AI-native Trove: Natural-language backend generation

[Akshay1018]

Proposal: Making Trove AI-Native with Natural-Language Backend Generation

Overview

Trove currently helps developers scaffold Node.js backend projects using supported frameworks, databases, authentication options, and development tooling.

I would like to explore the next step: making Trove AI-native so developers can describe the backend they want to build in natural language, and Trove can help translate those requirements into a validated project configuration and a working generated project.

The goal is not simply to add a chatbot. The goal is to make backend project setup more accessible, intelligent, and reliable.

Proposed Experience

Imagine a developer entering the following request:

Build a backend for a SaaS application using PostgreSQL, authentication, role-based access control, Docker, API documentation, and automated tests.

Trove could:

• Understand the requirements.
• Recommend a compatible framework and configuration.
• Ask clarifying questions only when necessary.
• Present a structured project plan for review.
• Generate the project using Trove's existing templates and supported capabilities.
• Run build and validation checks.
• Report what succeeded and what still needs attention.

Developers should retain control over the configuration and generated changes.

Proposed Features

1. Natural-language project requirements.
2. AI-assisted framework and database recommendations.
3. Compatibility validation before generation.
4. Editable project plans.
5. Configurable LLM providers.
6. Automated validation of generated projects.
7. Optional conversational CLI mode.

Design Principles

• Existing non-AI CLI workflows should continue to work.
• AI recommendations must be grounded in actual Trove capabilities.
• Project generation should remain deterministic wherever possible.
• The LLM should not independently invent unsupported integrations.
• Risky filesystem operations and external side effects should require appropriate approval.
• AI functionality should be optional for developers who prefer not to use it.

Questions for the Community

1. Would natural-language backend generation make Trove more useful to you?
2. Should AI be integrated into the existing CLI, offered as a separate command, or both?
3. Should Trove recommend a stack automatically or ask users to choose their preferred framework and database?
4. Which project requirements should be supported first?
5. What level of automation would you trust without manual approval?
6. Which LLM providers or local-model options would you want to use?
7. What would make this feature genuinely better than using a general-purpose AI coding assistant?

Suggested First Milestone

Start with a limited workflow that converts a natural-language request into a validated, editable project configuration. Once that works reliably, connect it to the existing project generator.

This discussion is intended to gather feedback and agree on the direction before defining implementation tasks.

I would appreciate feedback, use cases, architectural suggestions, and contributions from developers interested in AI-assisted backend development.

[bomtobe-hgkim]

trove plan could emit versioned JSON, including unsupported requests; apply only a reviewed plan. That keeps today’s CLI stable. I work on Orke at BOMTOBE; haven’t tried Trove.