Choosing a Home
In plain words: Guide: where an idea belongs
A routing guide for deciding whether a new idea needs a new App or fits as a Packaged Offer (Product), profile, Workflow, view, integration, typed reference or App Bundle (Suite).
- 1A routing table maps each kind of change (new customer, job, vertical, UI, engine) to its proper home.
- 2Shared machinery like the Web Fetcher, Voice or Vision does work, but the consuming app keeps the judgment.
- 3Media ownership follows who accepts the result (visual, music, language, picture sound), not the file format.
Words on this page, decoded click one for its card
“Help a garage diagnose cars and find parts” is an operator promise before it is an architecture. Follow the durable work: a technical service case needs an owner, a parts request may use another owner's result, and a promise to coordinate both live processes needs its own coordination contract.
Workflow owns the law. This page tests where the idea belongs.
One idea end to end
Mechanic is the proposed ProductPackaged Offer a garage would recognize. “Choose the next diagnostic test” is one use case. Workshop, still a candidate, would own the durable service case under its passenger-vehicle profile. One configured garage installation would be a Deployment.
An already-settled part requirement may pass to Scavenger as exact sourcing input. Only a result that must actively coordinate diagnostic and acquisition InvocationsJob Requests needs aan SuiteApp Bundle. None of those packaging or transport choices can supply a missing application owner.
The shortest routing test
| What actually changed? | Give it this home |
|---|---|
| a thematic route for finding related applications | an Area, with links to the existing CompositionApp owners |
| another customer, host, credential set, or local configuration | a Deployment from an exact ProductPackaged Offer revision or Composition-owned reference deployment-profile revision |
| another concrete job the operator needs to finish | a use case supported by a ProductPackaged Offer |
| another profession, market promise, default set, or support envelope | a ProductPackaged Offer identity or immutable revision, according to whether the stable operator promise changed |
| another repeatable score under the same application owner | a PatternWorkflow and immutable ScrollWorkflow Version revisions |
| another vertical, region, provider, or form factor under the same records, authority, recovery, and finish judgment | an exact owner-qualified profile |
| another UI, report, device view, or public explanation | a projection, with its own presentation owner |
| another engine, API, transport, capture surface, or implementation mechanism | an exact Provider, AnimatorModel Runtime/Connector, adapter, or client integration, plus a typed package Contribution through its explicit Core or Extension DomainPlug-in Area owner when registration is required |
| an already-settled result from another owner | an exact typed record, artifact, observation, or receipt reference |
| one live result must coordinate the lifecycle and settlement of several application owners | aan SuiteApp Bundle |
| genuinely new durable records, judgment, effects, outcomes, and recovery reused beyond one ProductPackaged Offer, client, or deployment | a candidate CompositionApp |
Name the owner and kind of every profile: Workshop service profile, Reach deployment profile, AnimatorModel Runtime capability profile. Profile is no global LychD entity. Pin its exact revision and retain the owning records, authority, recovery, and honest finish.
“Translate this” usually names a versioned Translation SpellStep or owner-declared transformation. Language Edition earns application ownership through durable source/segment alignment, casting, spoken performance, captions, conform, acceptance, and recovery. Interface localization remains a Frontend catalogue; common language machinery does not merge them.
A curated Portfolio, not a closed universe
The Portfolio should grow more slowly than ProductsPackaged Offers, profiles, use cases, and Deployments. A numerical cap would instead reward god-Compositions that accumulate incompatible records and safety law.
Before acceptance, follow a case beyond the conversation. Identify the record that must survive, the judgments and policies that govern it, and the endings it must distinguish: finished, refused, partial, or unknown. Explain how stopped work restarts or recovers.
Then test the proposed ownership. Find an existing owner, if one fits, and show why a PatternWorkflow, profile, ProductPackaged Offer, projection, Extension DomainPlug-in Area, package Contribution, Provider, or typed reference cannot carry the work. Name the smallest fixture that could disprove the design and the consumers who need it for independent reasons. Two ProductsPackaged Offers or use cases can corroborate reuse; they are neither quota nor proof. Decorative wrappers count for little.
Reuse without making a god
A profile can vary vocabulary and rules while preserving one lifecycle, authority, recovery, and finish. Workshop can admit a new trade only while the service case still fits. A vague universal casework schema cannot absorb Broker, legal, accounting, or municipal judgment without evidence of a shared root lifecycle.
Follow a serviced heater into a shared enclosure. Homestead keeps the place and its provision, even when a professional visits. The repair establishes neither a colony's seasonal state nor a crop's suitability: Keeper and Cultivator retain those histories and care judgments. Pets and livestock can use Keeper profiles; indoor plants and cultivated fungi can use Cultivator profiles. They need no universal biological state list.
A physical body poses the same test at another boundary. Familiar can reuse its contract while capability, safety, stop, and settlement law stay coherent. A different emergency or recovery boundary may require another owner.
