@oh-my-pi/pi-catalog 18.4.3 → 18.4.5

package/dist/types/discovery/factory-droid.d.ts

@@ -0,0 +1,28 @@
+import { type FactoryDroidRegistryModel } from "../compat/factory-droid.js";
+import type { FetchImpl, ModelSpec } from "../types.js";
+import { type AccountScope } from "../wire/factory-droid.js";
+export interface FactoryDroidModelDiscoveryOptions extends AccountScope {
+    /** OMP-stored WorkOS access token (from `/login factory-droid`), when present. */
+    apiKey?: string;
+    fetch?: FetchImpl;
+}
+/**
+ * Offline seed: the registry narrowed by the same availability rule live
+ * discovery applies, with no flags or org policy known yet. Feature-gated and
+ * explicit-opt-in models stay hidden until an online refresh proves access.
+ */
+export declare function factoryDroidSeedModels(scope: AccountScope): ModelSpec<"factory-droid-agent">[];
+/**
+ * Availability filter, not a catalog: Factory has no model-listing endpoint,
+ * so the KDL registry is narrowed live with `GET /api/feature-flags`
+ * (Statsig gates) and the org model policy in
+ * `GET /api/organization/managed-settings`. Returns null when no credential
+ * resolves or the flags fetch fails — callers keep an offline snapshot, not
+ * evidence of current entitlement. Missing policy denies explicit-opt-in models.
+ */
+export declare function fetchFactoryDroidModels(options?: FactoryDroidModelDiscoveryOptions): Promise<ModelSpec<"factory-droid-agent">[] | null>;
+/** Account scope plus the upstream rotation resolved for that account. */
+export interface FactoryDroidModelBuildOptions extends AccountScope {
+    apiProviders?: readonly string[];
+}
+export declare function buildFactoryDroidModel({ spec, policy }: FactoryDroidRegistryModel, options?: FactoryDroidModelBuildOptions): ModelSpec<"factory-droid-agent">;

package/dist/types/discovery/index.d.ts

@@ -1,5 +1,6 @@
 export * from "./antigravity.js";
 export * from "./codex.js";
+export * from "./factory-droid.js";
 export * from "./gemini.js";
 export * from "./gemini-cli.js";
 export * from "./gitlab-duo-workflow.js";

package/dist/types/discovery/protobuf.d.ts

@@ -21,12 +21,12 @@ export interface ProtoMessage {
 }
 /** A bidirectional codec for one protobuf message type. */
 export interface MessageCodec<T extends ProtoMessage = ProtoMessage> {
-    (value: T): Uint8Array;
+    (value: T): Uint8Array<ArrayBuffer>;
     (value: Uint8Array): T;
     /** Creates a message with protobuf defaults for omitted fields. */
     create(value?: Partial<T>): T;
     /** Encodes one message into protobuf wire bytes. */
-    encode(value: T): Uint8Array;
+    encode(value: T): Uint8Array<ArrayBuffer>;
     /** Decodes one protobuf message from wire bytes. */
     decode(value: Uint8Array): T;
     /** Converts a message to its protobuf JSON representation. */
@@ -91,7 +91,7 @@ export declare function pb<T extends ProtoMessage = ProtoMessage>(typeName: stri
 /** Creates a message using its codec's protobuf defaults. */
 export declare function create<TMessage extends ProtoMessage>(codec: MessageCodec<TMessage>, value?: Partial<TMessage>): TMessage;
 /** Encodes a message using its codec. */
-export declare function toBinary<TMessage extends ProtoMessage>(codec: MessageCodec<TMessage>, value: TMessage): Uint8Array;
+export declare function toBinary<TMessage extends ProtoMessage>(codec: MessageCodec<TMessage>, value: TMessage): Uint8Array<ArrayBuffer>;
 /** Decodes wire bytes using a message codec. */
 export declare function fromBinary<TMessage extends ProtoMessage>(codec: MessageCodec<TMessage>, value: Uint8Array): TMessage;
 /** Converts a message to protobuf JSON using its codec. */

package/dist/types/hosts.d.ts

@@ -165,5 +165,18 @@ export declare function isVertexExpressOpenAIUrl(baseUrl: string): boolean;
 export declare function isVertexRawPredictUrl(baseUrl: string): boolean;
 /** Azure OpenAI deployment-scoped path (`…/deployments/<name>/…`). */
 export declare function isAzureDeploymentsUrl(baseUrl: string): boolean;
+/**
+ * Amazon Bedrock API route on a bedrock-runtime or bedrock-mantle endpoint
+ * (public, FIPS, or PrivateLink hostname):
+ * - `anthropic`: the Anthropic Messages API under `/anthropic`.
+ * - `openai`: the OpenAI-compatible APIs under `/openai`, plus Mantle's
+ *   documented `/v1` base (`https://bedrock-mantle.<region>.api.aws/v1`).
+ *
+ * Hostnames are parsed strictly so proxies that embed these hosts in a path
+ * do not match.
+ */
+export declare function isBedrockRouteUrl(baseUrl: string | undefined, route: "openai" | "anthropic"): boolean;
+/** Amazon Bedrock's OpenAI-compatible routes; see {@link isBedrockRouteUrl}. */
+export declare function isBedrockOpenAIUrl(baseUrl: string | undefined): boolean;
 /** Alibaba DashScope consumer `compatible-mode` endpoint (rejects multimodal arrays for some text-only SKUs). */
 export declare function isDashscopeCompatibleModeUrl(baseUrl: string): boolean;

