AFlowUser Guide ← Back to aflow.adventurous.systems

AFlow User Guide

Welcome to AFlow — a visual canvas where you connect ideas like a flowchart and put AI to work at every step of designing: generating images, building 3D models, and analysing space, with a full record of who decided what.

This guide is for designers and students using the platform — no technical background needed.

This is the source for the published documentation at aflow.adventurous.systems/docs.

Where to start

What's in the guide

More node guides are added as features ship.

Need help?

If something isn't working as expected, note what you did and the message you saw, and contact your administrator or systems@adventurous.systems.

Getting started

This page takes you from signing in to running your first AI node.

1. Sign in

Open the app and sign in with the email and password you were given. The first time you sign in, you'll be asked to set your own password.

2. Meet the canvas

After signing in you land on a canvas — an open space where you build a flow. A new, empty project is created for you automatically. The toolbar across the top has your project name, save controls, your credit balance, and What's New and Docs links.

3. Add a node

Each node does one thing — generate an image, reason with an AI model, build a 3D model, and so on. Open the node toolbar, choose a node, and it appears on the canvas. Nodes are grouped by type (Input, AI Generation, AI Reasoning, Human Reasoning, 2D Transform, Viewer, Annotation, Topologic).

4. Fill it in and run it

Type or choose the node's inputs, then press its run control. The node shows a running state and then displays the result right on the node. AI runs use credits (see Credits & billing).

5. Connect nodes into a flow

Drag from one node's output handle to another node's input handle to pass a result along — for example, send a generated image into a node that turns it into a 3D model. Compatible connections snap together; incompatible ones are blocked automatically. See Building flows for more.

6. Your work saves automatically

Projects autosave a few seconds after you stop working, and when you switch tabs. You can also press Ctrl/Cmd + S. The toolbar shows Saving… / Saved! / Save Failed! so you always know the state. See Sessions & projects to open, rename, export, or share work.

A first flow to try

  1. Add a Text to Image node.
  2. Write a prompt, e.g. "A timber footbridge over a quiet river, morning mist, wide view."
  3. Pick a fast, low-cost model and run it.
  4. Add a Generate 3D node and connect the image into it to build a 3D version.

That's the core loop: add nodes, connect them, run, and refine.

Next

Building flows

A flow is a set of nodes connected together so a result from one feeds into the next. This page covers connecting, navigating, and organising your canvas.

Connecting nodes

  • Every node has handles — small dots on its edges. Outputs are on the right, inputs on the left.
  • Drag from an output handle to an input handle to connect them. A wire appears.
  • Handles are typed (text, image, mesh, and so on). AFlow only lets you connect compatible types — if a connection is blocked, the two handles aren't compatible. A coloured handle and a label tell you what a handle expects.
  • A node input that's connected shows "(from input)" — it's using the upstream result instead of anything you typed.

Deleting a wire

Click a wire to select it — a small × appears at its midpoint; click it to remove the wire. You can also select a wire and press Backspace/Delete.

Moving around the canvas

  • Pan: drag an empty part of the canvas.
  • Zoom: scroll, or use the zoom controls.
  • Move a node: drag it by its header.
  • Resize a node: drag a corner — handy for nodes with long text or image output.

Node panels and overlays

The toolbar includes a few views that help you understand a project:

  • Design decisions — highlights the decision points in your flow.
  • Governance — see and manage which parts of the design are locked as Non-Fungible (see below).
  • Provenance — a timeline of what happened in the project and who/what produced each result.
  • Debug — a log of recent actions, useful when reporting a problem.

Locking work (Fungible / Non-Fungible)

Each node can be marked Non-Fungible (locked) using the lock toggle in its header. Locking marks that part of the design as a deliberate, kept decision — it's preserved as important context as you continue, so later AI steps take your locked choices into account. Leave a node Fungible while you're still exploring; lock it once you've decided.

Tips

  • Keep separate ideas on separate nodes so you can compare them side by side.
  • Name your project (click the name in the toolbar) so it's easy to find later.
  • Use sticky notes (Annotation) to leave yourself reminders on the canvas.

Next

Text to Image

Generate an image from a written description.

What it is

The Text to Image node turns a text prompt into a generated image, right on your canvas. You describe what you want; the node produces a picture you can preview, download, or pass into other nodes.

