promptkit/docs/roadmap/namespaces.md

Namespaces, Collections, And Scoped Resource Lookup

Status: feature roadmap; not implemented. The reserved-slash naming convention, reference-resolution rules, decisions 1–3 (scope shape, validation, and integrated discovery filtering), and decision 4's move to application/operator-supplied configuration are accepted target behavior. Decision 5 selects mounted collections and shared source-version checking; its remaining policy and API choices are tracked in the collection roadmap, subject to the ConfigKit boundary. Exact Go names and option types remain delivery details.

Purpose

Let consumers select a namespace once and use convenient local names within that namespace across PromptKit operations. An assembled engine may contain multiple application prompt libraries and provider profile catalogs with overlapping local names. Resolution must remain unambiguous.

The motivating cases are:

• WeatherReporter and another application both define summarize.alerts. Each consumer selects its application namespace and uses the local name.
• OpenRouter and RakestrawHome both provide the model identified in the discussion as google/gemma-4-31b-it. Consumers distinguish the execution profiles by provider namespace while retaining convenient local naming.

Model names here illustrate identity and resolution, not provider availability or a promise of maintained catalog entries.

The primary need is organization and convenient lookup within assembled libraries, with backend, profile, and prompt collections managed independently of PromptKit software releases. Ansible collections are a conceptual reference for qualified names; a package manager and relocatable collection imports are not required for this feature.

Existing Foundation

The format reference defines prompt, profile, and backend keys as exact literal IDs. Punctuation is already allowed, and references are not trimmed or re-keyed. Merely adopting prefixed IDs therefore does not provide the requested scoped lookup behavior.

Existing prompt and profile listings expose assembled identities. The consumer guide describes current operations; their Go declarations and GoDoc remain authoritative until this feature lands.

Directory-based and embedded configuration loading already exist through DirectoryInputs and FilesystemInputs. The current engine also prepends configuration assets from the imported OpenRouter and RakestrawHome Go modules, plus PromptKit-owned profiles. Decision 4 removes that automatic inclusion and dependency coupling; it does not require a replacement for ConfigKit loading.

ADR 0001 assigns source loading, composition, and inheritance to ConfigKit. This proposal adds interpretation of consumer-supplied resource names against the final effective configuration, without a second source-composition pass.

Accepted Identity Convention

Reserve / exclusively as the namespace separator in PromptKit prompt, profile, and backend identifiers and references. A qualified identifier has exactly two nonempty components: namespace/local-name. Neither component can contain /. Dots remain available within local names. An unqualified identifier has no slash and remains valid under the existing nonblank-ID rule.

Reject resource identifiers or supplied references with multiple slashes or an empty qualified component, such as a/b/c, /name, and namespace/. These rules do not apply to provider model values, endpoints, resource paths, or prompt versions. Preserve exact spelling; do not silently trim or rewrite IDs.

Resource	Namespace	Local name	Qualified ID
Prompt	weatherreporter	summarize.alerts	weatherreporter/summarize.alerts
Prompt	homewatch	summarize.alerts	homewatch/summarize.alerts
Profile	openrouter	google.gemma-4-31b-it	openrouter/google.gemma-4-31b-it
Profile	rakestrawhome	google.gemma-4-31b-it	rakestrawhome/google.gemma-4-31b-it

Prompt, profile, and backend identities remain separate resource types. The same namespace can contain all three, and resources in it may come from multiple configuration sources. A namespace does not imply a directory, package, provider model identifier, or security boundary.

A profile ID identifies execution settings. Its configured model value is still the identifier passed to the provider; qualifying the profile ID does not modify that value. Both example profiles may retain the configured model google/gemma-4-31b-it. Different profiles may select the same model with different settings. Prompt version labels remain separate from namespaces. The selected collection format installs one definition per prompt ID; optional request-version handling remains a decision in the collection roadmap.

Retain unqualified and dotted IDs. Existing slash-containing IDs need review: those with one slash and nonempty components become qualified identities, while those outside the new grammar require migration. This intentional format break is accepted in favor of unambiguous resolution; the migration work is described below.

Accepted Consumer Scope

Provide a lightweight, immutable view over an existing engine with independent prompt, profile, and backend scopes. The following sketch illustrates the accepted shape; method/type names and final signatures remain provisional:

weather, err := engine.WithNamespaces(Namespaces{
    Prompt:  "weatherreporter",
    Profile: "openrouter",
})
// Handle err, then use local prompt and profile names through weather.

This sketch follows decision 1; these declarations do not exist yet.