package/dist/types/model-manager.d.ts

@@ -35,6 +35,16 @@ export interface ModelManagerOptions<TApi extends Api = Api, TModelsDevPayload =
     cacheTtlMs?: number;
     /** When true, a successful dynamic fetch is the complete provider catalog and prunes static-only models. */
     dynamicModelsAuthoritative?: boolean;
+    /**
+     * When true, a fresh cache never satisfies an online-eligible refresh: the
+     * dynamic fetch always runs (an explicit `"offline"` strategy is still
+     * honored), and the cache serves only as the fetch-failure fallback. For
+     * providers whose dynamic result encodes fast-changing live state that a
+     * TTL cache would replay wrongly — e.g. Factory discovery includes live
+     * organization model policy and upstream blocks. Cached eligibility is
+     * an offline snapshot, not current entitlement.
+     */
+    alwaysRefetchDynamicModels?: boolean;
     /** Cached model ids whose presence forces refresh when the static or migration-policy fingerprint changes. */
     dropCachedModelIdsOnStaticMismatch?: readonly string[];
     /**

package/dist/types/models.d.ts

@@ -1,9 +1,13 @@
 import MODELS from "./models.json";
-import type { Api, KnownProvider, Model, ModelCost, Usage } from "./types.js";
+import type { Api, KnownProvider, Model, ModelCost, ModelPricingStatus, Usage } from "./types.js";
 export type GeneratedProvider = keyof typeof MODELS;
+/** Whether `provider` has bundled rows in models.json (e.g. a provider id authored in KDL). */
+export declare function isGeneratedProvider(provider: string): provider is GeneratedProvider;
 export declare function getBundledModel<TApi extends Api = Api>(provider: GeneratedProvider, modelId: string): Model<TApi>;
 export declare function getBundledProviders(): KnownProvider[];
 export declare function getBundledModels(provider: GeneratedProvider): Model<Api>[];
+/** Resolve display semantics without mistaking an absent rate card for a free model. */
+export declare function getModelPricingStatus(model: Pick<Model<Api>, "cost" | "pricingStatus">): ModelPricingStatus;
 /** Return the recurring UTC tariff period, independently of its monetary multiplier. */
 export declare function getTimeBasedPricingPeriod(cost: ModelCost, timestamp?: number): "peak" | "off-peak" | undefined;
 /** Return the next actual peak/off-peak change strictly after the Unix-ms timestamp. */

package/dist/types/provider-models/cache-provider-id.d.ts

@@ -1,4 +1,5 @@
-export interface ModelCacheProviderIdOptions {
+import { type AccountScope } from "../wire/factory-droid.js";
+export interface ModelCacheProviderIdOptions extends AccountScope {
     apiKey?: string;
     baseUrl?: string;
 }

package/dist/types/provider-models/descriptor-types.d.ts

@@ -1,8 +1,9 @@
 import type { CompiledProviderDiscovery } from "../compat/types.js";
 import type { ModelManagerOptions } from "../model-manager.js";
 import type { Api, FetchImpl } from "../types.js";
+import type { AccountScope } from "../wire/factory-droid.js";
 /** Config passed to a provider's runtime model-manager factory. */
-export type ModelManagerConfig = {
+export type ModelManagerConfig = AccountScope & {
     apiKey?: string;
     baseUrl?: string;
     fetch?: FetchImpl;

package/dist/types/provider-models/openai-compat.d.ts

@@ -106,6 +106,27 @@ export interface HuggingfaceModelManagerConfig {
     fetch?: FetchImpl;
 }
 export declare function huggingfaceModelManagerOptions(config?: HuggingfaceModelManagerConfig): ModelManagerOptions<"openai-completions">;
+export interface HelmcodeModelManagerConfig {
+    apiKey?: string;
+    baseUrl?: string;
+    fetch?: FetchImpl;
+}
+/**
+ * Helmcode model manager: OpenAI-compatible chat completions at
+ * `api.helmcode.com/v1`. `/v1/models` also lists embedding, rerank, TTS, and
+ * STT models; the exclusion policy lives in `runtime/behavior.kdl`
+ * (`exclude-models provider="helmcode"`).
+ *
+ * `/v1/models` carries no capability data. Resold frontier ids (Claude, GPT,
+ * Gemini) take only capability facts from the first-party vendor's bundled
+ * row: reasoning, modalities, context window, output cap, and list price. The
+ * rest of that row (thinking shape, compat, native web search, tool dialects,
+ * cache semantics) describes the vendor's own API, not this chat-completions
+ * proxy; the host's `reasoning_effort` ladders and cache-write pricing live in
+ * `providers/helmcode.kdl`. Ids with no Helmcode or vendor row (e.g. a new
+ * open-weight model) inherit nothing from other gateways.
+ */
+export declare function helmcodeModelManagerOptions(config?: HelmcodeModelManagerConfig): ModelManagerOptions<"openai-completions">;
 export interface NvidiaModelManagerConfig {
     apiKey?: string;
     baseUrl?: string;

