@oh-my-pi/pi-coding-agent 15.10.9 → 15.10.11

package/dist/types/async/job-manager.d.ts

@@ -15,6 +15,12 @@ export interface AsyncJob {
      * supply an id (e.g. legacy tests, SDK consumers without an agent context).
      */
     ownerId?: string;
+    /**
+     * Job is registered but parked behind a caller-managed gate (e.g. a task
+     * batch semaphore). Queued jobs do not count toward the running-job limit
+     * until the caller invokes `markRunning()` from the run context.
+     */
+    queued?: boolean;
 }
 export interface AsyncJobManagerOptions {
     onJobComplete: (jobId: string, text: string, job?: AsyncJob) => void | Promise<void>;
@@ -32,6 +38,8 @@ export interface AsyncJobRegisterOptions {
     /** Registry id of the agent that owns this job; used to scope cancelAll. */
     ownerId?: string;
     onProgress?: (text: string, details?: Record<string, unknown>) => void | Promise<void>;
+    /** Register the job in queued state; see {@link AsyncJob.queued}. */
+    queued?: boolean;
 }
 /**
  * Filter applied to job query/cancel APIs. With `ownerId`, results are
@@ -50,10 +58,14 @@ export declare class AsyncJobManager {
     /** Reset the process-global instance. Test-only. */
     static resetForTests(): void;
     constructor(options: AsyncJobManagerOptions);
+    /** True when the running-job count has reached the configured cap. */
+    get atCapacity(): boolean;
     register(type: "bash" | "task", label: string, run: (ctx: {
         jobId: string;
         signal: AbortSignal;
         reportProgress: (text: string, details?: Record<string, unknown>) => Promise<void>;
+        /** Clear the queued flag once the job actually starts executing. */
+        markRunning: () => void;
     }) => Promise<string>, options?: AsyncJobRegisterOptions): string;
     /**
      * Cancel a single job by id. When `filter.ownerId` is set and does not
@@ -70,6 +82,12 @@ export declare class AsyncJobManager {
     watchJobs(jobIds: string[]): number;
     unwatchJobs(jobIds: string[]): number;
     acknowledgeDeliveries(jobIds: string[]): number;
+    /**
+     * Lift a foreground-wait suppression set via `acknowledgeDeliveries`. If the
+     * job already finished while suppressed (its delivery enqueue was skipped),
+     * re-enqueue the completion so the result is still delivered exactly once.
+     */
+    resumeDeliveries(jobIds: string[]): void;
     /**
      * Cancel running jobs. With `filter.ownerId` set, cancels only jobs the
      * matching agent registered; with no filter, cancels every running job

package/dist/types/cli/args.d.ts

@@ -1,7 +1,7 @@
 /**
  * CLI argument parsing and help display
  */
-import { type Effort } from "@oh-my-pi/pi-ai/effort";
+import { type Effort } from "@oh-my-pi/pi-catalog/effort";
 export type Mode = "text" | "json" | "rpc" | "acp" | "rpc-ui";
 export interface Args {
     cwd?: string;

package/dist/types/cli/dry-balance-cli.d.ts

@@ -1,5 +1,5 @@
 import type { Api, AssistantMessageEventStream, AuthCredentialSnapshotEntry, Context, Model, OAuthAccess, OAuthAccessResolution, SimpleStreamOptions } from "@oh-my-pi/pi-ai";
-import type { CanonicalModelVariant } from "../config/model-equivalence";
+import type { CanonicalModelVariant } from "@oh-my-pi/pi-catalog/identity";
 import { type CanonicalModelQueryOptions } from "../config/model-registry";
 import { Settings } from "../config/settings";
 export interface DryBalanceCommandArgs {

package/dist/types/cli/gallery-cli.d.ts

@@ -34,7 +34,7 @@ export declare function resolveFixture(name: string): GalleryFixture;
  * to the requested state, settles any async edit preview, then snapshots the
  * render and stops all animation timers.
  */
-export declare function renderGalleryState(name: string, fixture: GalleryFixture, state: GalleryState, width: number, expanded?: boolean): Promise<string[]>;
+export declare function renderGalleryState(name: string, fixture: GalleryFixture, state: GalleryState, width: number, expanded?: boolean): Promise<readonly string[]>;
 /**
  * Render the gallery. Iterates the renderer registry (or a single tool),
  * printing each requested lifecycle state under a labeled section — or, with

