Why it matters
A useful AI change must become maintainable source.
The first answer from a model can be helpful. The production problem is what happens next: testing the result, preserving the decision, keeping secrets out of source, and deploying the exact version you inspected.
The AI edits files, not a throwaway answer
A workspace starts from a repository, template, or imported source tree. The agent changes files inside that source context, so the result can be reviewed as a normal diff.
Every change has a runtime path
Build commands, start commands, ports, health checks, secrets, and managed resources stay close to the code through adios.yaml and workspace config.
The preview teaches you before production
Run the generated change, inspect logs, call the page or API route, and fix the source while the context is still fresh.
Git remains the system of record
Generated code should be committed, reviewed, reverted, and synced like any other product code. The workspace helps the agent participate in that workflow instead of bypassing it.
Example
Ask for a real change.
A developer can ask Adios to add a billing settings page to a Next.js app. The agent edits route files, UI components, and API handlers in the workspace. The developer reviews the diff, runs a preview, checks logs, and commits the change before promoting it.
create a workspace for my nextjs project
add a billing settings page
wire the form to /api/billing
run the preview and show me changed filesChoose the deployment path
Review the target framework, runtime, data service, or workflow before promoting source.
Deploy generated code
Move from AI-assisted edits to previews, logs, and production deploys.
Define the manifest
Keep build, start, port, health, and secret references beside source.
Read the quickstart
Deploy an existing app, inspect logs, and continue in a workspace.