package/dist/types/provider-models/special.d.ts

@@ -1,4 +1,5 @@
 import type { DevinModelDiscoveryOptions } from "../discovery/devin.js";
+import { type FactoryDroidModelDiscoveryOptions } from "../discovery/factory-droid.js";
 import type { ModelManagerOptions } from "../model-manager.js";
 import type { Api, FetchImpl, Model } from "../types.js";
 /** One Codex OAuth account to fetch a catalog for. */
@@ -55,6 +56,7 @@ export interface DevinModelManagerConfig {
     fetch?: DevinModelDiscoveryOptions["fetch"];
 }
 export declare function devinModelManagerOptions(config?: DevinModelManagerConfig): ModelManagerOptions<"devin-agent">;
+export declare function factoryDroidModelManagerOptions(config?: FactoryDroidModelDiscoveryOptions): ModelManagerOptions<"factory-droid-agent">;
 export declare function localModelManagerOptions(): ModelManagerOptions<"local-inference">;
 export declare function webModelManagerOptions(): ModelManagerOptions<"web-search">;
 /** Credentials and endpoint overrides for the TypeSafe catalog manager. */

package/dist/types/types.d.ts

@@ -2,7 +2,7 @@ import type { ModelIdentity } from "./compat/types.js";
 import type { Effort } from "./effort.js";
 export type { FetchImpl } from "@oh-my-pi/pi-utils";
 export type { KnownProvider } from "./provider-models/descriptors.js";
-export type KnownApi = "openai-completions" | "openai-responses" | "openrouter" | "openai-codex-responses" | "azure-openai-responses" | "anthropic-messages" | "bedrock-converse-stream" | "google-generative-ai" | "google-gemini-cli" | "google-vertex" | "ollama-chat" | "cursor-agent" | "gitlab-duo-agent" | "devin-agent" | "apple-foundation-models";
+export type KnownApi = "openai-completions" | "openai-responses" | "openrouter" | "openai-codex-responses" | "azure-openai-responses" | "anthropic-messages" | "bedrock-converse-stream" | "google-generative-ai" | "google-gemini-cli" | "google-vertex" | "ollama-chat" | "cursor-agent" | "factory-droid-agent" | "gitlab-duo-agent" | "devin-agent" | "apple-foundation-models";
 export type Api = KnownApi | (string & {});
 /** Catalog kinds used to isolate role-specific runners from session chat models. */
 export declare const MODEL_KINDS: readonly ["chat", "tiny", "image", "tts", "stt", "search", "judge", "embedding", "rerank", "video"];
@@ -206,12 +206,21 @@ export interface OpenAICompat {
     thinkingKeep?: "all" | false;
     /** Which reasoning content field to emit on assistant messages. Default: auto-detected. */
     reasoningContentField?: "reasoning_content" | "reasoning" | "reasoning_text";
+    /** Decode Mistral thinking parts in streamed content and replay them as typed assistant content. */
+    mistralReasoningContentParts?: boolean;
     /** Whether assistant tool-call messages must include reasoning content. Default: false. */
     requiresReasoningContentForToolCalls?: boolean;
     /** Whether all assistant messages must include reasoning content. Default: false. */
     requiresReasoningContentForAllAssistantTurns?: boolean;
     /** Whether the provider accepts a synthetic placeholder (e.g. ".") for missing reasoning_content on tool-call turns. Default: true. Set to false for providers like DeepSeek that validate the exact reasoning_content value. */
     allowsSyntheticReasoningContentForToolCalls?: boolean;
+    /**
+     * Value emitted for the reasoning field on tool-call turns when the provider
+     * requires it (`requiresReasoningContentForToolCalls`) but no reasoning was
+     * captured. Default: "". Some upstreams validate the exact value — the droid
+     * proxy's DeepSeek family requires a single space.
+     */
+    syntheticReasoningContentFallback?: string;
     /**
      * Replay preserved thinking blocks as `reasoning_content` (or the configured
      * `reasoningContentField`) on EVERY assistant turn that carried reasoning,
@@ -481,6 +490,15 @@ export interface AnthropicCompat {
      * `tools.<n>.custom.strict: Extra inputs are not permitted`.
      */
     disableStrictTools?: boolean;
+    /**
+     * The endpoint is Amazon Bedrock's Anthropic Messages API (`/anthropic` on
+     * bedrock-runtime or bedrock-mantle). Requests drop tool `strict`, fit
+     * `metadata.user_id` to Bedrock's request-metadata pattern, and may use
+     * on-demand compaction. Detected from the model `baseUrl`; set it in
+     * models.yml `compat` for a route the URL check cannot see (a proxy, an
+     * `ANTHROPIC_BASE_URL` reroute) or to `false` to opt out.