package/dist/types/cli/gallery-fixtures/types.d.ts

@@ -24,7 +24,7 @@ export interface GalleryFixture {
      * Custom gallery-only renderer for fixtures that are not one ToolExecutionComponent
      * (for example the read-group transcript component).
      */
-    renderState?: (state: GalleryFixtureState, width: number, expanded: boolean) => string[] | Promise<string[]>;
+    renderState?: (state: GalleryFixtureState, width: number, expanded: boolean) => readonly string[] | Promise<readonly string[]>;
     /**
      * Set for tools whose real `AgentTool` attaches `renderCall`/`renderResult`
      * directly on the instance (e.g. `task`). The harness then attaches

package/dist/types/cli/usage-cli.d.ts

@@ -0,0 +1,72 @@
+/**
+ * Usage CLI command handler.
+ *
+ * Handles `omp usage` — fetches provider usage reports for every
+ * authenticated account and prints a detailed per-account breakdown
+ * (limits, windows, reset times, plan metadata). Accounts whose
+ * credentials produced no usage report are listed too, so the output
+ * always covers the full credential pool.
+ */
+import type { UsageReport } from "@oh-my-pi/pi-ai";
+export interface UsageCommandArgs {
+    json?: boolean;
+    provider?: string;
+    redact?: boolean;
+}
+/** Identity slice of a stored credential, for "every account" coverage. */
+export interface UsageAccountIdentity {
+    provider: string;
+    type: "api_key" | "oauth";
+    email?: string;
+    accountId?: string;
+    projectId?: string;
+    enterpriseUrl?: string;
+}
+/**
+ * Minimal-reveal masks for identity strings (`--redact`).
+ *
+ * Every mask shows a two-character anchor. When two identities share the
+ * anchor, the mask additionally reveals the shortest "middle-out"
+ * differentiator — the shortest substring (closest to the string's middle on
+ * ties) that no colliding identity contains — as `an*`, `ca*9*`, `ca*nb*`.
+ * Prefix growth is deliberately avoided: it leaks the start of the local
+ * part (`can.boluk@*`) when a couple of mid-string characters suffice.
+ * Duplicate strings (same account on two providers) share a mask.
+ */
+export declare function buildRedactionMap(values: Iterable<string>): Map<string, string>;
+/**
+ * Stored credentials that no usage report could be attributed to.
+ *
+ * Conservative on purpose: when a provider's reports carry no identity at
+ * all (or the credential is an API key alongside existing reports), we
+ * can't attribute, so we don't claim the account is missing.
+ */
+export declare function collectUnreportedAccounts(reports: UsageReport[], accounts: UsageAccountIdentity[]): UsageAccountIdentity[];
+/** Per-window capacity stat: how much account quota is burned and left. */
+export interface ProviderWindowStat {
+    /** Compact window label, e.g. "5h", "7d". */
+    window: string;
+    durationMs?: number;
+    /** Accounts reporting a limit in this window. */
+    accounts: number;
+    /** Sum of each account's binding used fraction — accounts' worth of quota burned. */
+    usedAccounts: number;
+    /** Accounts' worth of quota still available across reporting accounts. */
+    remainingAccounts: number;
+}
+/**
+ * Aggregate one provider's reports into per-window quota capacity stats.
+ *
+ * Limits are bucketed by window duration (5h, 7d, ...). Within a bucket each
+ * account contributes its single highest used fraction — when an account has
+ * several meters on the same window (tiered/metered limits), the most-burned
+ * one is what binds.
+ */
+export declare function computeProviderWindowStats(reports: UsageReport[]): ProviderWindowStat[];
+/**
+ * Render the full text breakdown: per provider, per account, every limit
+ * with a bar, amounts, and reset times; unattributed credentials trail
+ * each provider section as "no usage data" rows.
+ */
+export declare function formatUsageBreakdown(reports: UsageReport[], accounts: UsageAccountIdentity[], nowMs: number, redaction?: Map<string, string>): string;
+export declare function runUsageCommand(cmd: UsageCommandArgs): Promise<void>;