When to use it

  • Exploring visual ideas early in a design — moods, materials, massing, atmosphere.
  • Producing reference imagery to react to or refine.
  • Feeding an image into a downstream node (for example, turning it into a 3D model).

How to use it

  1. Add the node. From the toolbar, add a Text to Image node to the canvas.
  2. Write a prompt. In the Prompt field, describe the image you want — be specific about subject, style, materials, lighting, and viewpoint. (You can also connect a text output from another node into the Prompt input instead of typing — the field shows "(from input)" when it's connected.)
  3. Choose a model. Pick an option from the Model dropdown. Each option shows its cost per run, and a small status dot indicates whether it's currently available. Different models suit different looks — try a fast, low-cost one while iterating, then a higher-quality one for a final.
  4. Set the aspect ratio. Choose a shape for the image (for example, 1:1 square or a wide/tall ratio).
  5. Run it. Start the node. It shows a running state, then the finished image appears in the node's preview. Each run uses credits.
  6. Use the result. Download the image, or drag from the node's Image output to another node to keep building your flow.

Example

Goal: a quick concept image of a small pavilion.

  1. Add a Text to Image node.
  2. Prompt: "A modern glass pavilion overlooking the sea, soft morning light, minimal steel frame, wide view."
  3. Model: pick a fast, low-cost option to iterate.
  4. Aspect ratio: a wide ratio for a landscape view.
  5. Run. Refine the wording and run again until it feels right, then switch to a higher-quality model for the final image.

Tips

  • Be concrete. "Weathered timber cabin in pine forest, dusk, warm interior light" beats "a nice cabin."
  • Iterate cheaply. Use a low-cost model while you're still wording the prompt; spend on a higher-quality model only for the keeper.
  • Name the view. Words like aerial, eye-level, close-up, wide strongly steer the result.
  • One idea per node. Keep separate concepts on separate nodes so you can compare them side by side.

Troubleshooting

  • "Provider is busy — please try again in a moment." The selected model is temporarily rate- limited. Wait a few seconds and run again, or pick another model.
  • A model's status dot isn't green. That model is degraded or unavailable right now — choose a different one.
  • The run failed and no image appeared. You aren't charged for a run that returns no output. Check your prompt and try again, or switch models. If it keeps failing, contact your administrator.
  • "Insufficient credits." Your balance is too low for this run. Top up (if in-app purchase is enabled) or ask an admin to add credits.

See the User Guide index for more. For how this documentation is kept current, see docs/CONTENT_PLAYBOOK.md.

Image to Image

Transform an existing image into a new one, guided by a prompt.

What it is

The Image to Image node takes an image you already have — an upload, or the output of another node — and produces a new version of it based on your instructions.

When to use it

  • Restyling a concept image (different material, mood, time of day).
  • Iterating on a generated image without starting from scratch.
  • Turning a rough sketch into a more finished image.

How to use it

  1. Add the node from the toolbar.
  2. Give it an image. Connect an image output (for example from Text to Image or an Image Input node) into the node's image input, or upload one.
  3. Describe the change in the prompt field — say what you want different, not just what the image already is.
  4. Choose a model from the dropdown. Each option shows its cost per run and a small availability dot.
  5. Run it. The new image appears in the node's preview; the result is also available on the node's image output to feed onward.

Example

Goal: restyle a daytime pavilion image as dusk.

  1. Connect your pavilion image into an Image to Image node.
  2. Prompt: "Same pavilion at dusk, warm interior lighting, long shadows, calm sky."
  3. Run, then refine the wording until the mood is right.

Tips

  • Describe the difference you want ("make it snowy", "warmer materials"), not the whole scene.
  • Strong, specific changes work better than vague ones.
  • Keep the original on its own node so you can compare before/after.

Troubleshooting

  • No image input connected. The node needs a source image — connect one or upload it first.
  • "Provider is busy." The model is briefly rate-limited; wait and retry, or pick another model.
  • The run failed with no output. You aren't charged for an empty result — adjust and try again.

See the User Guide index.

LLM Reason

Think through a design problem with an AI language model and get a written response.

What it is

The LLM Reason node sends your question or brief to an AI model and returns its reasoning as text you can read, copy, and pass into other nodes.

When to use it

  • Exploring options or trade-offs for a design decision.
  • Summarising or restructuring notes and requirements.
  • Generating a prompt to feed into an image or 3D node.