+     */
+    bedrockMessagesApi?: boolean;
     /**
      * Map adaptive thinking (`thinking: { type: "adaptive" }`) to
      * `{ type: "enabled", budget_tokens }`. Vertex AI rejects the `adaptive`
@@ -508,6 +526,13 @@ export interface AnthropicCompat {
      * `input_transformations` under the thinking-binding-controls beta.
      */
     supportsThinkingBindingControls?: boolean;
+    /**
+     * Whether the model replaces `thinking: { type: "disabled" }` with
+     * `thinking: { type: "between_tools" }` (Claude Sonnet 5.5). The disabled
+     * form is rejected with a 400; `between_tools` skips up-front thinking and
+     * only allows progress updates between tool calls.
+     */
+    supportsBetweenToolsThinking?: boolean;
     /**
      * Whether the model accepts a forced `tool_choice` (`{ type: "any" }` or
      * `{ type: "tool", name }`). Claude Fable/Mythos 5 reject forced tool use
@@ -578,6 +603,25 @@ export interface AnthropicCompat {
     stripImageInput?: boolean;
     /** Thinking-loop watchdog guard family applied to streamed reasoning. */
     thinkingLoopGuard?: "gemini" | "deepseek" | "xai";
+    /**
+     * Drop the enabled `thinking` config and replayed thinking blocks when the
+     * conversation is not thinking-led (a proxy contract for non-interleaved
+     * budget models). Default: false.
+     */
+    stripThinkingHistory?: boolean;
+    /**
+     * Send the `effort-2025-11-24` beta exactly when `output_config.effort`
+     * rides the request. Unset keeps the transport's legacy heuristic.
+     */
+    effortBeta?: boolean;
+    /**
+     * Wire form of a disabled-thinking turn: omit the field, send
+     * `{ type: "disabled" }`, or keep adaptive thinking. Unset keeps the
+     * direct-provider behavior.
+     */
+    disabledThinking?: "omit" | "disabled" | "adaptive";
+    /** The model is a fast-mode SKU: send `speed: "fast"` with the fast-mode beta. Default: false. */
+    fastMode?: boolean;
 }
 /**
  * Compatibility settings for Bedrock Converse prompt caching. Cache pricing is
@@ -679,9 +723,12 @@ export interface ResolvedOpenAISharedCompat {
     supportsForcedToolChoice: boolean;
     supportsNamedToolChoice: boolean;
     reasoningContentField?: OpenAICompat["reasoningContentField"];
+    mistralReasoningContentParts?: boolean;
     requiresReasoningContentForToolCalls: boolean;
     requiresReasoningContentForAllAssistantTurns: boolean;
     allowsSyntheticReasoningContentForToolCalls: boolean;
+    /** See {@link OpenAICompat.syntheticReasoningContentFallback}. */
+    syntheticReasoningContentFallback?: string;
     replayReasoningContent: boolean;
     qwenPreserveThinking: boolean;
     qwenTemplateReasoningEffort: boolean;
@@ -734,7 +781,7 @@ export interface ResolvedOpenAISharedCompat {
  * `buildModel`; request handlers read fields and never detect, resolve, or
  * allocate.
  */
-export type ResolvedOpenAICompat = ResolvedOpenAISharedCompat & Required<Omit<OpenAICompat, "supportsDeveloperRole" | "supportsReasoningEffort" | "reasoningEffortMap" | "supportsReasoningParams" | "supportsReasoningSummary" | "supportsSamplingParams" | "supportsPenaltyAndStopParams" | "thinkingFormat" | "kimiApiFormat" | "reasoningDisableMode" | "omitReasoningEffort" | "includeEncryptedReasoning" | "filterReasoningHistory" | "disableReasoningOnForcedToolChoice" | "disableReasoningOnToolChoice" | "disableReasoningWithTools" | "supportsToolChoice" | "supportsForcedToolChoice" | "supportsNamedToolChoice" | "reasoningContentField" | "requiresReasoningContentForToolCalls" | "requiresReasoningContentForAllAssistantTurns" | "allowsSyntheticReasoningContentForToolCalls" | "replayReasoningContent" | "qwenPreserveThinking" | "qwenTemplateReasoningEffort" | "requiresThinkingAsText" | "requiresMistralToolIds" | "requiresToolResultName" | "requiresAssistantAfterToolResult" | "requiresAssistantContentForToolCalls" | "stripDeepseekSpecialTokens" | "streamMarkupHealingPattern" | "reasoningDeltasMayBeCumulative" | "emptyLengthFinishIsContextError" | "usesOpenAIToolCallIdLimit" | "promptCacheSessionHeader" | "supportsPromptCacheBreakpoints" | "promptCacheBreakpointTtl" | "openRouterRouting" | "isOpenRouterHost" | "supportsStrictMode" | "supportsLongPromptCacheRetention" | "alwaysSendMaxTokens" | "clampOutputToModelMax" | "wireModelIdMode" | "vercelGatewayRouting" | "extraBody" | "toolStrictMode" | "toolSchemaFlavor" | "streamFirstEventTimeoutMs" | "streamIdleTimeoutMs" | "streamRevision" | "cacheControlFormat" | "thinkingKeep" | "strictResponsesPairing" | "supportsImageDetailOriginal" | "supportsConfigurationUpdate" | "supportsSteering" | "stripImageInput" | "thinkingLoopGuard" | "whenThinking">> & {