package/dist/types/commands/launch.d.ts

@@ -82,7 +82,7 @@ export default class Index extends Command {
         };
         thinking: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"string"> & {
             description: string;
-            options: import("@oh-my-pi/pi-ai").Effort[];
+            options: import("@oh-my-pi/pi-catalog").Effort[];
         };
         "hide-thinking": import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"boolean"> & {
             description: string;

package/dist/types/commands/read.d.ts

@@ -1,5 +1,5 @@
 /**
- * Show what the read tool will return for a given path.
+ * Show what the read tool will return for a path, URL, or internal URI.
  */
 import { Command } from "@oh-my-pi/pi-utils/cli";
 export default class Read extends Command {

package/dist/types/commands/usage.d.ts

@@ -0,0 +1,25 @@
+/**
+ * Show provider usage limits for every authenticated account.
+ */
+import { Command } from "@oh-my-pi/pi-utils/cli";
+export default class Usage extends Command {
+    static description: string;
+    static flags: {
+        json: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"boolean"> & {
+            char: string;
+            description: string;
+            default: boolean;
+        };
+        provider: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"string"> & {
+            char: string;
+            description: string;
+        };
+        redact: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"boolean"> & {
+            char: string;
+            description: string;
+            default: boolean;
+        };
+    };
+    static examples: string[];
+    run(): Promise<void>;
+}

package/dist/types/config/append-only-context-mode.d.ts

@@ -2,7 +2,8 @@
 export interface AppendOnlyContextModel {
     provider: string;
     baseUrl: string;
-    compat?: object;
+    /** Verbatim sparse compat config (explicit user intent), never the resolved record. */
+    compatConfig?: object;
 }
 /** Resolves whether append-only context should be active for a model and setting. */
 export declare function shouldEnableAppendOnlyContext(setting: "auto" | "on" | "off" | undefined, model: AppendOnlyContextModel | null | undefined): boolean;

package/dist/types/config/model-discovery.d.ts

@@ -0,0 +1,55 @@
+/**
+ * HTTP discovery protocols for configured and implicit providers — ollama,
+ * llama.cpp, lm-studio, openai-models-list, and new-api/one-api-style proxies.
+ * `ModelRegistry` owns the orchestration (status, state, caching) and calls
+ * `discoverModelsByProviderType` with a `DiscoveryContext`; built-in provider
+ * discovery lives in pi-catalog's provider-models.
+ */
+import type { FetchImpl } from "@oh-my-pi/pi-ai";
+import type { Api, Model } from "@oh-my-pi/pi-ai/types";
+import type { ModelSpec } from "@oh-my-pi/pi-catalog/types";
+import type { ProviderDiscovery } from "./models-config-schema";
+export declare const DISCOVERY_DEFAULT_MAX_TOKENS = 32768;
+export declare function getImplicitOllamaBaseUrl(): string;
+export declare function getOllamaContextLengthOverride(): number | undefined;
+/** Routes discovered-model `maxTokens` defaults around Anthropic's 3× output divisor. */
+export declare function discoveryDefaultMaxTokens(api: Api | undefined): number;
+export interface DiscoveryProviderConfig {
+    provider: string;
+    api: Api;
+    baseUrl?: string;
+    headers?: Record<string, string>;
+    compat?: ModelSpec<Api>["compat"];
+    discovery: ProviderDiscovery;
+    optional?: boolean;
+}
+/** Registry-provided capabilities the protocol probes need; never the registry itself. */
+export interface DiscoveryContext {
+    /** Injected fetch implementation (tests stub this). */
+    fetch: FetchImpl;
+    /**
+     * Resolve a provider's API key for `Authorization: Bearer …`. Returns
+     * undefined when no key is stored or it is a local/no-auth sentinel.
+     */
+    getBearerApiKey(provider: string): Promise<string | undefined>;
+}
+export declare function discoverModelsByProviderType(providerConfig: DiscoveryProviderConfig, ctx: DiscoveryContext): Promise<Model<Api>[]>;
+export declare function discoverOllamaModels(providerConfig: DiscoveryProviderConfig, ctx: DiscoveryContext): Promise<Model<Api>[]>;
+export declare function discoverLlamaCppModels(providerConfig: DiscoveryProviderConfig, ctx: DiscoveryContext): Promise<Model<Api>[]>;
+export declare function discoverOpenAIModelsList(providerConfig: DiscoveryProviderConfig, ctx: DiscoveryContext): Promise<Model<Api>[]>;
+/**
+ * Discover models from an Anthropic+OpenAI-compatible reseller proxy that
+ * exposes both `/v1/messages` and `/v1/chat/completions`, advertising each
+ * model's wire capabilities through `supported_endpoint_types` on
+ * `GET /v1/models` (new-api / one-api-style proxies).
+ *
+ * Routing per model:
+ *   supported_endpoint_types: ["anthropic", ...] -> api: "anthropic-messages"
+ *   supported_endpoint_types: ["openai"]         -> api: "openai-completions"
+ *   missing / neither                            -> provider-level api fallback
+ *
+ * Anthropic models share the same baseUrl; the Anthropic SDK strips a
+ * trailing `/v1` itself before appending `/v1/messages`, so the discovery
+ * URL (which ends in `/v1`) round-trips correctly.
+ */
+export declare function discoverProxyModels(providerConfig: DiscoveryProviderConfig, ctx: DiscoveryContext): Promise<Model<Api>[]>;