How to use it

  1. Add the node from the toolbar.
  2. Write your input — a question, brief, or instruction. You can also connect text from another node into the input (it shows "(from input)" when connected).
  3. Run it. The response appears in the node's text preview.
  4. Use the result. Select and copy the text, or connect the node's text output into another node (for example, into Text to Image as a prompt).

Example

Goal: turn a loose brief into an image prompt.

  1. Input: "I want a small lakeside studio for a painter — quiet, lots of north light, timber."
  2. Run. The node returns a richer description.
  3. Connect its output into a Text to Image node to visualise it.

Tips

  • Be specific about what you want back ("give me three options", "in one paragraph", "as a prompt").
  • Chain nodes: reason first, then generate — the text output makes a great prompt.
  • Keep separate questions on separate nodes so answers stay easy to compare.

Troubleshooting

  • The output isn't what you expected. Rephrase the input more specifically and run again.
  • "Provider is busy" / availability dot not green. The model is briefly unavailable — wait and retry, or choose another model.
  • The run failed with no output. You aren't charged for an empty result — try again.

See the User Guide index.

LLM Critique

Get structured feedback on a piece of work from an AI model.

What it is

The LLM Critique node reviews an input — text, or a result from another node — and returns considered feedback: strengths, weaknesses, and suggestions.

When to use it

  • Pressure-testing a design idea or written brief.
  • Getting a second opinion before committing to a direction.
  • Producing review notes you can act on or share.

How to use it

  1. Add the node from the toolbar.
  2. Provide what to critique. Type it in, or connect another node's output into the input.
  3. Optionally add focus. If the node offers a guidance field, say what to focus on (for example "comment on circulation and daylight").
  4. Run it. The critique appears in the node's text preview and on its text output.

Example

Goal: review a concept before refining it.

  1. Connect an LLM Reason output (your concept description) into an LLM Critique node.
  2. Focus: "Is this practical for a small budget? What would you change?"
  3. Run, then feed the suggestions back into a new reasoning or generation step.

Tips

  • Give the model something concrete to react to — the more specific the input, the sharper the critique.
  • Tell it what to focus on so feedback stays relevant.
  • Critique → revise → re-generate is a strong loop.

Troubleshooting

  • Feedback is too general. Add focus guidance and give it more concrete input.
  • "Provider is busy" / dot not green. Wait and retry, or pick another model.
  • Failed with no output. No charge for an empty result — try again.

See the User Guide index.

Generate 3D models

Turn a text description or an image into a 3D model you can view, download, and use downstream.

What it is

AFlow can build a 3D model from either a written prompt (Text to 3D) or an existing image (Image to 3D). The result is a 3D model shown in an interactive viewer on the node, which you can rotate and download.

When to use it

  • Quickly getting a 3D form from a concept image or description.
  • Producing a model to analyse, refine, or hand off to other software.

How to use it

From an image (Image to 3D)

  1. Add an Image to 3D node.
  2. Connect an image (for example a Text to Image result) into its image input, or upload one. A clear single subject on a simple background works best.
  3. Run it. The 3D model appears in the node's viewer.

From text (Text to 3D)

  1. Add a Text to 3D node.
  2. Describe the object — keep it to one clear subject.
  3. Run it.

Working with the result

  • Rotate / inspect the model in the node's viewer.
  • Download it with the download control. Models are saved in a standard 3D format (.glb) that opens in common 3D tools.
  • Pass it on by connecting the node's 3D output into a downstream node.

Example

Goal: a 3D massing from a concept image.

  1. Generate a concept with Text to Image.
  2. Connect it into an Image to 3D node and run.
  3. Rotate to check the form, then download the .glb.

Tips

  • One clear subject, simple background — busy images produce messier models.
  • Iterate on the image first; a better image makes a better 3D result.
  • 3D generation costs more credits than a 2D image — get the input right before running.

Troubleshooting

  • The model looks rough or incomplete. Try a cleaner input image or a simpler description.
  • Large upload won't attach. Very large files are rejected — use a smaller image.
  • "Provider is busy" / dot not green. Wait and retry, or try again later.
  • Failed with no output. You aren't charged for an empty result — adjust the input and retry.

See the User Guide index.

Sessions & projects

Everything you build lives in a project (also called a session) — one canvas of nodes and their results.

Creating and naming

A new, empty project is created for you when you sign in. To rename it, click the project name in the toolbar and type a new one. Naming projects makes them easy to find later.