+export type ResolvedOpenAICompat = ResolvedOpenAISharedCompat & Required<Omit<OpenAICompat, "supportsDeveloperRole" | "supportsReasoningEffort" | "reasoningEffortMap" | "supportsReasoningParams" | "supportsReasoningSummary" | "supportsSamplingParams" | "supportsPenaltyAndStopParams" | "thinkingFormat" | "kimiApiFormat" | "reasoningDisableMode" | "omitReasoningEffort" | "includeEncryptedReasoning" | "filterReasoningHistory" | "disableReasoningOnForcedToolChoice" | "disableReasoningOnToolChoice" | "disableReasoningWithTools" | "supportsToolChoice" | "supportsForcedToolChoice" | "supportsNamedToolChoice" | "reasoningContentField" | "mistralReasoningContentParts" | "requiresReasoningContentForToolCalls" | "requiresReasoningContentForAllAssistantTurns" | "allowsSyntheticReasoningContentForToolCalls" | "syntheticReasoningContentFallback" | "replayReasoningContent" | "qwenPreserveThinking" | "qwenTemplateReasoningEffort" | "requiresThinkingAsText" | "requiresMistralToolIds" | "requiresToolResultName" | "requiresAssistantAfterToolResult" | "requiresAssistantContentForToolCalls" | "stripDeepseekSpecialTokens" | "streamMarkupHealingPattern" | "reasoningDeltasMayBeCumulative" | "emptyLengthFinishIsContextError" | "usesOpenAIToolCallIdLimit" | "promptCacheSessionHeader" | "supportsPromptCacheBreakpoints" | "promptCacheBreakpointTtl" | "openRouterRouting" | "isOpenRouterHost" | "supportsStrictMode" | "supportsLongPromptCacheRetention" | "alwaysSendMaxTokens" | "clampOutputToModelMax" | "wireModelIdMode" | "vercelGatewayRouting" | "extraBody" | "toolStrictMode" | "toolSchemaFlavor" | "streamFirstEventTimeoutMs" | "streamIdleTimeoutMs" | "streamRevision" | "cacheControlFormat" | "thinkingKeep" | "strictResponsesPairing" | "supportsImageDetailOriginal" | "supportsConfigurationUpdate" | "supportsSteering" | "stripImageInput" | "thinkingLoopGuard" | "whenThinking">> & {
     vercelGatewayRouting?: OpenAICompat["vercelGatewayRouting"];
     extraBody?: OpenAICompat["extraBody"];
     cacheControlFormat?: OpenAICompat["cacheControlFormat"];
@@ -804,9 +851,23 @@ export interface ResolvedOpenAIResponsesCompat extends ResolvedOpenAISharedCompa
  */
 export type ResolvedOpenRouterCompat = ResolvedOpenAICompat & ResolvedOpenAIResponsesCompat;
 /** Fully-resolved anthropic-messages compat view (same contract as `ResolvedOpenAICompat`). */
-export type ResolvedAnthropicCompat = Required<Omit<AnthropicCompat, "streamIdleTimeoutMs" | "thinkingLoopGuard">> & {
+export type ResolvedAnthropicCompat = Required<Omit<AnthropicCompat, "streamIdleTimeoutMs" | "thinkingLoopGuard" | "bedrockMessagesApi" | "effortBeta" | "disabledThinking" | "stripThinkingHistory" | "fastMode">> & {
+    /** Effort-beta override; undefined keeps the transport's legacy heuristic. */
+    effortBeta?: AnthropicCompat["effortBeta"];
+    /** Disabled-thinking wire form; undefined keeps the direct-provider behavior. */
+    disabledThinking?: AnthropicCompat["disabledThinking"];
+    /** Strip thinking history on non-thinking-led turns; undefined behaves as false. */
+    stripThinkingHistory?: AnthropicCompat["stripThinkingHistory"];
+    /** Fast-mode SKU; undefined behaves as false. */
+    fastMode?: AnthropicCompat["fastMode"];
     /** Thinking-loop watchdog guard family applied to streamed reasoning. */
     thinkingLoopGuard?: AnthropicCompat["thinkingLoopGuard"];
+    /**
+     * Bedrock's Anthropic Messages API (see {@link AnthropicCompat.bedrockMessagesApi}).
+     * `true` when detected or configured; otherwise unset, so rows baked before
+     * this field keep matching.
+     */
+    bedrockMessagesApi?: boolean;
     /**
      * Stream-watchdog idle-timeout fallback in ms for slow reasoning hosts; 0 disables the idle watchdog.
      * Undefined defers to `PI_STREAM_IDLE_TIMEOUT_MS`, then the legacy
@@ -923,6 +984,8 @@ export interface TokenCost {
     cacheRead: number;
     cacheWrite: number;
 }
+/** User-facing interpretation of a model's token-rate card. */
+export type ModelPricingStatus = "fixed" | "free" | "included" | "variable" | "unknown";
 /**
  * Rates applied to the full request when its prompt exceeds `inputThreshold`,
  * or reaches it when `inputThresholdInclusive` is true. Prompt input is the