The view would share the engine's captured configuration, clients, and execution machinery. Selecting a namespace would not rebuild the engine or duplicate its backend capacity accounting. Consumers could use different views concurrently without changing a shared current-namespace setting.

The accepted resolution rules, independent of the final scope API, are:

1. A supplied name without a slash is qualified using the selected scope for its resource type, then looked up exactly. Without that scope, it is looked up as an exact unqualified ID.
2. A missing scoped resource does not trigger a search in other namespaces or among unqualified IDs. Loading another collection cannot redirect lookup.
3. A valid supplied name containing one slash is fully qualified and is looked up exactly, regardless of scope. Malformed resource references are rejected.
4. Scope applies to caller-supplied resource names. Stored references, including prompt default profiles, profile backends, inheritance references, and backend smoke-test default profiles, retain their existing exact meaning. They are not rebound to the invoking consumer's namespace.
5. Omitted optional IDs retain their existing selection/default behavior; an omission is not converted into a namespace prefix or a lookup for an empty local name.
6. Where outputs, inspection, errors, or diagnostic reports identify a selected resource, preserve its resolved full identity. Any exposed local name should be distinguishable from that identity.

Keep qualification consistent across execution, preparation, request validation, prompt/profile inspection, and applicable smoke-test selections. Namespace selection must not change request override precedence, the chosen request-version semantics, or provider model strings.

Local And Qualified References In One Operation

Syntax alone distinguishes local and qualified references. Do not decide whether to qualify a name by checking which candidates exist; that would make resolution depend on loaded content.

Within a weatherreporter prompt scope, summarize.alerts resolves only to weatherreporter/summarize.alerts, while homewatch/summarize.alerts explicitly selects that other library's prompt. A qualified profile override similarly bypasses the profile scope. Scoped operations accept both forms without a separate reference type or requiring a return to the original engine for cross-namespace access.

The original engine remains the way to select an exact unqualified ID without applying a scope. No escape syntax for unqualified IDs is introduced here.

Separate Resource Scopes

A consumer can select prompt shorthand under weatherreporter and profile shorthand under openrouter in the same request. Each resource type has an independent scope, as defined in decision 1. A prompt's configured default profile needs no caller profile scope because the stored reference is already exact. Scope syntax and existence handling are defined in decision 2.

Discovery And Collection Organization

Namespace enumeration and filtering are integrated into PromptKit's listing APIs. Consumers select a filter and receive matching full resource identities; they do not need to fetch all resources and filter them themselves. Defaults, explicit filter overrides, and result semantics are defined in decision 3.

Derive namespace membership from the qualified identifier's namespace component: weather must not include weatherreporter/.... Nested slash-separated namespaces are outside the accepted grammar.

Collection namespace/release metadata and an API for reporting effective collection declarations are desired additions. They are distinct from resource namespace enumeration. Their configuration shape and specification-version handling remain pending in decision 5.

Composition And Compatibility

Qualified identities reduce accidental collisions between independently authored libraries. Definitions with the same exact qualified identity still follow the existing ConfigKit overlay and inheritance rules; namespaces do not create a separate collision policy.

Leave source composition and configuration inspection with ConfigKit. Do not introduce implicit qualification of definition keys, mounting under caller-chosen prefixes, or rewriting of stored references as part of consumer scope. Those would be separate configuration-design work.

ConfigKit owns all configuration processing before handing off the final, fully resolved configuration. PromptKit must not add a collection-manifest preflight, source-level uniqueness check, or independent version-checking pass around that processing. Its domain interpretation and validation consume final effective values. Source history remains available through explicit ConfigKit inspection, not as a way to enforce extra assembly rules.

In particular, two documents contributing metadata for the same collection namespace are governed by ordinary ConfigKit composition semantics. ConfigKit may reject conflicts or combine contributions according to their layers; a repeated source declaration is not an additional PromptKit error. The earlier proposal to reject duplicate declared namespaces before composition is withdrawn.

Existing unscoped exact lookup remains available for IDs valid under the new grammar. Compatibility with arbitrary slash-containing resource IDs is intentionally broken. The migration must:

• Audit maintained catalogs and representative consumer definitions for invalid IDs and for valid one-slash IDs whose namespace meaning needs review.
• Choose explicit old-to-new mappings and check for collisions. Do not automatically replace slashes with dots: distinct existing IDs could collapse to the same new identity.
• Update renamed keys and their references together, including inheritance, prompt default profiles, profile backend references, backend smoke-test defaults, and affected examples and consumer request IDs.
• Keep provider model strings unchanged. Renaming a profile does not rename the upstream model.
• Document consumer action and coordinate the PromptKit format change with catalog adoption under the release policy. Its catalog profile-data exception does not cover this general format/API behavior change; deliberately retire its built-in-profile preservation promise when the accepted removal lands. Current policy and format documents remain unchanged until implementation.

