English source · 3 min

Your first Veyndo Commerce round: change, review, publish

Explore one connected workflow before reading the complete feature list. This is a working prototype; use synthetic data and simulated payments.

Watch without an account

Watch the 1:56 AI commerce overview with English captions. It introduces the agent workspace, visual design and app platform with actual UI captures, synthetic data and simulated payments. The films have subtitles and no sound. All product films and capture notes.

Run your own local tour

The container path needs Git, Docker with Compose and OpenSSL. It does not require Rust, Node, Python, a model download or provider credentials.

git clone --depth 1 https://github.com/sthamann/veyndo.git
cd veyndo
./scripts/try.sh

Open http://127.0.0.1:8788/, then Studio → Create shop. Create a personal local owner account and a separate synthetic shop. No account or registration on veyndo.ai is needed. Your local data stays in the Compose project's named volume.

The published image targets Linux amd64. ARM hosts need Docker's x86 emulation; native ARM packaging and its acceptance are not claimed. If the public image is not available, build from source. The first pull may take time. Choose another local port/project without touching other development shops:

TRY_PORT=8789 TRY_PROJECT=veyndo-my-tour ./scripts/try.sh

Set up your own shop

The optional Set up your shop entry in Studio offers six review areas, an inline company form and a saved return point. It links to the existing catalogue, design, checkout, legal and test tools. Illustrated setup guide and current local acceptance.

Complete the round

In the shop's storefront, choose a coat size and place one simulated order. Inspect that saved order in Studio. This exercises the existing commerce path.
In Studio's environment controls, create a sandbox from your synthetic shop and switch into it. Check the displayed environment before editing.
Open a product and change its description to a clearly fictional tour note. Save it in the sandbox. The live product must still have its original description.
Open Review changes. Select that product change only, inspect it against the live baseline, then explicitly publish the selected change.
Return to the live product and verify the description. Live inventory is preserved by content publication. A concurrent live edit should cause a conflict, rather than being silently overwritten.

This first round works without an LLM. After configuring a model, you can ask for a bounded product proposal and inspect its fields before approval. Model configuration or a generated summary alone does not prove a completed action.

Let fictional customers test the private change

After configuring a model, stay in the full private sandbox and open Shop knowledge → Customer simulator, or Shop design → Test with customers. Describe one shopping goal, edit the customer budgets and run the test. Inspect the saved native observations and test orders. Change one private setting, choose Use these customers again, and compare the two runs before selecting a release. Illustrated simulator guide and evidence.

A public shopper can separately ask for a gift or comparison, edit the visual product selection and use the ordinary product options and checkout. Visual shopping guide.

./scripts/try.sh stop
./scripts/try.sh

Reuse the same TRY_PORT and TRY_PROJECT values if you changed them. Stopping does not delete orders or accounts. Do not remove the data volume if you need them. The preview tag moves; use VEYNDO_IMAGE=ghcr.io/sthamann/veyndo:sha-COMMIT_SHA with an actual published full commit SHA when you need an exact source version. Choose that SHA from a successful Publish tested local preview image run. Documentation-only commits do not create another image. A manual Run workflow on main can rebuild intentionally; a documentation commit landing during a build does not prevent the tested runtime from advancing the preview tag.

Tell us where you got stuck or what you built or report a reproducible bug. Include your source/image version, operating system and the first unexpected step. Keep credentials and customer data out of public reports.

Sandbox boundaries · Build useful apps · Connect an agent · First contributions