@@ -980,6 +1043,17 @@ export interface ModelAccountAccess {
      */
     cyberPrograms?: readonly string[];
 }
+/** One Cursor `RequestedModel.parameters` entry recovered from rich discovery. */
+export interface CursorModelParameter {
+    id: string;
+    value: string;
+}
+/** Cursor wire target selected after local effort/variant routing. */
+export interface CursorModelRoute {
+    modelId: string;
+    parameters: readonly CursorModelParameter[];
+    maxMode?: boolean;
+}
 export interface Model<TApi extends Api = Api> {
     id: string;
     /** Role-specific runner capability; omitted for ordinary chat models. */
@@ -1088,6 +1162,29 @@ export interface Model<TApi extends Api = Api> {
      * bundled/config rows and on single-account discovery.
      */
     accountAccess?: Readonly<Record<string, ModelAccountAccess>>;
+    /** Cursor `RequestedModel.parameters` for this model's default variant. */
+    cursorModelParameters?: readonly CursorModelParameter[];
+    /**
+     * Per-wire-id Cursor routes recovered from `AvailableModels`. Generic effort
+     * collapse retains these while routing selects one key at request time.
+     */
+    cursorModelRoutes?: Readonly<Record<string, CursorModelRoute>>;
+    /** Cursor's account-relative model price/multiplier; not a per-token USD rate. */
+    cursorPrice?: number;
+    /** Cursor requires data retention to invoke this model. */
+    cursorRequiresDataRetention?: boolean;
+    /** Cursor explicitly reports support for its Agent surface. */
+    cursorSupportsAgent?: boolean;
+    /** Cursor explicitly reports support for sandboxed execution. */
+    cursorSupportsSandboxing?: boolean;
+    /** Factory Droid: account-resolved upstream rotation (first entry is the default `x-api-provider`). */
+    factoryDroidApiProviders?: string[];
+    /** Factory Droid: `configured_order` when live routing chose the rotation; absent means registry order. */
+    factoryDroidRoutingSource?: "configured_order";
+    /** Factory Droid Standard Credits base rate (relative per-token weight, not dollars). */
+    factoryDroidCredits?: number;
+    /** Canonical organization resolved during Factory discovery. */
+    factoryDroidOrgId?: string;
     cost: ModelCost;
     /**
      * Prompt-cache entry lifetime per retention tier, in seconds. Populated only
@@ -1097,9 +1194,16 @@ export interface Model<TApi extends Api = Api> {
      * Custom models and provider overrides opt in via models.yml `promptCache`.
      */
     promptCache?: ModelPromptCache;
+    /**
+     * Interpretation of an all-zero token-rate card. Omitted zero-rate cards
+     * are unknown; any non-zero rate is always treated as fixed pricing.
+     */
+    pricingStatus?: Exclude<ModelPricingStatus, "fixed">;
     /** Premium Copilot requests charged per user-initiated request (defaults to 1). */
     premiumMultiplier?: number;
     contextWindow: number | null;
+    /** Preserve the host's supplied window instead of applying inferred expansion or reference-price caps. */
+    contextWindowAuthoritative?: boolean;
     /** Optional larger prompt window available when extended context is enabled. */
     maxContextWindow?: number;
     maxTokens: number | null;
@@ -1143,6 +1247,12 @@ export interface Model<TApi extends Api = Api> {
     useResponsesLite?: boolean;
     /** Codex Code Mode restriction: model expects tools routed through a programmatic exec surface (mirrors codex-rs `tool_mode`). */
     toolMode?: "code_mode_only";
+    /**
+     * Service-tier ids the provider advertises for this model (Codex discovery
+     * `service_tiers[].id`, e.g. `priority`, `ultrafast`). Absent when the
+     * provider publishes no per-model tier list.
+     */
+    serviceTiers?: readonly string[];
     /** Preferred model to switch to when context promotion is triggered (model id or provider/id). */
     contextPromotionTarget?: string;
     /** Preferred model to use only for compaction (model id or provider/id); the active session model is unchanged. */
@@ -1166,6 +1276,8 @@ export interface Model<TApi extends Api = Api> {
     isNew?: boolean;
     /** Upstream marks this model as beta / preview quality. */
     isBeta?: boolean;
+    /** Authenticated provider catalog marks this as the account's default model. */
+    isProviderDefault?: boolean;
     /** Upstream marks this model as one of its recommended picks. */
     isRecommended?: boolean;
     /** Canonical thinking capability metadata for this model. */