Renaming a backend changes its identity; duplicating it under another ID can separate capacity accounting. Do not use blanket aliases or duplicate catalog entries as an automatic compatibility mechanism. Exact release versions and catalog mappings remain implementation-planning work.

The architecture policy continues to assign CLI syntax, flags, and application policy to consumers such as Scriptorium. PromptKit would supply reusable scope and resolution semantics.

Work Not Included In The Initial Proposal

• Package installation, collection registries, and dependency resolution.
• Caller-selected collection mount points or import alias rewriting.
• Implicit namespace search paths or fallback to whichever resource exists.
• Namespace-based access control or filesystem isolation.
• Changes to model names, inference behavior, or configuration precedence.

These boundaries limit the initial proposal; they do not prohibit future work if a concrete consumer need emerges.

Delivery Outline

1. Resolve decision 5's ConfigKit integration question and audit existing identifier usage against the accepted grammar. Record migration mappings, application provisioning guidance, and any separately required ConfigKit work. Finalize Go names and option types for accepted decisions 1–3.
2. Remove automatic catalog inclusion, catalog-module dependencies, and built-in profiles under decision 4. Use caller-selected configuration sources and update examples/tests to supply their own definitions.
3. Implement the smallest public scope surface with shared name qualification and existing engine collaborators. Cover supported operations consistently.
4. Add integrated namespace filtering and representative consumer examples. Implement collection metadata/inventory only after decision 5 is resolved.
5. Verify compatibility and scoped resolution through focused offline tests, following the testing policy. Update GoDoc, the format reference, consumer guide, and relevant internal documentation as behavior lands; retire this roadmap when its work is complete.

Acceptance cases for the settled rules; collection metadata/versioning cases follow the eventual decision 5:

• Two prompt scopes resolve the same local name to their respective prompts; a missing name never resolves from the other scope or an unqualified entry.
• Two provider profile scopes resolve the same local name, such as google.gemma-4-31b-it, to different profiles without changing the configured provider model value google/gemma-4-31b-it.
• A qualified reference bypasses the selected scope without double qualification; adding another collection does not change local or qualified resolution.
• Multiple slashes and empty qualified components are rejected in resource IDs and references, while provider model values retain their existing syntax.
• Prompt shorthand and a profile override use their intended scopes, while stored default references and omitted selections retain their meaning.
• Scoped views coexist concurrently and share backend capacity with ordinary calls on the underlying engine.
• Scope creation rejects invalid syntax, accepts well-formed absent namespaces, and treats empty scope fields as no scope. Missing resources fail at lookup; qualified references remain usable even when the default namespace is absent.
• Listing without an explicit filter uses the resource's view scope, or lists all resources when no scope is selected. Explicit named, unqualified-only, and all-namespaces filters override that default for the current listing only.
• A valid namespace with no matches produces an empty listing. Namespace filtering respects resource type and exact namespace boundaries.
• Exact-ID calls using valid IDs and intentional overlays retain their behavior. Migrated catalog references resolve without collisions; the collection roadmap owns migration to one installed prompt and optional version labels.
• Listings preserve full IDs and separate optional prompt version labels. Qualified results round-trip through either engine or view; unqualified results round-trip through the engine without a corresponding scope.
• Representative validation, inspection, execution, and smoke-test paths use the same qualification rules without duplicating every test at every layer.
• Engines use caller-selected definitions without automatically contributing provider catalogs or PromptKit-owned profiles. Ordinary directory and embedded configuration sources retain ConfigKit composition semantics.
• PromptKit does not add source-history-based duplicate-namespace rejection or independently pre-validate collection definitions before the final handoff.

Decisions And Pending Design

Decisions 1–3 are accepted; their alternatives are retained as considered but not selected. Decision 4 records the accepted change in collection ownership and supersedes the earlier catalog-only migration proposal. Decision 5 retains the unresolved metadata/versioning question and its alternatives.

1. Public Scope Shape (Accepted)

Accepted: an immutable view with independent resource scopes. Provide one engine operation that accepts a small value with prompt, profile, and backend namespace fields and returns a scoped view or an error. The illustrative WithNamespaces and Namespaces names above are provisional. Empty fields mean no scope for that resource type. The view captures the value and delegates to the same engine; it does not expose a mutable namespace setting. Apply the prompt field to caller-supplied prompt IDs, the profile field to profile IDs, and the backend field to backend IDs wherever those selections are supported. Qualified references bypass the corresponding field.