package/dist/types/config/model-registry.d.ts

@@ -1,234 +1,22 @@
-import { type ModelRefreshStrategy } from "@oh-my-pi/pi-ai/model-manager";
-import type { Api, Context, Model, SimpleStreamOptions, ThinkingConfig } from "@oh-my-pi/pi-ai/types";
+import type { Api, Context, Model, ModelSpec, SimpleStreamOptions, ThinkingConfig } from "@oh-my-pi/pi-ai/types";
 import type { AssistantMessageEventStream } from "@oh-my-pi/pi-ai/utils/event-stream";
+import { type ModelRefreshStrategy } from "@oh-my-pi/pi-catalog/model-manager";
 import type { ApiKeyResolver, FetchImpl } from "@oh-my-pi/pi-ai";
 import type { OAuthCredentials, OAuthLoginCallbacks } from "@oh-my-pi/pi-ai/oauth/types";
-import { type ThemeColor } from "../modes/theme/theme";
+import { type CanonicalModelIndex, type CanonicalModelRecord, type CanonicalModelVariant, type ModelEquivalenceConfig } from "@oh-my-pi/pi-catalog/identity";
 import type { AuthStorage } from "../session/auth-storage";
 import { type ApiKeyResolverOptions } from "./api-key-resolver";
-import { type ConfigError, ConfigFile } from "./config-file";
-import { type CanonicalModelIndex, type CanonicalModelRecord, type CanonicalModelVariant, type ModelEquivalenceConfig } from "./model-equivalence";
-import { type Settings } from "./settings";
+import type { ConfigError } from "./config-file";
 export type { CanonicalModelIndex, CanonicalModelRecord, CanonicalModelVariant, ModelEquivalenceConfig };
 export declare const kNoAuth = "N/A";
 export declare function isAuthenticated(apiKey: string | undefined | null): apiKey is string;
-export type ModelRole = "default" | "smol" | "slow" | "vision" | "plan" | "designer" | "commit" | "task";
-export interface ModelRoleInfo {
-    tag?: string;
-    name: string;
-    color?: ThemeColor;
-}
-export declare const MODEL_ROLES: Record<ModelRole, ModelRoleInfo>;
-export declare const MODEL_ROLE_IDS: ModelRole[];
-/** Alias for ModelRoleInfo - used for both built-in and custom roles */
-export type RoleInfo = ModelRoleInfo;
-/**
- * Return the canonical set of known roles for selector/carousel UI.
- *
- * Built-ins always come first. Configured cycle order, model assignments, and
- * tag metadata can introduce additional custom roles without requiring duplicate
- * entries across settings.
- */
-export declare function getKnownRoleIds(settings: Settings): string[];
-/**
- * Get role info for a role name (built-in or custom).
- * Configured metadata overrides built-in defaults when present.
- */
-export declare function getRoleInfo(role: string, settings: Settings): RoleInfo;
-export declare const ModelsConfigFile: ConfigFile<{
-    providers?: Record<string, {
-        baseUrl?: string | undefined;
-        apiKey?: string | undefined;
-        api?: "anthropic-messages" | "azure-openai-responses" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
-        headers?: Record<string, string> | undefined;
-        compat?: {
-            supportsStore?: boolean | undefined;
-            supportsDeveloperRole?: boolean | undefined;
-            supportsMultipleSystemMessages?: boolean | undefined;
-            supportsReasoningEffort?: boolean | undefined;
-            reasoningEffortMap?: {
-                minimal?: string | undefined;
-                low?: string | undefined;
-                medium?: string | undefined;
-                high?: string | undefined;
-                xhigh?: string | undefined;
-            } | undefined;
-            maxTokensField?: "max_completion_tokens" | "max_tokens" | undefined;
-            supportsUsageInStreaming?: boolean | undefined;
-            requiresToolResultName?: boolean | undefined;
-            requiresMistralToolIds?: boolean | undefined;
-            requiresAssistantAfterToolResult?: boolean | undefined;
-            requiresThinkingAsText?: boolean | undefined;
