@oh-my-pi/pi-catalog 18.0.1 → 18.0.4

package/CHANGELOG.md

@@ -2,11 +2,26 @@
 
 ## [Unreleased]
 
+## [18.0.4] - 2026-08-24
+
+### Fixed
+
+- Fixed default reasoning effort for `cursor/cursor-grok-4.5` and `cursor/cursor-grok-4.6` so requests without an explicit effort setting default to `-medium` instead of `-low`, preventing rate limit rejections on Cursor's Start plan ([#9478](https://github.com/can1357/oh-my-pi/issues/9478)).
+- Fixed aliased OpenCode Zen Ox Alpha models exposing incorrect effort levels, ensuring the gateway's native `low`, `high`, and `max` tiers are correctly mapped and reachable ([#9349](https://github.com/can1357/oh-my-pi/issues/9349)).
+- Fixed missing rate card tier for public xAI and SuperGrok models with prompt lengths exceeding 200K tokens ([#9512](https://github.com/can1357/oh-my-pi/issues/9512)).
+
+## [18.0.2] - 2026-08-23
+
+### Fixed
+
+- Fixed OpenRouter auxiliary requests (e.g. session-title generation) failing with `400 Reasoning is mandatory for this endpoint and cannot be disabled` on mandatory-reasoning models such as `stealth/ox-alpha`. Live discovery now honors the endpoint's `reasoning.mandatory` flag, clamping thinking-off to the lowest supported effort instead of sending `reasoning: { enabled: false }` ([#9415](https://github.com/can1357/oh-my-pi/issues/9415)).
+
 ## [18.0.1] - 2026-08-23
 
 ### Added
 
 - Fixed `google-gemini-cli` model refresh returning only bundled models for Gemini Code Assist Standard accounts, whose credential is not authorized for the Antigravity `fetchAvailableModels` endpoint (HTTP 403). Discovery now falls back to the account's own `retrieveUserQuota` list on Cloud Code Assist, surfacing models such as `gemini-3.5-flash` ([#9315](https://github.com/can1357/oh-my-pi/issues/9315)).
+- Added Amazon Bedrock guardrail metadata to model definitions for Converse requests.
 
 ### Fixed
 
@@ -16,9 +31,6 @@
 - Fixed `opencode-go/deepseek-v4-flash` exposing the generic `minimal`/`low`/`medium`/`high`/`xhigh` thinking ladder instead of DeepSeek V4's real `low`/`high`/`max` tiers. The model is pinned to the Responses transport (the Go gateway serves it only at `/responses`), which the DeepSeek effort branch did not admit, so it fell through to the default ladder; the branch now covers the `openai-responses` transport like every other host ([#9134](https://github.com/can1357/oh-my-pi/issues/9134)).
 - Fixed protobuf map decoding corrupting entries when a key is `__proto__`, which dropped that argument and replayed spurious numeric arguments ([#9394](https://github.com/can1357/oh-my-pi/issues/9394)).
 
-### Added
-
-- Added Amazon Bedrock guardrail metadata to model definitions for Converse requests.
 ## [18.0.0] - 2026-08-22
 
 ### Added

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

@@ -67,4 +67,15 @@ export declare function resolveWireModelId<TApi extends Api>(model: ApiModel<TAp
  * thinking-off requests on `thinking.requiresEffort` models.
  */
 export declare function minimumSupportedEffort<TApi extends Api>(model: ApiModel<TApi>): Effort | undefined;
+/**
+ * Clamp target for effort-less requests on `thinking.requiresEffort` models:
+ * the effort whose wire route equals the model's default wire id
+ * (`requestModelId`), so a collapsed row clamps to the tier it already
+ * advertises as its default rather than the numerically lowest supported tier
+ * (e.g. Cursor Grok 4.5/4.6 default to `medium`, the only tier the Start plan
+ * serves). Falls back to {@link minimumSupportedEffort} when no route matches
+ * the default id — families whose default already is the minimum, or that
+ * expose no routing, are unaffected.
+ */
+export declare function defaultSupportedEffort<TApi extends Api>(model: ApiModel<TApi>): Effort | undefined;
 export {};

package/dist/types/models.d.ts

@@ -1,11 +1,14 @@
 import MODELS from "./models.json";
-import type { Api, KnownProvider, Model, Usage } from "./types.js";
+import type { Api, KnownProvider, Model, ModelCost, Usage } from "./types.js";
 export type GeneratedProvider = keyof typeof MODELS;
 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>[];
 /** Price a prompt as fully uncached input under its active context-length tier. */
-export declare function calculateUncachedInputCost(cost: Model["cost"], promptInputTokens: number): number;
+export declare function calculateUncachedInputCost(cost: ModelCost, promptInputTokens: number): number;
+/** Price one usage record from a token rate card, including active context tiers. */
+export declare function calculateUsageCost(cost: ModelCost, usage: Usage): Usage["cost"];
+/** Price one usage record from its model's token rate card. */
 export declare function calculateCost<TApi extends Api>(model: Model<TApi>, usage: Usage): Usage["cost"];
 /**
  * Check if two models are equal by comparing both their id and provider.

