the query as an AST was a great idea — the code is simpler and makes more sense. We can now parse and generate simple queries! 💪
We also mapped out the journey of a Viborm query, and here’s what we landed on:
query => query-parser => database-adapter (for syntax specificity) => provider-adapter (managing connection and query execution)
This design promotes strong database agnosticism. For example, the SQL adapter will be able to handle scalar lists using JSON operations, even if the database doesn’t support them natively.
Day 3 😭 of vibecoding an ORM.
Because Claude think i'm always "absolutely right" we spent the day coding a query parser and an AST. Turns out it works but it is not completely useful. Went to the trash this morning.
The idea was to have a DB agnostic AST that will be used by database adapters to build SQL statements.
But we can think of the query as being the AST itself. We just have to find a clever way to design the database adapter around it and the already implemented type system.
Morality:
When vibecoding, do not trust AI when it says your ideas are awesome, they are probably not.
Vibecoding allows you to make more mistakes in a shorter period of time and that's awesome ! A good thing that I didn't spent a week or more working on this AST.
✌️ Day 2 was dedicated to the type inference of the client.
The goal was to have a 1:1 parity with the @prisma client without code generation, relaying only on type inference from a schema.
Claude almost nailed the type system in one shot ! I only had to adjust some generics and conditions.
I do think asking it to write a careful implementation guide before help a lot.
https://t.co/nGnYIR9X2o
Circular reference of model is working !
Self reference is breaking Typescript thought. Maybe some wizard can help me with that ? @FabianHiller@mattpocockuk ?😄