-            reasoningContentField?: "reasoning" | "reasoning_content" | "reasoning_text" | undefined;
-            requiresReasoningContentForToolCalls?: boolean | undefined;
-            allowsSyntheticReasoningContentForToolCalls?: boolean | undefined;
-            requiresAssistantContentForToolCalls?: boolean | undefined;
-            supportsToolChoice?: boolean | undefined;
-            disableReasoningOnForcedToolChoice?: boolean | undefined;
-            disableReasoningOnToolChoice?: boolean | undefined;
-            thinkingFormat?: "openai" | "openrouter" | "qwen" | "qwen-chat-template" | "zai" | undefined;
-            openRouterRouting?: {
-                only?: string[] | undefined;
-                order?: string[] | undefined;
-            } | undefined;
-            vercelGatewayRouting?: {
-                only?: string[] | undefined;
-                order?: string[] | undefined;
-            } | undefined;
-            extraBody?: Record<string, unknown> | undefined;
-            cacheControlFormat?: "anthropic" | undefined;
-            supportsStrictMode?: boolean | undefined;
-            toolStrictMode?: "all_strict" | "none" | undefined;
-        } | undefined;
-        authHeader?: boolean | undefined;
-        auth?: "apiKey" | "none" | "oauth" | undefined;
-        discovery?: {
-            type: "llama.cpp" | "lm-studio" | "ollama" | "openai-models-list" | "proxy";
-        } | undefined;
-        models?: {
-            id: string;
-            name?: string | undefined;
-            api?: "anthropic-messages" | "azure-openai-responses" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
-            baseUrl?: string | undefined;
-            reasoning?: boolean | undefined;
-            thinking?: {
-                minLevel: "high" | "low" | "medium" | "minimal" | "xhigh";
-                maxLevel: "high" | "low" | "medium" | "minimal" | "xhigh";
-                mode: "anthropic-adaptive" | "anthropic-budget-effort" | "budget" | "effort" | "google-level";
-                defaultLevel?: "high" | "low" | "medium" | "minimal" | "xhigh" | undefined;
-                levels?: ("high" | "low" | "medium" | "minimal" | "xhigh")[] | undefined;
-            } | undefined;
-            input?: ("image" | "text")[] | undefined;
-            cost?: {
-                input: number;
-                output: number;
-                cacheRead: number;
-                cacheWrite: number;
-            } | undefined;
-            premiumMultiplier?: number | undefined;
-            contextWindow?: number | undefined;
-            maxTokens?: number | undefined;
-            omitMaxOutputTokens?: boolean | undefined;
-            headers?: Record<string, string> | undefined;
-            compat?: {
-                supportsStore?: boolean | undefined;
-                supportsDeveloperRole?: boolean | undefined;
-                supportsMultipleSystemMessages?: boolean | undefined;
-                supportsReasoningEffort?: boolean | undefined;
-                reasoningEffortMap?: {
-                    minimal?: string | undefined;
-                    low?: string | undefined;
-                    medium?: string | undefined;
-                    high?: string | undefined;
-                    xhigh?: string | undefined;
-                } | undefined;
-                maxTokensField?: "max_completion_tokens" | "max_tokens" | undefined;
-                supportsUsageInStreaming?: boolean | undefined;
-                requiresToolResultName?: boolean | undefined;
-                requiresMistralToolIds?: boolean | undefined;
-                requiresAssistantAfterToolResult?: boolean | undefined;
-                requiresThinkingAsText?: boolean | undefined;
-                reasoningContentField?: "reasoning" | "reasoning_content" | "reasoning_text" | undefined;
