Closes #6406 ## What changes Publishing no longer waits for advisory content-database and MDX diagnostics. Those checks run alongside the publish request, so a slow diagnostic does not prevent publishing from starting. Permissions, pending content saves, and blocking project checks still run before publishing. The server continues to enforce publish authorization and build integrity. ### Publish dialog - Add a green **Validate** button beside the blue **Publish** button, with equal widths. - Show waiting states and disable the actions while their work is running. - Let users validate without publishing. Successful validation shows a persistent “Validation passed. Ready to publish.” toast. - Clear the button's **Validated** state when the popover closes. Reopening still reflects an active publish job. - Move **Add new domain**, **Export**, and **Publish settings** into the header's ellipsis menu. - Show yellow progress icons for publishing domains, green checkmarks for completed publishes, red icons for failures, and neutral icons for unpublished domains. Built-in and custom domains share the status mapping. ### CLI and MCP Add `check-publish-readiness` to check the saved project without starting a publish. It checks permissions, rejects empty or unusable domain selections, reuses the existing build-integrity checks, and returns content diagnostics. Broken resource references no longer produce a false `valid: true` result. ### Content checks Run independent collection reads with bounded concurrency instead of processing every collection sequentially. This reduces serial waiting, but does not eliminate the underlying storage reads or content processing. ## Checklist - [x] Keep authorization and blocking project checks in place. - [x] Run advisory content diagnostics alongside publishing. - [x] Add standalone validation, waiting states, and popover-local validation results. - [x] Consolidate secondary actions in the publish menu. - [x] Distinguish pending, completed, failed, and unpublished domain states. - [x] Add CLI/MCP readiness checks using existing integrity and verified-domain logic. - [x] Bound concurrent collection reads. - [x] Add regression tests, including invalid domains, missing resources, status transitions, and popover state. - [x] Demonstrate the latest regression tests failing without the fixes and passing with them. - [x] Review documentation impact: update CLI operation guidance and metadata, and document domain status colors in the publishing guide. - [x] Run focused Builder, CLI, and asset-uploader tests, Builder typechecking, lint, formatting, and the generated-API check. ## Verification and limits The latest UI check passed all 49 publish tests. The preceding readiness fix passed 98 focused Builder tests and 130 CLI tests; collection coverage passed 112 tests. Typechecking, lint, formatting, and generated-API checks passed. The latest fixes do not affect fixture inputs or generated site output. Storybook screenshots and agent evaluations were not run because approval was not given. A live production publish was not performed during the latest verification. Publish can finish before advisory diagnostics do. Returned warnings appear afterward and cannot stop a publish that has already started. Standalone validation still waits for diagnostics, and actual publishing still requires content synchronization, server checks, and build/deployment work. This does not promise that every publishing timeout is eliminated.
Protocol
@webstudio-is/protocol owns Webstudio data exchange protocols.
The project bundle is the external export/import artifact used by sync, local
.webstudio/data.json, project import, transferred asset files, and prebuild.
This package owns the assembled bundle schemas, transfer RPC payloads, and
bundle compatibility/version helpers.
Domain models such as pages, assets, and serialized builds remain owned by their domain packages and are composed here through schema-only entrypoints. Do not copy those schemas into this package.
Terminology
Project bundle is the authoritative external project artifact. It is used by
sync, .webstudio/data.json, project import, transferred asset files, and
prebuild.
Data envelope is a serialized single-key object whose key is a versioned
Webstudio marker, for example { "@webstudio/instance/v0.1": data }. The
envelope identifies the payload version and lets consumers detect malformed
Webstudio-owned data before falling back to other formats.
Transfer data is the domain payload inside a data envelope. Examples include
InstanceTransferData, InstancesTransferData, and PageTransferData.
Transfer data is portable and detached from a project; insertion/import remaps
ids and merges it into authoritative project state. Runtime-owned transfer
formats live in @webstudio-is/project-build/transfer.
Clipboard data is a UI transport concern. Builder copy/paste may serialize transfer data into a data envelope and place it on the browser clipboard, but runtime and protocol code should name the reusable format as data envelope or transfer data, not clipboard.