This directly serves the select-once workflow and the common mixed case of an application prompt namespace with a provider profile namespace. It also keeps scope out of stored configuration and makes concurrent use independent of call order. The cost is an additional facade surface whose supported operations must remain consistent with the engine. Keep qualification shared rather than duplicating selection logic in each wrapper.

Initially omit an ambiguous one-string InNamespace shorthand. Repeating the same namespace in several fields happens once when creating a view and states exactly which resource types it affects. A convenience operation can follow if that repetition is a demonstrated burden.

Considered alternative, not selected: one namespace per immutable view, applied to all resource types. A one-string operation would scope every unqualified prompt, profile, and backend selection; callers would use qualified references for resources in other namespaces. This is smaller and reads naturally for self-contained application collections. It is less convenient when prompts and profiles habitually come from different namespaces, because profile overrides would repeatedly need qualification. Prefer it if consumers primarily use prompts' configured defaults and rarely select profiles themselves.

Considered alternative, not selected: explicit scope values supplied per operation. Add scope to request values and appropriate inspection/listing options instead of returning a view. This makes every invocation self-contained and suits applications that choose a different scope for every request. It spreads scope through more call sites and API shapes, and consumers would need their own reusable wrapper to achieve the requested select-once experience. The immutable view is therefore the better default for the use case discussed so far.

2. Scope Validation (Accepted)

Accepted: validate syntax when creating the view; resolve existence when an operation selects a resource. In the scope value, the empty string means no scope. Reject a nonempty namespace containing / or consisting only of whitespace. Otherwise preserve its exact spelling, with no trimming, case folding, or additional character whitelist. This distinguishes omission from an unusable namespace without inventing normalization rules. Report invalid scope construction through PromptKit's existing public error conventions.

Allow a well-formed namespace even when it has no current members. Its scoped listing would be empty, and selection of a missing local resource would produce the ordinary lookup error. Qualified selections would remain usable regardless of the view's default scope. An empty optional request ID continues to mean omission under that operation's existing contract; it is never qualified.

This keeps scope as a naming context and supports uniform application setup across installations with different library contents. It also avoids treating successful scope construction as validation of a prompt or execution profile. The tradeoff is that a misspelled but syntactically valid namespace is detected only when listed or used, rather than during view construction.

Considered alternative, not selected: require each selected namespace to have resources of its corresponding type when creating the view. For an immutable assembled engine, this can catch spelling mistakes immediately and gives startup checks useful feedback without attempting execution. However, it rejects empty optional collections and a view intended to use qualified references outside its default namespace. A namespace having one member also does not establish that the particular required prompt/profile exists. Prefer this approach only if eager inventory validation is itself an intended consumer contract; it should not be an accidental consequence of implementing scope creation.

3. Discovery (Accepted)

Accepted: integrated namespace filtering and full-identity results. Enumerate the distinct named namespaces present in final effective prompts, profiles, and backends. Allow filtering that enumeration by resource type so a prompt picker need not show profile-only namespaces. Provide scoped listings for the supported resource types, adding a backend inventory surface as part of this work if needed. Namespace filtering is an explicit option in the library listing API, not a consumer-side filtering helper. Preserve existing unfiltered engine calls and their result contracts; the exact Go option shape remains to be specified.

Selection	Listing behavior
No explicit filter on the engine	All namespaces, including unqualified resources
No explicit filter on a scoped view	The selected scope for that resource type; all resources of that type if no scope is selected
Explicit named namespace	Resources in exactly that namespace
Explicit unqualified-only filter	Resources whose IDs have no namespace
Explicit all-namespaces filter	All resources of that type, including unqualified resources

An explicit filter replaces the view's default for the current listing; it does not intersect with that default or mutate the view. A named namespace filter uses the namespace syntax rules in decision 2. A well-formed namespace with no matching resources returns an empty listing. Represent default, named, unqualified-only, and all-namespaces selections distinctly in the API; do not overload an empty string to mean several of them.

Return full IDs with optional prompt version labels separately; the full ID identifies the installed prompt. Do not invent a named namespace such as default for unqualified resources. Namespace enumeration excludes that unnamed group. Use deterministic ordering consistent with existing listing contracts.

