Photo β Pattern β Futures β First Move
See what a place could become.
Upload one property photo, answer the obvious questions, and get a believable first plan: what the place is trying to become, where the ideas sit, and what to test first.
Start here
Create your result board.
Upload a photo or run the self-test. VerdeAI will build a clear board with six futures, a recommendation, and one first move.
Run the shaded garden self-test. VerdeAI will simulate photo upload, starter clue, analysis, report, Vision Board, and a saved project card.
v6.2 keeps a local recovery copy while you test, so closing the page should not mean starting from zero.
Tester pass
Share this without explaining the whole thing.
Ask testers to do one quick loop: upload a real photo, tap the closest starter clue, read one overlay card, then copy the tester summary or send quick feedback. For app checks, use the public tester page or shaded garden self-test first.
- Upload or demo
- Tap one clue
- Read overlay
- Save or export
Share this first
One photo. One clue. One visual idea.
This is the simplified page for real testers. They should not need to understand the whole app. Upload a photo or run the shaded self-test, then look at the plant overlay and first move. AI rendering is being prepared, but the safe beta still shows concept overlays until you choose a paid provider.
- 1Upload or run self-test
- 2Tap one starter clue
- 3Screenshot the plant overlay
- 4Copy the tester summary
Plant overlay mockup, not a final AI render yet. It is meant to show where ideas could sit on the actual photo.
Generated workspace
Six possible futures
The cards will become more specific once you run an analysis.
Property DNA
What VerdeAI noticed
Biggest warning
Run the analysis first. The garden oracle refuses to guess before coffee.
Site clues
Add a main problem and property note to make the result sharper.
Copy this after testing. v6.2 keeps the current analysis stable, improves the public dashboard polish, and prepares render prompts without charging for AI.
Refine selected future
These controls change style only. In v6.0 the analysis stays locked and autosaved, so design refinements cannot reset the situation, pattern, main problem, or overlay labels.
Safe to test: tick Cleaner / minimal, then check Report or Vision Board. The result should stay on the same analysis.
One clean handoff
Copy the tester result, send it with one screenshot, and ask for one simple reply: useful, confusing, or not believable yet.
- Run a real photo or self-test.
- Tap Copy clean tester result.
- Send it with one screenshot of the board.
- Ask: βWould you try the first move?β
Shareable test summary
Use this when sending VerdeAI to an early tester by text, Messenger, Facebook, or a feedback form. Keep it short: one board screenshot, one copied result, one question.
Real rendering, safely locked until the backend is ready.
VerdeAI still uses free concept boards. v6.2 makes the next rendering step easier to understand: pick one future, preview the prompt and likely cost, then only render after a secure backend confirms everything.
Recommended path: start with one provider and one image.
VerdeAI is ready to plan real rendering, but paid rendering stays disabled until a backend proxy, server-side API key, and cost confirmation are in place.
Good first candidate for realistic landscape-style images with simple per-image testing.
Needs backend proxy and Replicate API key.Strong prompt-following option, but pricing and image workflow should be tested carefully before all-six rendering.
Needs backend proxy and server-side OpenAI API key.Credit-based image provider option. Useful later if it gives better cost or style control for property concepts.
Needs backend proxy and server-side Stability API key.- Choose one provider only: start with Replicate / FLUX.
- Render one future image only.
- Show the estimated cost before rendering.
- Keep the concept-board fallback if the render fails.
- Require a second confirmation before rendering all six futures.
First provider setup checklist
/api/render/estimate works before /api/render is unlocked.
GitHub Pages is fine for the app, not for secret API keys.
GitHub Pages can host VerdeAI's public interface, but it cannot safely hide provider API keys. Real rendering needs a backend/proxy layer.
Recommended first path: keep the frontend where it is, add one small proxy.
GitHub Pages can keep hosting VerdeAI's public interface. The first real render test only needs a small serverless proxy that hides the provider key, estimates cost, and renders one future after confirmation.
Netlify Functions or Cloudflare Workers
Use whichever is easier to set up from your current accounts. Both can run a small /api/render proxy without putting secret API keys inside the browser.
- Best first provider: Replicate / FLUX.
- First test: one future image only.
- Fallback: keep the concept board if rendering fails.
- Cost rule: show estimated cost before any paid render.
Good match if VerdeAI is later moved to Netlify or you want a familiar static-site + function setup.
Needs: Netlify project, environment variable for provider key, one serverless render function.Good fit if you want to keep the current frontend and add a separate lightweight render proxy.
Needs: Cloudflare account, Worker secret for provider key, CORS allowed for the VerdeAI site.Strong option if the project later becomes a fuller web app with API routes and deployments together.
Needs: Vercel project, environment variable, API route deployment.Most control, but more setup, maintenance, and hosting decisions. Not the best first test.
Needs: server hosting, deploy process, environment variables, monitoring, and CORS setup.First real render path: one small proxy, one provider, one image.