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

@@ -166,6 +166,11 @@ export interface XaiModelManagerConfig {
     baseUrl?: string;
     fetch?: FetchImpl;
 }
+/**
+ * Applies xAI's long-context rate card and mirrors exact public-model prices
+ * onto matching SuperGrok catalog rows.
+ */
+export declare function applyXaiCatalogPricing(models: readonly ModelSpec[]): ModelSpec[];
 export declare function xaiModelManagerOptions(config?: XaiModelManagerConfig): ModelManagerOptions<"openai-responses">;
 export interface XaiOAuthModelManagerConfig {
     apiKey?: string;

package/dist/types/types.d.ts

@@ -705,12 +705,15 @@ export interface TokenCost {
     cacheWrite: number;
 }
 /**
- * Rates applied to the full request when its prompt exceeds `inputThreshold`.
- * Prompt input is the sum of uncached, cached-read, cache-write, and
- * provider-orchestration input tokens.
+ * Rates applied to the full request when its prompt exceeds `inputThreshold`,
+ * or reaches it when `inputThresholdInclusive` is true. Prompt input is the
+ * sum of uncached, cached-read, cache-write, and provider-orchestration input
+ * tokens.
  */
 export interface LongContextTokenCost extends TokenCost {
     inputThreshold: number;
+    /** Whether the long-context tier starts exactly at `inputThreshold`. */
+    inputThresholdInclusive?: boolean;
 }
 /** Base token rates plus an optional long-context tier. */
 export interface ModelCost extends TokenCost {

package/dist/types/variant-collapse.d.ts

@@ -21,6 +21,15 @@ export interface EffortVariantFamily {
      * when it equals the logical id).
      */
     members: readonly string[];
+    /**
+     * Preferred default wire id: overrides the member-order default for the
+     * collapsed spec's `requestModelId` when this member is live. Lets a
+     * mandatory-reasoning family advertise a canonical tier that is not its
+     * numeric floor (Cursor Grok 4.5/4.6 default to `-medium`, the only tier the
+     * Start plan serves; the `-low` floor is refused). Ignored when absent from
+     * the input or retired.
+     */
+    defaultMember?: string;
     /**
      * Wire ids upstream no longer serves (e.g. a deployment killed while
      * discovery still advertises it). Fresh collapsing never routes to them,

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.0.1",
+	"version": "18.0.4",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Stencil Labs, Inc.",
@@ -34,11 +34,11 @@
 		"gen:proto": "bun scripts/generate-protocols.ts"
 	},
 	"dependencies": {
-		"@oh-my-pi/omptype": "18.0.1",
-		"@oh-my-pi/pi-utils": "18.0.1"
+		"@oh-my-pi/omptype": "18.0.4",
+		"@oh-my-pi/pi-utils": "18.0.4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "18.0.1",
+		"@oh-my-pi/pi-ai": "18.0.4",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/model-thinking.ts

@@ -559,15 +559,19 @@ function isSakanaFuguReasoningModel<TApi extends Api>(spec: ModelSpec<TApi>): bo
  * `opencode-zen`) reason through the wire-exact `low`/`high`/`max` ladder with
  * mandatory thinking: the gateway rejects `minimal`/`medium`/`xhigh`
  * (`[1210] ... please use low, high, or max`), the same dialect it already
- * serves for GLM-5.3 and Kimi K3. Other hosts proxying an `ox-alpha` SKU
- * (Kilo, NanoGPT, Venice, OpenRouter) expose their own vocabularies and are
- * left untouched. See issue #9349.
+ * serves for GLM-5.3 and Kimi K3. The SKU also ships under unrelated aliased
+ * ids (`opencode-zen/x-preview-f-free`), so the stencil display name is
+ * matched as well. Other hosts proxying an `ox-alpha` SKU (Kilo, NanoGPT,
+ * Venice, OpenRouter) expose their own vocabularies and are left untouched.
+ * See issue #9349.
  */
 function isOpenCodeGatewayOxAlphaModel<TApi extends Api>(spec: ModelSpec<TApi>): boolean {
-	return (
-		(spec.provider === "opencode-go" || spec.provider === "opencode-zen") &&
-		/(?:^|\/)ox-alpha(?:-|$)/i.test(bareModelId(spec.id))
-	);
+	if (spec.provider !== "opencode-go" && spec.provider !== "opencode-zen") return false;
+	if (/(?:^|\/)ox-alpha(?:-|$)/i.test(bareModelId(spec.id))) return true;
+	// Aliased gateway ids surface the same SKU under an unrelated id; the
+	// display name ("Ox Alpha …") is the stable cross-id signal. Anchored on
+	// both sides so "Box Alpha"/"Ox Alphabet" cannot false-positive.
+	return /\box[ _-]?alpha\b/i.test(spec.name ?? "");
 }
 
 function isDeepseekReasoningModel<TApi extends Api>(spec: ModelSpec<TApi>): boolean {
@@ -923,3 +927,25 @@ export function minimumSupportedEffort<TApi extends Api>(model: ApiModel<TApi>):
 	}
 	return efforts[0];
 }
+
+/**
+ * Clamp target for effort-less requests on `thinking.requiresEffort` models:
+ * the effort whose wire route equals the model's default wire id
+ * (`requestModelId`), so a collapsed row clamps to the tier it already
+ * advertises as its default rather than the numerically lowest supported tier
+ * (e.g. Cursor Grok 4.5/4.6 default to `medium`, the only tier the Start plan
+ * serves). Falls back to {@link minimumSupportedEffort} when no route matches
+ * the default id — families whose default already is the minimum, or that
+ * expose no routing, are unaffected.
+ */
+export function defaultSupportedEffort<TApi extends Api>(model: ApiModel<TApi>): Effort | undefined {
+	const routing = model.thinking?.effortRouting;
+	const defaultWireId = model.requestModelId;
+	if (routing !== undefined && defaultWireId !== undefined) {
+		const efforts = model.thinking?.efforts;
+		for (const effort of THINKING_EFFORTS) {
+			if (efforts?.includes(effort) && routing[effort] === defaultWireId) return effort;
+		}
+	}
+	return minimumSupportedEffort(model);
+}