Qualified result IDs can be passed unchanged to either scoped or unscoped operations. Unqualified IDs remain literal unqualified identities: when an explicit all-namespaces or unqualified-only filter exposes one through a scoped view, use the original engine or a view without that resource scope to select it exactly. A listing filter does not change subsequent execution scope. This preserves the accepted lookup rules without introducing an escape syntax.

Integrated filtering gives consumers a direct way to discover the desired resources, while full identities keep logs and subsequent selections unambiguous. Namespace/local-name display decomposition is separate from filtering; a public splitting helper is not required by this decision.

Considered alternative, not selected: add a richer inventory API returning full ID, namespace, local name, resource type, and prompt version where applicable. Preserve existing list methods and expose the structured form separately. This is attractive for Scriptorium-style library browsers: grouping and display require no parsing, and one inventory operation can cover all resource kinds. It creates another public representation and overlaps existing listing APIs, so its value should justify that maintenance. Prefer it if downstream clients routinely need a combined catalog view rather than a resource picker or namespace filter. In either approach, the full ID remains the authoritative selection value.

4. Collection Ownership And Built-In Removal (Accepted)

Accepted: load collections as application/operator-supplied configuration, and remove PromptKit's built-in profiles as a breaking change. Remove the mandatory OpenRouter and RakestrawHome catalog Go-module dependencies and their automatic contribution to engine configuration. PromptKit owns the definition format and execution machinery; collection publishers own their configuration repositories, and applications/operators select and provision the collections they use. One or more explicitly selected directories can supply backends, profiles, and prompts through the existing ConfigKit integration.

Remove promptkit-openrouter-standard and promptkit-openrouter-auto, including their implicit catalog dependencies. Do not replace them with another mandatory starter bundle. Applications may embed their own defaults using the same configuration format and loader, then layer operator configuration as desired. Ordinary application overrides remain valid configuration and need not declare a collection release of their own.

This decouples collection updates from PromptKit rebuilds and releases and avoids imposing provider inventories on every consumer. The tradeoff is that applications and deployments must supply their required configuration and manage compatible collection revisions. PromptKit does not gain a collection installer, automatic downloader, or implicit directory search policy.

Retain the captured-engine lifecycle: configuration changes take effect when the application constructs and validates a replacement engine. A change to an installed directory does not mutate an existing engine's captured definitions. Distribution as configuration does not make new provider protocols or diagnostic adapters data-only features; those can still require implementation work.

Prepare explicit identifier mappings, replacement configuration examples, and consumer migration instructions. Update the built-in-profile promise in the release policy and its canonical format documentation when removal lands. Use a release permitted to break configuration and API contracts under the current release policy: a minor release before v1, or a major release at/after v1. Actual catalog mappings, packaging arrangements, and release versions remain delivery work; independent collection releases need not track future PromptKit software releases.

Superseded alternatives: retaining catalog Go modules with renamed profiles, or retaining a small PromptKit-owned starter bundle, would simplify initial provisioning but preserve some provider-data coupling. The accepted direction places optional defaults in consuming applications and removes those core dependencies instead.

5. Collection Metadata And Specification Versioning (Open)

Selected direction: mounted collections. Proposal 3 is the active configuration roadmap. It uses collection-local mapping keys, one installed definition per prompt, and an optional flat prompt version. Input destinations, relative imports, and shared source-version checks are implemented in ConfigKit's reviewed development branch. PromptKit still needs to adopt a suitable dependency release and implement its new schema.

Use shared specification checking with IfPresent and distinct collection release_version metadata. Collection entries can describe their layout through relative imports; applications select their sources and namespaces. PromptKit interprets the final collection/resource keys as qualified IDs after ConfigKit's handoff. It adds no manifest pre-parser, source-level namespace ownership rule, or history-based validation pass. Stored references remain exact public IDs.

Proposals 1 and 2 were not selected and have been retired. Their original contents and the separately preserved ConfigKit identified-record request are recorded in the collection roadmap's retirement note. The record-sequence request is independent future upstream work and does not block the selected mapping design.

The collection roadmap owns the candidate schema, examples, version policies, metadata inventory, and migration requirements. Inventory describes effective collection declarations captured by the engine, not historical manifests or proof of unchanged published releases. Source explanation remains ConfigKit's inspection concern.

Still open: the six remaining PromptKit decisions cover main declaration requirements, exposing stricter checking, resources named version, request-version assertions, cross-collection inheritance, and consumer APIs/packaging. Selecting proposal 3 and retiring its alternatives does not settle those choices. Removal of bundled catalogs and built-in profiles remains accepted.