Swift Tasks is a browser-based, Bring-Your-Own-Key (BYOK) AI vibe-coding platform that converts a natural-language description into a complete, multi-file web application through token-by-token streaming generation, with a live rendered preview that updates as generation proceeds. The platform's central architectural contribution is fully isolated multi-project workspaces: per-project Zustand store instances on the client (implemented as a factory-and-registry pattern, Map ) and projectId-scoped database rows with cascade deletes ensure that chat history, files, model configuration, and generation state never leak between projects. Swift Tasks supports five categories of large-language-model (LLM) providers – a free platform demo model (GLM-4.6), OpenRouter, OpenAI, Anthropic, and custom OpenAI-compatible endpoints such as Groq, Together, DeepSeek and Ollama – with model identifiers auto-fetched from each provider's /models API and automatic fallback to the demo model on configuration, authentication, or region errors, without ever forwarding the user's key to the fallback provider. Generation executes as a detached, server-side job managed by a JobManager singleton, so that closing the browser tab does not terminate an in-progress build; the client reattaches through Server-Sent Events (SSE) using an in-memory ring buffer and a resumable cursor. All API keys and arbitrary secrets are encrypted at rest with AES-256-GCM, returned to the client only in masked form, and revealed solely through an explicit user action. The generated application is rendered inside a sandboxed iframe that has no access to the parent window, cookies, or local storage, and is shipped to production through one-click Netlify deployment, ZIP export, or self-contained HTML, alongside deployment guides for Vercel, Cloudflare Pages, Surge, and GitHub Pages. The system is implemented with Next.js 16 (App Router), TypeScript 5, Prisma ORM over Neon serverless PostgreSQL, shadcn/ui with Tailwind CSS 4, Monaco Editor, the Vercel AI SDK, and Zustand. By combining real-time generation, complete project isolation, secure BYOK key management, sandboxed preview, and one-click deployment in a single zero-configuration platform, Swift Tasks offers an open, self-hostable, and cost-accessible alternative to proprietary AI coding assistants for individual developers, students, and small teams.
Pamisetty Maria Bala Akshay Raj, Nethi Adarsh Ram· Zenodo (CERN European Organi...· 0 citations
Swift Tasks is a browser-based, Bring-Your-Own-Key (BYOK) AI vibe-coding platform that converts a natural-language description into a complete, multi-file web application through token-by-token streaming generation, with a live rendered preview that updates as generation proceeds. The platform's central architectural contribution is fully isolated multi-project workspaces: per-project Zustand store instances on the client (implemented as a factory-and-registry pattern, Map ) and projectId-scoped database rows with cascade deletes ensure that chat history, files, model configuration, and generation state never leak between projects. Swift Tasks supports five categories of large-language-model (LLM) providers – a free platform demo model (GLM-4.6), OpenRouter, OpenAI, Anthropic, and custom OpenAI-compatible endpoints such as Groq, Together, DeepSeek and Ollama – with model identifiers auto-fetched from each provider's /models API and automatic fallback to the demo model on configuration, authentication, or region errors, without ever forwarding the user's key to the fallback provider. Generation executes as a detached, server-side job managed by a JobManager singleton, so that closing the browser tab does not terminate an in-progress build; the client reattaches through Server-Sent Events (SSE) using an in-memory ring buffer and a resumable cursor. All API keys and arbitrary secrets are encrypted at rest with AES-256-GCM, returned to the client only in masked form, and revealed solely through an explicit user action. The generated application is rendered inside a sandboxed iframe that has no access to the parent window, cookies, or local storage, and is shipped to production through one-click Netlify deployment, ZIP export, or self-contained HTML, alongside deployment guides for Vercel, Cloudflare Pages, Surge, and GitHub Pages. The system is implemented with Next.js 16 (App Router), TypeScript 5, Prisma ORM over Neon serverless PostgreSQL, shadcn/ui with Tailwind CSS 4, Monaco Editor, the Vercel AI SDK, and Zustand. By combining real-time generation, complete project isolation, secure BYOK key management, sandboxed preview, and one-click deployment in a single zero-configuration platform, Swift Tasks offers an open, self-hostable, and cost-accessible alternative to proprietary AI coding assistants for individual developers, students, and small teams.
Pamisetty Maria Bala Akshay Raj, Nethi Adarsh Ram· Zenodo (CERN European Organi...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