-                requiresReasoningContentForToolCalls?: boolean | undefined;
-                allowsSyntheticReasoningContentForToolCalls?: boolean | undefined;
-                requiresAssistantContentForToolCalls?: boolean | undefined;
-                supportsToolChoice?: boolean | undefined;
-                disableReasoningOnForcedToolChoice?: boolean | undefined;
-                disableReasoningOnToolChoice?: boolean | undefined;
-                thinkingFormat?: "openai" | "openrouter" | "qwen" | "qwen-chat-template" | "zai" | undefined;
-                openRouterRouting?: {
-                    only?: string[] | undefined;
-                    order?: string[] | undefined;
-                } | undefined;
-                vercelGatewayRouting?: {
-                    only?: string[] | undefined;
-                    order?: string[] | undefined;
-                } | undefined;
-                extraBody?: Record<string, unknown> | undefined;
-                cacheControlFormat?: "anthropic" | undefined;
-                supportsStrictMode?: boolean | undefined;
-                toolStrictMode?: "all_strict" | "none" | undefined;
-            } | undefined;
-            contextPromotionTarget?: string | undefined;
-        }[] | undefined;
-        modelOverrides?: Record<string, {
-            name?: string | undefined;
-            reasoning?: boolean | undefined;
-            thinking?: {
-                minLevel: "high" | "low" | "medium" | "minimal" | "xhigh";
-                maxLevel: "high" | "low" | "medium" | "minimal" | "xhigh";
-                mode: "anthropic-adaptive" | "anthropic-budget-effort" | "budget" | "effort" | "google-level";
-                defaultLevel?: "high" | "low" | "medium" | "minimal" | "xhigh" | undefined;
-                levels?: ("high" | "low" | "medium" | "minimal" | "xhigh")[] | undefined;
-            } | undefined;
-            input?: ("image" | "text")[] | undefined;
-            cost?: {
-                input?: number | undefined;
-                output?: number | undefined;
-                cacheRead?: number | undefined;
-                cacheWrite?: number | undefined;
-            } | undefined;
-            premiumMultiplier?: number | undefined;
-            contextWindow?: number | undefined;
-            maxTokens?: number | undefined;
-            omitMaxOutputTokens?: boolean | undefined;
-            headers?: Record<string, string> | undefined;
-            compat?: {
-                supportsStore?: boolean | undefined;
-                supportsDeveloperRole?: boolean | undefined;
-                supportsMultipleSystemMessages?: boolean | undefined;
-                supportsReasoningEffort?: boolean | undefined;
-                reasoningEffortMap?: {
-                    minimal?: string | undefined;
-                    low?: string | undefined;
-                    medium?: string | undefined;
-                    high?: string | undefined;
-                    xhigh?: string | undefined;
-                } | undefined;
-                maxTokensField?: "max_completion_tokens" | "max_tokens" | undefined;
-                supportsUsageInStreaming?: boolean | undefined;
-                requiresToolResultName?: boolean | undefined;
-                requiresMistralToolIds?: boolean | undefined;
-                requiresAssistantAfterToolResult?: boolean | undefined;
-                requiresThinkingAsText?: boolean | undefined;
-                reasoningContentField?: "reasoning" | "reasoning_content" | "reasoning_text" | undefined;
-                requiresReasoningContentForToolCalls?: boolean | undefined;
-                allowsSyntheticReasoningContentForToolCalls?: boolean | undefined;
-                requiresAssistantContentForToolCalls?: boolean | undefined;
