About the [bolt.diy] New Features Discussion category

bolt.diy [bolt.diy] New Features Discussion
1

Use this category to talk about pull requests, new feature requests, and discussions around additions to bolt.diy!

2 Likes
2

I have a list of features that I would like to know if already possible or requires specific implementation:

  • Transcribe YT videos (post URL for text)
  • Excel sheets integration (Google Sheets)
  • create professional videos (Video Generator, Stable Diffusion, Pictory.ai)
  • Website scraping (based on keywords, mindmap)
  • AI Playground (generate responses from multiple LLMs simultaneously)
  • screenshot to code (e.g. V0)
  • 3D object generation (e.g. stability.ai)
  • Music generator (to create own remixes)
  • Image generator (Flux)
  • PDF chat (also split and merge)

I would appreciate to know if the artifact and LLMs can already handle these? Please only concise replies for all points, not just general basic answers (unless Bolt.diy can do them all via e.g. boltstarter templates).

Thanks!

3

I was watching this YouTube video
This NEW AI Coding Platform BEATS Bolt.new (Lovable AI)
Do we have any plans to connect to Backend-as-a-Service (BaaS) such as Superbase or Firebase. If we get Auth, consume API endpoints and DB connected to our generated project, it can become a game-changer.

1 Like
4

Love your use cases here! bolt.diy can be used to build the frontend for all of these things, but I wouldn’t use it to build a backend which would be necessary in general for each of these.

My recommendation would be to build a frontend with bolt.diy and then a backend with an AI IDE like Windsurf/Cursor. The IDE could also tell you how to integrate the frontend with the backend so the whole process should be pretty smooth!

5

Yes this is on the roadmap! Starting with Supabase.

2 Likes
6

Thanks Cole, so it’s not possible to build both frontend and backend yet using bolt.diy alone, maybe in conjunction with additional agents or boltstarter.com?

I’m not agile enough with windsurf or cursor.

7

Hi Cole, thanks for the great work. Here are some suggestions for new features:

  • implement monorepo structure with nx for easy code scaffolding, generation, etc.
  • add .devcontainers configuration for easy local spin up of local containerized development environment with tools like devpod or vscode remote development (vscode devcontainers)
  • add unit and e2e tests (and possibly update the contributing guidelines) to ensure all application usage scenarios and workflows work as expected (unit and e2e testing setups for an app can be easily scaffolded with nx jest, nx playwright generators
  • use LaVague for easy e2e tests with AI Web Agents
  • add codegate for safety considerations of local AI assistants (secrets encryption, dependency risk mitigation, security reviews)
  • would be nice to have unit-tests not just for the app itself, but also for LLMs (consider hundreds of different cases, to ensure an automated way to test proper LLM performance for certain use cases). Also makes sense to broaden the perception (and implement it with tooling) of unit tests to both of the app and the LLM, so that all parts taking part in the delivery of functionality can be ensured to work as expected.
  • consider using caddy or nest.js as a server, AuthCrunch for auth
  • consider deploying the app to vercel, pointing to publicly available LLMs (Grok free tier? / other) for demo purposes, and adding the link to the github repo
  • consider using SafeLine to protect the web app from attacks and exploits
  • consider developing a desktop app with Rust and Tauri (check out nx rust)
  • consider next.js for frontend (with MUI as a UI-library) for webapp, expo for mobile app (with react native paper as a UI-library). Both MUI and react native paper support Material Design Guidelines (Paper supports material you (material design v3), MUI supports material design v2, support for v3 WIP
  • consider automatic creation and maintenance of architecture decisions record with log4brains for the apps created by bolt.diy (MADRs could be updated on every prompt that changes the architecture of the generated app)
  • consider execution of the code generation not only on prompts, but also on tickets (user stories, enabler stories) and ticket sequences. With a built-in project (issue tracker) to persist and track tickets. (Imagine, a description of the app functionality in a sequence of persisted user stories + codegen on it).
  • consider adding a project / issue tracker to bolt.diy for persistence of tickets (user / enabler stories, bugs, etc.)
  • consider generating unit and e2e tests when running the codegen prompts (or executing tickets / sequences)
  • consider generating a knowledge base aside with the code (with the explanations of the techstack choices, features, app components etc.) for the generated apps
  • consider adding a bolt.diy specific LLM evaluation benchmark with code generation tasks and evaluation leaderboard for LLMs with clear ratings (which LLMs performs best on the bolt codegen tasks)
1 Like
8

Elasticsearch search engine (Apache Lucene)

1 Like