package/dist/types/wire/codex.d.ts

@@ -7,10 +7,12 @@ export declare const CODEX_BASE_URL = "https://chatgpt.com/backend-api";
  * Updated from npm by `bun run check-spoofed-versions --update`.
  *
  * The backend version-gates model availability against this value on both
- * `/models?client_version=` and `/responses`; 0.155.1 exposes GPT-6 Sol and Luna.
- * An older pin silently hides newer SKUs from discovery.
+ * `/models?client_version=` and `/responses`, independently of a model's
+ * declared `minimal_client_version`: GPT-6.1 Sol declares 0.153.0 yet is
+ * omitted from `/models` at 0.155.1 and listed at 0.159.0. An older pin
+ * silently hides newer SKUs from discovery.
  */
-export declare const CODEX_CLIENT_VERSION = "0.155.1";
+export declare const CODEX_CLIENT_VERSION = "0.159.0";
 export declare const OPENAI_HEADERS: {
     readonly BETA: "OpenAI-Beta";
     /** Codex feature-negotiation header; values identify opt-in wire protocols. */

package/dist/types/wire/cursor.d.ts

@@ -0,0 +1,20 @@
+/**
+ * Cursor wire constants shared by authenticated catalog discovery, the pi-ai
+ * provider, and account usage. Keep the announced client build aligned with the
+ * released Cursor agent CLI because the service gates protocol features on it.
+ */
+/** Default host for Cursor's Connect RPCs and account API. */
+export declare const CURSOR_DEFAULT_BASE_URL = "https://api2.cursor.sh";
+/** Released Cursor agent CLI build whose protocol surface this client mirrors. */
+export declare const CURSOR_CLIENT_VERSION = "cli-2026.09.02-c22c1a3";
+export declare const CURSOR_RUN_PATH = "/agent.v1.AgentService/Run";
+export declare const CURSOR_RUN_SSE_PATH = "/agent.v1.AgentService/RunSSE";
+export declare const CURSOR_BIDI_APPEND_PATH = "/aiserver.v1.BidiService/BidiAppend";
+export declare const CURSOR_GET_USABLE_MODELS_PATH = "/agent.v1.AgentService/GetUsableModels";
+export declare const CURSOR_GET_DEFAULT_MODEL_PATH = "/agent.v1.AgentService/GetDefaultModelForCli";
+export declare const CURSOR_AVAILABLE_MODELS_PATH = "/aiserver.v1.AiService/AvailableModels";
+/** Headers shared by Cursor's authenticated CLI RPCs. */
+export declare function cursorClientHeaders(apiKey: string, options?: {
+    clientVersion?: string;
+    contentType?: "application/proto" | "application/connect+proto";
+}): Record<string, string>;

package/dist/types/wire/factory-droid.d.ts

@@ -0,0 +1,45 @@
+/**
+ * Client version reported to Factory's API. When bumping it, recapture the
+ * native request corpus with `packages/ai/scripts/capture-factory-droid-native.ts`
+ * against the matching CLI binary and make `factory-droid-native-parity.test.ts` pass.
+ */
+export declare const FACTORY_DROID_CLIENT_VERSION = "0.230.0";
+/** Wire protocols the Factory proxy multiplexes; `api-routes provider="factory-droid"` picks one per model. */
+export declare const FACTORY_DROID_WIRES: readonly ["openai-completions", "openai-responses", "anthropic-messages", "google-generate"];
+/** Wire protocol used by the Factory proxy for each model. */
+export type FactoryDroidWire = (typeof FACTORY_DROID_WIRES)[number];
+/**
+ * Upstream routers the proxy dispatches to, sent as the `x-api-provider`
+ * header. The KDL compiler validates the rotation and region axes against it.
+ */
+export declare const FACTORY_DROID_UPSTREAMS: readonly ["fireworks", "baseten", "mistral", "anthropic", "azure_anthropic", "vertex_anthropic", "bedrock_anthropic", "openai", "azure_openai", "bedrock_openai", "google", "xai", "snowflake", "databricks"];
+/** Inference serving region; independent from account/API-host residency. */
+export type FactoryDroidRegion = "global" | "us" | "eu";
+/**
+ * Account scope a Factory credential resolves to. Residency (`region`) selects
+ * the API host; `inferenceRegion` selects eligible upstreams and capacities.
+ * An EU-host account can legitimately have global inference.
+ */
+export interface AccountScope {
+    region?: string;
+    inferenceRegion?: FactoryDroidRegion;
+    orgId?: string;
+}
+/** Inference region for an account: explicit scope wins, else EU residency infers EU, else global. */
+export declare function resolveFactoryDroidInferenceRegion(scope: {
+    region?: string;
+    inferenceRegion?: FactoryDroidRegion;
+}): FactoryDroidRegion;
+/** Factory API host per residency region; EU accounts are served from the EU region. */
+export declare function factoryDroidApiBaseUrl(region: string | undefined): string;
+/** Per-wire base URL for an account region; the stream layer appends the path suffix. */
+export declare function factoryDroidWireBaseUrl(wire: FactoryDroidWire, region: string | undefined): string;
+/**
+ * Client identity headers shared by discovery, usage and inference requests.
+ * Callers add their own auth, `Accept`, routing and SDK headers.
+ */
+export declare function factoryDroidClientHeaders(orgId: string | undefined): Record<string, string>;
+/** Model-cache namespace for a Factory credential and its account scope. */
+export declare function factoryDroidModelCacheProviderId(options: {
+    apiKey?: string;
+} & AccountScope): string;

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.4.3",
+	"version": "18.4.5",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"keywords": [
 		"ai",
@@ -93,12 +93,12 @@
 		"gen:proto": "bun scripts/generate-protocols.ts"
 	},
 	"dependencies": {
-		"@oh-my-pi/omptype": "18.4.3",
-		"@oh-my-pi/pi-utils": "18.4.3"
+		"@oh-my-pi/omptype": "18.4.5",
+		"@oh-my-pi/pi-utils": "18.4.5"
 	},
 	"devDependencies": {
 		"@bgotink/kdl": "0.4.0",
-		"@oh-my-pi/pi-ai": "18.4.3",
+		"@oh-my-pi/pi-ai": "18.4.5",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/build.ts

