Hacker News Reader

anorak27on Postgres with Builtin File Systems

You can do the same and a bit more with sqlite through agentfs

https://github.com/tursodatabase/agentfs

anorak27on Ask HN: Share your personal website

https://anilturaga.github.io/

I write blogs with interactive components on research and personal projects.

anorak27on Show HN: Diagram as code tool with draggable customizations

Wonderful project.

There's also mermaidjs to excalidraw https://github.com/excalidraw/mermaid-to-excalidraw

anorak27on React Flow, open source libraries for node-based UIs with React or Svelte

I made a desktop UI using react flow with resizing and deletion of windows.

https://github.com/Anilturaga/Generative-UI

anorak27on Show HN: Nestable.dev – local whiteboard app with nestable canvases, deep links

This is very cool! Thanks for sharing

anorak27on Show HN: Nestable.dev – local whiteboard app with nestable canvases, deep links

Tldraw's default version of nesting is embedding another full canvas. The benefit would be that the embedded canvas is fully live.

Nestable enables nesting at page management level and encourages deep links to connect things between canvases. This approach has proven to be much more scalable in all of my workflows

anorak27on Show HN: Nestable.dev – local whiteboard app with nestable canvases, deep links

Thanks for trying it out!

1. Tldraw support of embedded canvases it not great when you are power using 2. You can open the sidebar from clicking the page name in the center top of the canvas 3. Initially I had the sidebar be a side component that pushes the canvas to the right but the change in aspect ratio was jarring because I was constantly opening and closing the sidebar 4. Made with tldraw is mentioned on the right bottom of the canvas. I am not trying to hide it in any way.

anorak27on Show HN: Nestable.dev – local whiteboard app with nestable canvases, deep links

Haven't heard of muse before but it looks great!

Nestable approach with canvas management is more similar to notion than muse.

Nestable also has deep linking across the app so that you can leave hyperlinks to other pages or shapes from anywhere to enable better organisation and management.

Canvases aren't generally used as knowledge bases because more often than not, it gets really hard navigating them and nestable wishes to solve that.

In terms of making a business of out nestable, I have no plans for it. This is fully local and the only charge I incur would be for hosting which is very minimal.

anorak27on Show HN: Nestable.dev – local whiteboard app with nestable canvases, deep links

Thanks for trying it out!

Zooming in to reveal things will only make it more ambiguous since the right depth at which we hide away content will vary based on the content.

We can more intuitively build this with nestable using deep links. Each layer/level can be shown in one canvas and a deeplink to another canvas that captures a more granular level of any of the components would be a much scalable and generic approach.

anorak27on Ask HN: What are you working on? (March 2025)

I've just completed solving Cursor IDE model completion problem called Character prefix conditioning.

https://github.com/Anilturaga/Character-Prefix-Conditioning

anorak27on New tools for building agents

I have built myself a much simpler and powerful version of the responses API and it works with all LLM providers.

https://github.com/Anilturaga/aiide

anorak27on Show HN: FlakeUI

Reminds me of prezi[0]. It would be great if there is an open source version of prezi similar to reveal js.

[0] https://prezi.com/p/p6evz0gdy5dr/ux-design-tips-for-product-...

anorak27on Aiide: Pragmatic framework to build LLM Copilots

Sorry but could you please elaborate?

anorak27on Logica – Declarative logic programming language for data

There's also Malloy[0] from Google that compiles into SQL

> Malloy is an experimental language for describing data relationships and transformations.

[0]: https://github.com/malloydata/malloy

anorak27on Why Search is not a solved problem yet

I agree. However, I felt all the solutions in this space have converged to offering similar user experiences.

I personally wouldn't mind having the memory feature[1] in ChatGPT integrated with my search engine so that the results are more related to my task.

[1]-https://openai.com/index/memory-and-new-controls-for-chatgpt...