Saving

Projects autosave a few seconds after you stop working, and when you switch browser tabs. You can also save anytime with Ctrl/Cmd + S. The toolbar shows Saving… / Saved! / Save Failed! so you always know the state. If you try to close the tab with unsaved changes, your browser warns you.

Opening a project

Use the Open control in the toolbar to browse your saved projects and reopen one. You can search the list by name.

Exporting and importing

  • Export: download a project as a single .aflow file (it bundles the canvas, your generated artefacts, and the project's history). Use this to back up work or move it between accounts.
  • Import / restore: bring an .aflow file back in as a new project.

Only import .aflow files from people you trust, just as you would any downloaded file.

Sharing & collaboration

You can invite someone into a project:

  1. Create an invite for the project from the toolbar.
  2. Share the one-time invite code with your collaborator.
  3. They enter the code to join and can then work in the project with you.

Tips

  • Name projects clearly so the Open list stays easy to scan.
  • Export important work as an .aflow backup at milestones.
  • Keep one idea per project when you want a clean record of a single exploration.

Troubleshooting

  • "Save Failed!" Check your connection and try again (Ctrl/Cmd + S). Your work stays on screen.
  • An invite code didn't work. Codes are one-time use — ask for a fresh one.
  • Can't open a project. Make sure you're signed in to the right account; contact your admin if it persists.

See the User Guide index.

Credits & billing

AFlow uses credits to pay for AI runs. This page explains how they work.

What uses credits

Running an AI node — generating an image, reasoning with a model, building a 3D model — uses credits. Browsing, editing, connecting nodes, and saving are free. Each AI node shows the cost of its selected model so you can choose before you run.

Your balance

Your current balance is shown in the toolbar. The colour gives you a quick sense of how much is left, and you can hover it for the exact amount. The balance updates after each run.

If a run fails

If an AI run returns no result, you aren't charged for it. You only pay for successful output.

Running low

If you don't have enough credits for a run, you'll see an "Insufficient credits" message. Depending on how your organisation is set up, you can:

  • Top up in the app — if in-app purchase is enabled, use the top-up option from the toolbar and follow the secure checkout. Your balance updates once payment completes.
  • Ask an administrator to add credits to your account or your group.

Tips

  • Iterate cheaply. Use a fast, low-cost model while you're still refining a prompt, then switch to a higher-quality model for the final run.
  • Check the cost shown on the node before running a more expensive model (3D generation costs more than a 2D image).
  • Get the input right first — a good input image or prompt avoids spending on re-runs.

Troubleshooting

  • "Insufficient credits." Top up or ask an admin; then run again.
  • Balance didn't change after a top-up. Give it a moment and refresh; if it still looks wrong, contact your administrator.

See the User Guide index.

FAQ & troubleshooting

Signing in

I can't sign in. Check your email and password. The first time you sign in you'll be asked to set a new password — use that one afterwards. If you're still stuck, contact your administrator.

I was signed out unexpectedly. Sessions expire for security. You'll see a notice on the sign-in screen; just sign in again and your saved work is still there.

Signing in with a Microsoft account doesn't work. Some institutional Microsoft accounts require extra approval that isn't available yet. Use your email-and-password sign-in instead, or ask your administrator.

Running nodes

"Provider is busy — please try again in a moment." The selected AI model is briefly rate-limited. Wait a few seconds and run again, or choose another model.

A model's status dot isn't green. That model is degraded or unavailable right now — pick a different one.

A run failed and nothing appeared. You aren't charged for a run that returns no output. Check your input and try again, or switch models. If it keeps failing, contact your administrator.

A connection between two nodes won't form. The two handles aren't compatible types. Check that the output you're dragging from matches what the target input expects.

Credits

"Insufficient credits." Your balance is too low for this run. Top up (if in-app purchase is enabled) or ask an admin to add credits.

My balance looks wrong. Give it a moment and refresh. If it's still off, contact your administrator.

Saving & projects

"Save Failed!" Check your connection and press Ctrl/Cmd + S again. Your work stays on screen.

I lost track of a project. Use Open in the toolbar and search by name. Naming projects makes them easier to find.

Can I move work between accounts? Yes — Export a project as an .aflow file and Import it into the other account.

Files & uploads

My upload was rejected. Very large files aren't accepted — use a smaller image or model file.

Still stuck?

Note what you did and the exact message you saw, and contact your administrator or systems@adventurous.systems.

See the User Guide index.