Keep GitHub Pages as VerdeAI's public frontend for now. Add one small backend/proxy only for rendering, start with Replicate / FLUX, estimate cost first, then test one future image after confirmation.
GitHub Pages frontend + small backend proxy + Replicate / FLUX
This is the practical first test path because it keeps the current site working, avoids exposing API keys, and limits spending to one confirmed image.
index.html, JavaScript, or GitHub Pages./api/render/estimate and /api/render. The browser calls VerdeAI's backend, not the image provider directly.Create or choose the backend host, add the private provider key there, deploy the render endpoint, test the estimate endpoint, then approve one paid image test. Nothing in v7.1 makes paid calls.
Do this later, only when one paid image test is worth it.
This guide turns the backend plan into plain steps. Nothing here connects a paid provider yet; it simply shows the safest order for the first real render test.
What will cost money?
- Provider image generation once real rendering is enabled.
- Possibly backend hosting if free limits are exceeded.
- Repeated renders, especially all-six future sets.
What stays free?
- The current GitHub Pages frontend.
- Concept boards, mock render previews, reports, sharing, and saved projects.
- Prompt preview and cost-estimate rehearsal before provider connection.
Do not paste API keys into index.html, JavaScript files, GitHub Pages settings, screenshots, or public GitHub files. Real provider keys belong only in a backend host's secret environment variables.
Choose a future provider plan
Rendering is not connected yet. These controls only prepare the plan, prompts, mock results, and cost warnings. Nothing paid happens from this static page.
Safety rule: API keys belong in server environment variables only. VerdeAI will not make paid image calls until a backend proxy is configured and you confirm the cost.
Before paid renders
Choose one future before spending anything
This is a safe simulation. Pick a future, preview the property-specific prompt, check the estimated cost, then confirm a mock render. No provider is contacted in v7.1. This is only a rehearsal for the future paid render flow.
Cost rule: real rendering will start with one image only. All six futures stay locked behind a second confirmation when a backend provider is connected.
What is needed before real images?
This is the practical path before VerdeAI can turn one concept board into a real AI-rendered image. It stays disabled until each item is handled safely.
Developer detailsBackend proxy contract
Future endpoint shape
The browser should call VerdeAI's own backend. The backend then calls the image provider using a private API key stored in server environment variables.
GET /api/render/providersPOST /api/render/estimatePOST /api/renderPOST /api/render
{
futureId,
provider,
prompt,
imageReference,
confirmCost: true,
maxCostUsd,
count: 1,
fallback: "concept-board"
}
Generated from the current property analysis, selected future, starter clue, visible-site language, and design refinements. Copy one prompt to test manually with a provider later, or use the mock render buttons to check the flow without spending money.
Netlify-ready static package
- Unzip this package.
- Drag the full
verdeai-v6.2folder into Netlify. - No build command is required.
- Open the generated link and test shaded self-test β result board β copy tester result β AI Setup safe-mode check.
Included files
index.htmlstyles/main.cssjs/app.jsconfig.jsnetlify.tomland_redirectsCHANGELOG.mdandBUILD_STATUS.md
Generated result board
Your property futures board.
Every upload, demo, or self-test becomes one clear result: todayβs property, six futures, VerdeAIβs recommendation, and one practical first move.
After a board is ready, this button takes testers straight to the result.
Result summary
Upload, demo, or self-test to create your board.
VerdeAI will summarise the best future, why it fits, and the first practical move here.
Generated from the current photo, selected clues, and VerdeAI pattern logic. These are concept boards, not final AI renders yet.
v7.1 keeps AI rendering optional and secondary. No paid image calls are made here; every result board uses honest concept boards until a safe backend rendering provider is connected.
60-second tester pass
v6.0 is designed so an early tester can create a Property Futures Board, copy a short result, and give one useful sentence of feedback without you explaining every button.
Use this first when you just want to check the app works without uploading the same photo again.
- Upload one real property photo or use demo mode.
- Use the Next best action button if unsure.
- Tap the closest starter suggestion. It runs the analysis automatically.
- Read the first overlay card and the Best first move.
- Open Export and copy the tester invite or summary.
- Save feedback before leaving, or tell the owner what felt fake or useful.
Questions for testers
- Did the six futures make you see the property differently?
- Did the result feel specific to the photo and situation?
- Which future would you actually try?
- Was one first action clear enough?
- Did the clue coaching make it clear that this is not full AI vision yet?
- What felt fake, confusing, or unfinished?
This records useful tester actions such as photo upload, starter clue, analysis, save, and design refinement.
Climate profile
Add a postcode and run analysis to generate a basic climate clue.
Maintenance calendar
Quick view of the current future, overlay labels, design direction, and first experiment. The labels now sync with the latest analysis result.
Saved in this browser/domain only. v6.0 also autosaves the current session after upload, clue, analysis, design, and report changes.