What to build after architecture.md for your Dreamflow app

Once your architecture.md is in a solid place, the next meaningful step is connecting a backend. This is the moment when your app stops behaving like a prototype and starts behaving like a product. Screens gain real data. Logic anchors to actual state. The flows you’ve planned matter more and more.

Dreamflow includes Firebase and Supabase out of the box, so you can wire up data and auth without leaving the IDE. And because the agent uses architecture.md as its source of truth, your backend structure stays aligned with your plan—not whatever a one-off prompt might have generated.

Here’s the workflow teams follow when moving from a blueprint to a working product.

1. Start by planning your schema

Before connecting to anything, ask the agent to propose a backend schema grounded in your architecture.md. This keeps your tables, fields, and relationships consistent with the flows you’ve already defined.

Copy-and-paste prompt:

Based on architecture.md, propose a backend schema (tables/collections + fields). 

Save it in architecture.md. 

Ask clarifying questions if needed.

The goal is simple: tighten the feedback loop between your blueprint and your data.

2. Connect your backend inside Dreamflow

Pick the backend that matches how you think:

• Firebase if you want an all-in-one stack (Auth, Firestore, Storage).
• Supabase if you prefer Postgres and SQL workflows.

Open the module, click connect, and Dreamflow handles the client setup. No boilerplate, no manual config files.

Firebase steps

1. Open the Firebase module in the left panel.

2. Click "Connect to Firebase".

3. Sign in.

4. Create a new project or link an existing one.

Supabase steps

1. Open the Supabase module in the left panel.

2. Follow the connect flow.

3. Link your Supabase project.

Dreamflow generates client setup automatically—no boilerplate, no config files.

3. Build your first real data flow

Start small. Choose one contained path from your architecture.md—something you can test end-to-end.

Implement a single CRUD flow for [MAIN_OBJECT]. 

Include list + detail + create. 

Add loading and empty states. 

Keep changes aligned with architecture.md.

Examples of [MAIN_OBJECT]:

• Habit
• Product
• Task
• Entry

This single flow is usually enough to unlock real momentum:

• device builds with live data
• usable test sessions
• clearer iteration cycles

Once one slice works, the rest of the app has a blueprint to follow.

This is usually enough to unlock real iteration: device builds with live data, shareable test sessions, and feedback that reflects how the product actually behaves.

Where to go next

With a backend connected, architecture.md becomes the map the agent follows—state, data dependencies, and feature behavior all flow from it. Each prompt now extends the system you’ve defined.

👉 Connect your backend

‍

‍