@oh-my-pi/pi-catalog 16.0.3 → 16.0.4

package/CHANGELOG.md

@@ -2,6 +2,12 @@
 
 ## [Unreleased]
 
+## [16.0.4] - 2026-06-17
+
+### Fixed
+
+- Fixed GLM-5.2 catalog thinking metadata for Zhipu/BigModel so the top effort is exposed as `xhigh` and maps to provider-native `max`. ([#2833](https://github.com/can1357/oh-my-pi/issues/2833))
+
 ## [16.0.2] - 2026-06-16
 
 ### Fixed

package/dist/types/identity/family.d.ts

@@ -49,6 +49,8 @@ export declare function isOpenAIModelId(modelId: string): boolean;
  * allowlist.
  */
 export declare function isReasoningGlmModelId(modelId: string): boolean;
+/** GLM-5.2+ coding SKUs accept `reasoning_effort` in addition to binary thinking. */
+export declare function isGlm52ReasoningEffortModelId(modelId: string): boolean;
 /** GLM vision SKUs — the `v` that attaches to the version (`glm-4v`, `glm-4.5v`). */
 export declare function isGlmVisionModelId(modelId: string): boolean;
 /**

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "16.0.3",
+	"version": "16.0.4",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -34,11 +34,11 @@
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "^2.12.0",
-		"@oh-my-pi/pi-utils": "16.0.3",
+		"@oh-my-pi/pi-utils": "16.0.4",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "16.0.3",
+		"@oh-my-pi/pi-ai": "16.0.4",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/openai.ts

@@ -12,6 +12,7 @@ import {
 	isAnthropicNamespacedModelId,
 	isClaudeModelId,
 	isDeepseekModelIdOrName,
+	isGlm52ReasoningEffortModelId,
 	isKimiK26ModelId,
 	isKimiModelId,
 	isMimoModelIdOrName,
@@ -82,6 +83,7 @@ export function buildOpenAICompat(spec: ModelSpec<"openai-completions">): Resolv
 	const isCerebras = modelMatchesHost(hostModel, "cerebras");
 	const isZai = modelMatchesHost(hostModel, "zai");
 	const isZhipu = modelMatchesHost(hostModel, "zhipu");
+	const supportsZaiReasoningEffort = (isZai || isZhipu) && isGlm52ReasoningEffortModelId(spec.id);
 	const isKilo = modelMatchesHost(hostModel, "kilo");
 	const isKimiModel = isKimiModelId(spec.id);
 	const isMoonshotNative = modelMatchesHost(hostModel, "moonshotNative");
@@ -136,6 +138,8 @@ export function buildOpenAICompat(spec: ModelSpec<"openai-completions">): Resolv
 	const useMaxTokens =
 		isMistral ||
 		isMoonshotNative ||
+		isZai ||
+		isZhipu ||
 		hostMatchesUrl(baseUrl, "chutes") ||
 		hostMatchesUrl(baseUrl, "fireworks") ||
 		isDirectDeepseekApi;
@@ -202,7 +206,7 @@ export function buildOpenAICompat(spec: ModelSpec<"openai-completions">): Resolv
 		// OpenAI's reasoning-API surface.
 		supportsDeveloperRole: isOpenAIHost || isAzureHost,
 		supportsMultipleSystemMessages: supportsMultipleSystemMessagesDefault,
-		supportsReasoningEffort: !isGrok && !isZai && !isZhipu && !isXiaomiMimo,
+		supportsReasoningEffort: !isGrok && !isXiaomiMimo && (!(isZai || isZhipu) || supportsZaiReasoningEffort),
 		// GitHub Copilot's chat-completions endpoint rejects reasoning params wholesale.
 		supportsReasoningParams: provider !== "github-copilot",
 		reasoningEffortMap: {},

package/src/identity/family.ts

@@ -105,6 +105,17 @@ export function isReasoningGlmModelId(modelId: string): boolean {
 	}
 	return semverGte(glm.version, "4.5");
 }
+/** GLM-5.2+ coding SKUs accept `reasoning_effort` in addition to binary thinking. */
+export function isGlm52ReasoningEffortModelId(modelId: string): boolean {
+	const glm = parseGlmModel(bareModelId(modelId));
+	if (!glm || glm.vision) {
+		return false;
+	}
+	if (glm.variant !== "base" && glm.variant !== "air" && glm.variant !== "turbo") {
+		return false;
+	}
+	return semverGte(glm.version, "5.2");
+}
 
 /** GLM vision SKUs — the `v` that attaches to the version (`glm-4v`, `glm-4.5v`). */
 export function isGlmVisionModelId(modelId: string): boolean {

package/src/model-thinking.ts

@@ -23,6 +23,7 @@ import {
 import {
 	findThinkingVariantToken,
 	isDeepseekModelIdOrName,
+	isGlm52ReasoningEffortModelId,
 	isMinimaxM2FamilyModelId,
 	isOpenAIGptOssModelId,
 	supportsAdaptiveThinkingDisplay,
@@ -76,6 +77,13 @@ const DEEPSEEK_REASONING_EFFORT_MAP: Readonly<EffortMap> = {
 const FIREWORKS_REASONING_EFFORT_MAP: Readonly<EffortMap> = {
 	[Effort.Minimal]: "none",
 };
+const ZAI_GLM_52_REASONING_EFFORT_MAP: Readonly<EffortMap> = {
+	[Effort.Minimal]: "none",
+	[Effort.Low]: "high",
+	[Effort.Medium]: "high",
+	[Effort.High]: "high",
+	[Effort.XHigh]: "max",
+};
 
 /**
  * Effort → wire-value map for the 5-tier adaptive scale (Opus 4.7+ and
@@ -259,11 +267,19 @@ function sameEffortList(left: readonly Effort[], right: readonly Effort[]): bool
 }
 
 function getModelDefinedEfforts<TApi extends Api>(spec: ModelSpec<TApi>): readonly Effort[] | undefined {
+	if (spec.api === "openai-completions" && isZaiGlm52ReasoningEffortModel(spec)) {
+		return DEFAULT_REASONING_EFFORTS_WITH_XHIGH;
+	}
 	return spec.api === "openai-completions" && (isMinimaxM2FamilyModelId(spec.id) || isOpenAIGptOssModelId(spec.id))
 		? LOW_MEDIUM_HIGH_REASONING_EFFORTS
 		: undefined;
 }
 
+function isZaiGlm52ReasoningEffortModel<TApi extends Api>(spec: ModelSpec<TApi>): boolean {
+	if (!isGlm52ReasoningEffortModelId(spec.id)) return false;
+	return modelMatchesHost(spec, "zai") || modelMatchesHost(spec, "zhipu");
+}
+
 function readCompatEffortMap(compat: CompatOf<Api>): EffortMap | undefined {
 	if (compat === undefined || !("reasoningEffortMap" in compat)) {
 		return undefined;
@@ -288,6 +304,9 @@ function inferDetectedEffortMap<TApi extends Api>(
 	if (spec.provider === "groq" && spec.id === "qwen/qwen3-32b") {
 		return GROQ_QWEN3_32B_REASONING_EFFORT_MAP;
 	}
+	if (isZaiGlm52ReasoningEffortModel(spec)) {
+		return ZAI_GLM_52_REASONING_EFFORT_MAP;
+	}
 	if (isDeepseekReasoningModel(spec)) {
 		return DEEPSEEK_REASONING_EFFORT_MAP;
 	}