AI application builder · Sandboxed execution
Vibe Coding Platform
An adapted open-source coding workspace designed to turn natural-language prompts into full-stack projects with sandboxed execution, live preview, file browsing, and command logs.
- Role
- Developer · OSS Adaptation
- When
- 2026
- Mode
- Digital product

01 / Context
The problem
AI-assisted application building involves more than returning source code. The interface must coordinate model selection, generated files, command execution, runtime feedback, and a preview while keeping untrusted code isolated from the main application.
Approach
I adapted and deployed an open-source workspace that organises the workflow into four connected areas: an AI chat, a live preview, a sandbox file explorer, and command logs. Vercel AI Gateway provides a common model layer, while Vercel Sandbox defines the intended boundary for generated code execution.
02 / Contribution
What I contributed
- 01
Adapted the open-source application structure for a public portfolio prototype.
- 02
Prepared a multi-model workflow for Claude, GPT, and Grok through Vercel AI Gateway.
- 03
Organised the interface around chat, preview, remote files, command output, and error-recovery settings.
- 04
Deployed the prototype and published its source while documenting that the public AI flow requires service credentials.
03 / Decisions
Technical decisions
Keep execution isolated
Use Vercel Sandbox as the intended runtime boundary instead of executing generated code inside the portfolio or main application process.
Show the whole workflow
Keep conversation, preview, files, and logs visible together so users can follow how an application changes.
Label the limitation
Present the deployment as a prototype because the public interface does not currently expose the credentials required for AI generation.
04 / Project gallery
Evidence, interfaces, and process.


05 / Evidence
Outcome
- Deployed a public prototype of the adapted coding workspace.
- Published the source code with the platform structure and required service configuration visible for review.
- Established a foundation for later API configuration, sandbox execution, and end-to-end validation.