-                supportsToolChoice?: boolean | undefined;
-                disableReasoningOnForcedToolChoice?: boolean | undefined;
-                disableReasoningOnToolChoice?: boolean | undefined;
-                thinkingFormat?: "openai" | "openrouter" | "qwen" | "qwen-chat-template" | "zai" | undefined;
-                openRouterRouting?: {
-                    only?: string[] | undefined;
-                    order?: string[] | undefined;
-                } | undefined;
-                vercelGatewayRouting?: {
-                    only?: string[] | undefined;
-                    order?: string[] | undefined;
-                } | undefined;
-                extraBody?: Record<string, unknown> | undefined;
-                cacheControlFormat?: "anthropic" | undefined;
-                supportsStrictMode?: boolean | undefined;
-                toolStrictMode?: "all_strict" | "none" | undefined;
-            } | undefined;
-            contextPromotionTarget?: string | undefined;
-        }> | undefined;
-        disableStrictTools?: boolean | undefined;
-        transport?: "pi-native" | undefined;
-    }> | undefined;
-    equivalence?: {
-        overrides?: Record<string, string> | undefined;
-        exclude?: string[] | undefined;
-    } | undefined;
-}>;
 /** Provider override config (baseUrl, headers, apiKey, compat, transport) without custom models */
 interface ProviderOverride {
     baseUrl?: string;
     headers?: Record<string, string>;
     apiKey?: string;
     authHeader?: boolean;
-    compat?: Model<Api>["compat"];
+    compat?: ModelSpec<Api>["compat"];
     transport?: Model<Api>["transport"];
 }
 /**
@@ -262,6 +50,11 @@ export interface CanonicalModelQueryOptions {
     availableOnly?: boolean;
     candidates?: readonly Model<Api>[];
 }
+/** A canonical record (with query-filtered variants) plus the variant model selected for it. */
+export interface CanonicalModelSelection {
+    record: CanonicalModelRecord;
+    model: Model<Api>;
+}
 /**
  * Model registry - loads and manages models, resolves API keys via AuthStorage.
  */
@@ -306,6 +99,14 @@ export declare class ModelRegistry {
      */
     getAll(): Model<Api>[];
     getCanonicalModels(options?: CanonicalModelQueryOptions): CanonicalModelRecord[];
+    /**
+     * One-pass equivalent of `getCanonicalModels` + `resolveCanonicalModel` per
+     * record. The per-query state (candidate-selector set, availability memo,
+     * provider rank, candidate order) is built once, so the whole catalog
+     * resolves in O(records + candidates) instead of O(records × candidates).
+     * This is the path the model selector hydrates from synchronously on open.
+     */
+    getCanonicalModelSelections(options?: CanonicalModelQueryOptions): CanonicalModelSelection[];
     getCanonicalVariants(canonicalId: string, options?: CanonicalModelQueryOptions): CanonicalModelVariant[];
     resolveCanonicalModel(canonicalId: string, options?: CanonicalModelQueryOptions): Model<Api> | undefined;
     getCanonicalId(model: Model<Api>): string | undefined;
@@ -400,7 +201,7 @@ export interface ProviderConfigInput {
     api?: Api;
     streamSimple?: (model: Model<Api>, context: Context, options?: SimpleStreamOptions) => AssistantMessageEventStream;
     headers?: Record<string, string>;
-    compat?: Model<Api>["compat"];
+    compat?: ModelSpec<Api>["compat"];
     authHeader?: boolean;
     /** Streaming transport override — see {@link Model.transport}. */
     transport?: Model<Api>["transport"];
@@ -435,7 +236,7 @@ export interface ProviderConfigInput {
         contextWindow: number;
         maxTokens: number;
         headers?: Record<string, string>;
-        compat?: Model<Api>["compat"];
+        compat?: ModelSpec<Api>["compat"];
         contextPromotionTarget?: string;
         premiumMultiplier?: number;
     }>;