@@ -76,6 +76,11 @@ function isInputModalities(value: unknown): value is ("text" | "image")[] {
 function applyCatalogAssignments<TApi extends Api>(model: Model<TApi>, catalog: Record<string, unknown>): void {
 	const kind = MODEL_KINDS.find(value => value === catalog.kind);
 	if (kind !== undefined) model.kind = kind;
+	if (catalog.contextWindowAuthoritative === true) {
+		model.contextWindowAuthoritative = true;
+	} else {
+		delete model.contextWindowAuthoritative;
+	}
 	const webSearch = catalog.webSearch;
 	if (
 		webSearch === "gemini" ||
@@ -120,6 +125,15 @@ function applyCatalogAssignments<TApi extends Api>(model: Model<TApi>, catalog:
 	if (editPromptVariant === "full" || editPromptVariant === "compact") {
 		model.editPromptVariant = editPromptVariant;
 	}
+	const pricingStatus = catalog.pricingStatus;
+	if (
+		pricingStatus === "free" ||
+		pricingStatus === "included" ||
+		pricingStatus === "variable" ||
+		pricingStatus === "unknown"
+	) {
+		model.pricingStatus = pricingStatus;
+	}
 	const requiresCursorToolSchemaProjection = catalog.requiresCursorToolSchemaProjection;
 	if (requiresCursorToolSchemaProjection === true) {
 		model.requiresCursorToolSchemaProjection = true;
@@ -137,6 +151,11 @@ function applyCatalogAssignments<TApi extends Api>(model: Model<TApi>, catalog:
 	} else {
 		delete model.supportsAssistantPrefill;
 	}
+	// The same KDL output-cap contract gates host options and the Responses
+	// transport. Materialize it because the transport consumes the model field.
+	if (typeof catalog.omitMaxOutputTokens === "boolean" && model.omitMaxOutputTokens === undefined) {
+		model.omitMaxOutputTokens = catalog.omitMaxOutputTokens;
+	}
 	const contextPromotionTarget = catalog.contextPromotionTarget;
 	if (typeof contextPromotionTarget === "string" && model.contextPromotionTarget === undefined) {
 		model.contextPromotionTarget = contextPromotionTarget;

package/src/compat/anthropic.ts

@@ -5,7 +5,7 @@
  * pi-ai both consume these; URL matching is the one detection surface that
  * stays in code.
  */
-import { hostMatchesUrl } from "../hosts";
+import { hostMatchesUrl, isBedrockRouteUrl } from "../hosts";
 
 const OFFICIAL_ANTHROPIC_URL = "https://api.anthropic.com";
 
@@ -27,6 +27,15 @@ const VERTEX_ANTHROPIC_URL_MARKER = /aiplatform\.googleapis\.com\/.+\/publishers
 const BEDROCK_ANTHROPIC_URL_MARKER = /(?:^|\/\/|\.)bedrock-runtime\.[a-z0-9-]+\.amazonaws\.com/i;
 const AZURE_ANTHROPIC_URL_MARKER = /(?:^|\/\/|\.)[a-z0-9-]+\.(?:inference|services)\.ai\.azure\.com/i;
 
+/**
+ * Amazon Bedrock's Anthropic Messages API: the `/anthropic` path on the
+ * bedrock-runtime or bedrock-mantle endpoint. It rejects the tool `strict`
+ * field and enforces Anthropic signatures on replay.
+ */
+export function isBedrockAnthropicRoute(baseUrl?: string): boolean {
+	return isBedrockRouteUrl(baseUrl, "anthropic");
+}
+
 /**
  * Azure AI Inference / Foundry Anthropic route
  * (`<resource>.inference.ai.azure.com`, `<resource>.services.ai.azure.com`).
@@ -53,6 +62,7 @@ export function isAnthropicSigningProxyUrl(baseUrl?: string): boolean {
 			(CLOUDFLARE_ANTHROPIC_GATEWAY_URL_MARKER.test(baseUrl) ||
 				VERTEX_ANTHROPIC_URL_MARKER.test(baseUrl) ||
 				BEDROCK_ANTHROPIC_URL_MARKER.test(baseUrl))) ||
+		isBedrockAnthropicRoute(baseUrl) ||
 		isAzureAnthropicRoute(baseUrl)
 	);
 }

package/src/compat/auth-ids.ts

@@ -24,6 +24,7 @@ export type AuthProviderId =
 	| "deepseek"
 	| "devin"
 	| "exa"
+	| "factory-droid"
 	| "firepass"
 	| "fireworks"
 	| "github-copilot"
@@ -35,6 +36,7 @@ export type AuthProviderId =
 	| "google-gemini-cli"
 	| "google-vertex"
 	| "groq"
+	| "helmcode"
 	| "huggingface"
 	| "kagi"
 	| "kilo"
@@ -113,6 +115,7 @@ export type LoginProviderId =
 	| "deepseek"
 	| "devin"
 	| "exa"
+	| "factory-droid"
 	| "firepass"
 	| "fireworks"
 	| "github-copilot"
@@ -121,6 +124,7 @@ export type LoginProviderId =
 	| "gmi-cloud"
 	| "google-antigravity"
 	| "google-gemini-cli"
+	| "helmcode"
 	| "huggingface"
 	| "kagi"
 	| "kilo"