Shared machinery, separate judgments
Start with the work that can be shared without moving its judgment. ScoutWeb Fetcher can search/fetch/render/crawl while an application decides source policy, interpretation, rank, and consequence. Mail/platform delivery carries a message; the application still decides recipient purpose, disclosure, review, reply meaning, and follow-up. EchoVoice/PrismVision & Media similarly supply admitted media effects while their consumers judge interpretation and acceptance.
The same boundary applies to transport and physical work. Tether/Veil provide roads without creating identity or grants. Legion can carry a bounded task while the body retains fresh local refusal. A model or tool provider therefore never owns application decision policy.
Applications connect through exact requests, observations, artifacts, and receipts. They do not exchange ambient databases, credentials, SigilsID Badges, sessions, or judgment. In household work, Wellbeing supplies personal food needs and confirmed consumption; Keeper and Cultivator supply care needs, material-use evidence, and collected yields. Homestead retains inventory/provision and commissioned site effects.
Calling that equipment IoT identifies an integration mechanism. Calling the subjects fauna and flora groups them within the Stewardship & Acquisition Area. Neither label establishes an application owner. The stewardship handoff explains when shared equipment becomes live coordination. Scavenger separately keeps irregular listings, sellers, major commitments, parcels, and property diligence. Across presentation targets, Avatar follows the same principle: it groups presentation while Reach, Blockworld, Spectre, and Familiar retain their local work.
Media owners do not follow file extensions
A single engine or compound container can produce several roles. The role's acceptance determines its owner:
| Office | Owns | Does not inherit |
|---|---|---|
| Prism · Extension DomainPlug-in Area | visual/spatial effect contracts, technical result settlement, provenance, and derivative facts | a visual commission or creative acceptance |
| Echo · Extension DomainPlug-in Area | speech capture, STT/TTS, speech chronology, delivery, and playback facts | translation, casting, music, or application purpose |
| Atelier · CompositionApp | visual commission, direction, image, VFX, motion, and accepted visual package | music, timed-language editions, editorial cut, or publication |
| Riffmaw · CompositionApp | musical composition, instrumental and vocal performance, arrangement, mix/master, and musical cue map | ordinary speech, dubbing, standalone foley, or picture sound |
| Language Edition · CompositionApp | source-aligned language versions, translation/adaptation judgment, spoken performance, captions, dialogue conform, and restricted language-version packaging | Persona identity, music, editorial recut, or publication |
| Avatar · CompositionApp | Persona-linked presentation eligibility, Morphe selection, and projection bindings, including exact eligible voice references | Persona lineage, raw media, speech engines, or target-world authority |
| Foundry · CompositionApp | interactive world sound bound to exact game events, states, places, listeners, and runtime behavior | music, ordinary speech, picture-bound post-production, or a reusable universal sound library |
| Broadcast · CompositionApp | canonical source words and claims, picture-bound sound, final editorial timeline/render/mux, release, and correction | upstream visual, musical, voice, or provider truth |
A Translation SpellStep has an authority-qualified publisher; SpellweaverWorkflow Engine validates placement without taking linguistic or application truth. Compound model returns likewise keep one technical parent and independent visual/music/dialogue/picture/world-sound semantic admission. No generic reusable-sound CompositionApp is accepted; a future owner must prove its own lifecycle rather than hide in an existing craft.
An unresolved ownership need stops at study
If an external system already owns the needed truth, define an exact typed integration. If the state is bounded client UI or device-local presentation, leave it there. If LychD needs distinct durable application records, judgment, effects, finish, and recovery, prepare a candidate study.
Until an exact external authority or accepted CompositionApp revision resolves the need, affected use cases remain candidate or unsupported. A ProductPackaged Offer cannot temporarily own the missing truth or admit that path as supported.
How a leaf should read
Give the reader one representative journey, then the exact identity, PatternWorkflow catalogue, inputs, outcomes, records, handoffs, and stopping line needed to understand it. Classify foreign relations as settled reference, unmanaged external precondition, or live SuiteApp Bundle seam. Keep local authority/privacy/effect/recovery and the smallest falsifying fixture; link shared architecture and delivery instead of repeating them.
Native records, policies, and finish judgment enter src/lychd/compositions/<identity>/**; clients/<target>/** can project them without acquiring their authority. A stable key and separate CompositionApp revision, such as example.application revision 1, differs from PatternWorkflow example.perform_work@1. Workflow owns frozen revision, successor, and retirement law. State owns registration and execution evidence.
From study to matter
Product/use-case study exposes a need
→ candidate Composition tests the owner and defeat conditions
→ Portfolio acceptance establishes design truth
→ source enters only with an executable first-party contract
→ State of Work records what is actually delivered
A placeholder package cannot advance this sequence. There is no CryptData Folder compositions/ loader or Markdown discovery. State records CompositionApp registration, selection, scheduling, and execution evidence; its Loom entry retains the bounded source-registry projection.
Continue with Products and Suites, or return to the Portfolio.