package/dist/types/config/model-resolver.d.ts

@@ -1,12 +1,23 @@
 /**
- * Model resolution, scoping, and initial selection
+ * Model resolution, scoping, and initial selection.
+ *
+ * Layering:
+ * - `matchModel` is the single matching engine. Order: exact `provider/id`
+ *   reference (with OpenRouter routed/date fallbacks) → exact canonical id →
+ *   exact bare id → provider-scoped fuzzy → substring with alias-vs-dated pick.
+ * - `parseModelPatternWithContext`/`parseModelPattern` layer the selector
+ *   grammar on top: trailing `:level` thinking suffixes (`splitThinkingSuffix`)
+ *   and `@upstream` provider routing (`splitUpstreamRouting`).
+ * - Everything else (`resolveModelFromString`, `resolveModelOverride*`,
+ *   `resolveRoleSelection`, `resolveModelScope`, `resolveCliModel`,
+ *   `findSmolModel`/`findSlowModel`) adapts inputs — roles, settings patterns,
+ *   CLI flags, scope globs — onto that pipeline.
  */
 import { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import { type Api, type Effort, type KnownProvider, type Model } from "@oh-my-pi/pi-ai";
-import { type ModelRegistry, type ModelRole } from "./model-registry";
+import type { Api, Effort, Model } from "@oh-my-pi/pi-ai";
+import { type ModelRegistry } from "./model-registry";
+import { type ModelRole } from "./model-roles";
 import type { Settings } from "./settings";
-/** Default model IDs for each known provider */
-export declare const defaultModelPerProvider: Record<KnownProvider, string>;
 export interface ScopedModel {
     model: Model<Api>;
     thinkingLevel?: ThinkingLevel;
@@ -41,11 +52,6 @@ type CliModelRegistry = Pick<ModelRegistry, "getAll"> & Partial<CanonicalModelRe
 type InitialModelRegistry = Pick<ModelRegistry, "getAvailable" | "find">;
 type RestorableModelRegistry = Pick<ModelRegistry, "getAvailable" | "find" | "getApiKey">;
 export declare function getModelMatchPreferences(settings?: Partial<Pick<Settings, "get" | "getStorage">>): ModelMatchPreferences;
-/**
- * Find an exact explicit provider/model match.
- * Bare model ids are handled separately so canonical ids can coalesce variants.
- */
-export declare function findExactModelReferenceMatch(modelReference: string, availableModels: Model<Api>[]): Model<Api> | undefined;
 export interface ParsedModelResult {
     model: Model<Api> | undefined;
     /** Thinking level if explicitly specified in pattern, undefined otherwise */

package/dist/types/config/model-roles.d.ts

@@ -0,0 +1,28 @@
+/**
+ * Built-in model roles and role metadata helpers.
+ */
+import { type ThemeColor } from "../modes/theme/theme";
+import type { Settings } from "./settings";
+export type ModelRole = "default" | "smol" | "slow" | "vision" | "plan" | "designer" | "commit" | "task";
+export interface ModelRoleInfo {
+    tag?: string;
+    name: string;
+    color?: ThemeColor;
+}
+export declare const MODEL_ROLES: Record<ModelRole, ModelRoleInfo>;
+export declare const MODEL_ROLE_IDS: ModelRole[];
+/** Alias for ModelRoleInfo - used for both built-in and custom roles */
+export type RoleInfo = ModelRoleInfo;
+/**
+ * Return the canonical set of known roles for selector/carousel UI.
+ *
+ * Built-ins always come first. Configured cycle order, model assignments, and
+ * tag metadata can introduce additional custom roles without requiring duplicate
+ * entries across settings.
+ */
+export declare function getKnownRoleIds(settings: Settings): string[];
+/**
+ * Get role info for a role name (built-in or custom).
+ * Configured metadata overrides built-in defaults when present.
+ */
+export declare function getRoleInfo(role: string, settings: Settings): RoleInfo;