Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
Local catalog consolidation
The catalog detail page offers Local consolidation for two explicitly selected, same-type ingredients owned by this instance. Both records must already be unlocked. A matching name is not evidence of identity; the operator selects the source and confirms the reviewed result. The page being viewed is the survivor.
Consolidation unions lists and asks for a choice for every unequal populated scalar, including fields inside array objects with the same stable ID. It keeps both names discoverable, preserves images and attachments, and retains the complete original histories and provenance in a local receipt. A second portrait becomes a reference so it remains visible without replacing the survivor's portrait.
The source stays unchanged and live for peers. On this instance its stable ID resolves to the survivor, and local list/search/facet results show one identity. Exports preserve both original identities and their media. This is not a federated deletion or a cross-owner merge. Consolidating a foreign-owned record requires an owner-coordinated operation; this API refuses it.
Embedded project/canon copies are preserved as contextual records. A retained source-ID canon snapshot cannot overwrite the consolidated catalog payload. A subsequent explicit catalog edit can refresh matching embedded canon through the existing projection path. Sheet rendering uses consolidated catalog data rather than delegating to an older embedded variant. Existing image jobs attaching to the old ID resolve to the survivor.
API workflow
All paths are relative to /api/catalog and use the existing authenticated API.
No endpoint unlocks a record or performs generation.
POST /ingredients/mergewithsourceId,survivorIdanddryRun: truereturnsproposed,conflicts,counts, andpreviewToken. Each conflict names a JSON-pointerpathand thesourceandsurvivorvalues.- Set
fieldChoicesto a map from those paths tosourceorsurvivorand request a fresh preview. For example,{ "/name": "survivor" }retains the survivor's display name. Do not reuse a token after changing choices. - Submit the same IDs and choices with
dryRun: falseand the reviewedpreviewToken. Apply checks the full record, media, link, relation and revision snapshot inside a database transaction. Any intervening change returns409 CATALOG_MERGE_STALE; obtain and review a new preview. GET /ingredients/:id/merge-historyreturns paginated receipt summaries (limit,offset, andnextOffset: nullwhen exhausted).GET /ingredient-merges/:receiptIdloads one original snapshot on demand. Revision history also includes source revisions with their originalingredientId, so provenance stays visible and restore is still available.POST /ingredient-merges/:receiptId/recoverwith{}restores the original identities and associations only if the applied state is still unchanged and both records remain locally owned and unlocked. It retains audit history and emits tombstones for added reference/media/relation links. It refuses to overwrite later edits, jobs, attachments, or locks. There is no force option.
An identity participating in an active consolidation cannot enter another one; recover first. This prevents redirect chains and cycles. A receipt remains after recovery. Source rows and active survivors cannot be deleted while a consolidation is active. Legacy peer updates cannot rewrite the retained source.
Rollout
This feature requires the server code and its additive boot schema (the local
catalog_ingredient_merges receipt table, resolver function and protective
trigger), plus the client build. The schema does not enqueue work or consolidate
records. A normal authorized deployment/restart is required before the endpoints
exist on a running instance. Code merge alone does not deploy it. Each subsequent
catalog consolidation is a separate explicit operator action.