# fromspec — Product Overview

> **Spec it. See it. Ship it.**

## What it does (brief)

fromspec is a **spec-first AI app builder**. Instead of prompt → code, it produces four reviewable layers in order — a structured **Spec**, a live interactive **Prototype**, an **Architecture** blueprint (data models, API routes, file structure), and finally a full-stack **codebase + Docker**. The spec is the source of truth; the code is a compiled artifact you own. Chat is available at every layer, so you refine intent and the downstream layers regenerate.

## Who it's for

- **Roles:** founders / technical co-founders, product managers, full-stack and backend developers, agencies and consultancies building client MVPs, internal-tools teams.
- **Industries:** SaaS startups, digital agencies, and any team that ships web apps (auth + payments + notifications cover the common SaaS shape).
- **Buyer profile:** people who want AI speed but refuse a black box or a platform they can't leave.

## Problems it solves

1. **Black-box AI generation** — you get thousands of lines you didn't review and don't understand. fromspec makes you approve a spec and see a working prototype *before* production code exists.
2. **Wrong-thing-fast** — "Prompt → 🤞 → code → that's not what I wanted." Gating on Spec + Prototype catches misalignment when it's cheap to fix.
3. **Vendor lock-in** — most AI builders deploy to *their* platform. fromspec hands you a Dockerfile; `docker compose up` runs it anywhere.
4. **Toy backends** — component generators stop at the frontend. fromspec generates real architecture (FastAPI + Postgres, or Next.js full-stack).
5. **Blind regeneration** — changing one thing nukes everything. fromspec does spec-versioned, diff-based, contextual updates.

## Key benefits / outcomes

- **Four layers of confidence** before a line of production code — a reviewable spec, a live prototype, architecture blueprints, then the codebase.
- **Idea → working prototype in minutes**, production code shortly after.
- **You own the code.** Deploy to Vercel, Railway, GitHub Actions → any VM, or self-host. No platform tax.
- **Production-grade, not a demo:** FastAPI + PostgreSQL + Docker, Tailwind + shadcn/ui.
- **Iterate safely** with spec version history + diff viewer.

## Features / use cases worth highlighting

- **Toggleable capability modules** (no code-from-scratch): Auth, Payments (Stripe / Razorpay per region), Email notifications — injected per stack (FastAPI or Next.js).
- **BYOLLM:** bring your own Anthropic / OpenAI / Gemini / self-hosted Ollama key.
- **One-click deploy** + Push to GitHub.
- **Auto-generated Issues** previewing what building the spec entails.
- **Use case:** validate a SaaS MVP with stakeholders on a clickable prototype, then ship the owned codebase — same afternoon.

## What it solves *differently* than others

The category sells you an **output**; fromspec sells you **a process with gates**.

| Other AI tools | fromspec |
|---|---|
| Prompt → code (black box) | Chat → spec → review → prototype → code (transparent, gated) |
| Generate UI components | Generate full-stack architecture |
| Frontend-only / toy backends | FastAPI + Postgres + Docker |
| "Deploy to our platform" (lock-in) | `docker compose up` anywhere |
| One-shot regeneration | Refine at any layer, regenerate downstream via diffs |

> **Lovable gives you a demo. v0 generates components. They lock you in.**
> **fromspec gives you a spec you approved, a prototype you tested, architecture you inspected — and a codebase you own.**

**One-sentence version:** the only AI builder that makes you confident *before* it writes code, and leaves you with no lock-in *after*.