@oh-my-pi/pi-catalog 16.3.0 → 16.3.2

package/CHANGELOG.md

@@ -2,6 +2,19 @@
 
 ## [Unreleased]
 
+## [16.3.2] - 2026-07-02
+
+### Fixed
+
+- Fixed ZenMux model discovery to run without a `ZENMUX_API_KEY`, so newly published ZenMux models (for example `anthropic/claude-fable-5-free`) auto-update into the runtime `models.db` cache instead of waiting on a regenerated `models.json`.
+- Fixed ZenMux runtime discovery to query the `/api/v1/models` endpoint even when the resolved provider base URL points at the Anthropic-compatible route, so discovery no longer requests a non-existent `/api/anthropic/models` path.
+
+## [16.3.1] - 2026-07-02
+
+### Removed
+
+- Removed reasoning suppression prompt logic for GPT-5 models
+
 ## [16.3.0] - 2026-07-02
 
 ### Breaking Changes

package/dist/types/compat/openai.d.ts

@@ -20,8 +20,6 @@ interface OpenAIResponsesSpecLike {
  * as well as URL — bundled `azure` models carry no baseUrl (the deployment host
  * is per-resource, resolved at runtime) — while OpenAI/Copilot developer-role
  * and prompt-cache detection stay URL-keyed, as the historical call sites were.
- * The GPT-5 reasoning-suppression prompt keys on the model name, matching the
- * historical request-time check.
  */
 export declare function buildOpenAIResponsesCompat(spec: OpenAIResponsesSpecLike): ResolvedOpenAIResponsesCompat;
 export declare function buildOpenRouterCompat(spec: ModelSpec<"openrouter">): ResolvedOpenRouterCompat;

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

@@ -407,8 +407,10 @@ export declare const CATALOG_PROVIDERS: readonly [{
     readonly defaultModel: "anthropic/claude-opus-4.8";
     readonly envVars: readonly ["ZENMUX_API_KEY"];
     readonly createModelManagerOptions: (config: ModelManagerConfig) => import("..").ModelManagerOptions<import("..").Api, unknown>;
+    readonly allowUnauthenticated: true;
     readonly catalogDiscovery: {
         readonly label: "ZenMux";
+        readonly allowUnauthenticated: true;
     };
 }, {
     readonly id: "zhipu-coding-plan";

package/dist/types/types.d.ts

@@ -286,14 +286,6 @@ export interface OpenAICompat {
     strictResponsesPairing?: boolean;
     /** Whether the Responses API accepts the `detail: "original"` image hint. Default: auto-detected (false for GitHub Copilot, which rejects it with a 400). */
     supportsImageDetailOriginal?: boolean;
-    /**
-     * Append a trailing no-reasoning developer item when the caller did not
-     * request reasoning, suppressing default reasoning on models that cannot
-     * disable it via request params (Responses APIs only; see
-     * https://community.openai.com/t/need-reasoning-false-option-for-gpt-5/1351588/7).
-     * The prompt must not look like an execution or tool budget. Default: auto-detected (GPT-5-family model names).
-     */
-    requiresReasoningSuppressionPrompt?: boolean;
     /** Whether streamed reasoning deltas for the same field may repeat the full cumulative text snapshot. Default: false. */
     reasoningDeltasMayBeCumulative?: boolean;
     /** Strip leaked DeepSeek chat-template special tokens from visible content deltas. Default: auto-detected. */
@@ -463,7 +455,7 @@ 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" | "thinkingFormat" | "reasoningDisableMode" | "omitReasoningEffort" | "includeEncryptedReasoning" | "filterReasoningHistory" | "disableReasoningOnForcedToolChoice" | "disableReasoningOnToolChoice" | "supportsToolChoice" | "supportsForcedToolChoice" | "supportsNamedToolChoice" | "reasoningContentField" | "requiresReasoningContentForToolCalls" | "requiresReasoningContentForAllAssistantTurns" | "allowsSyntheticReasoningContentForToolCalls" | "replayReasoningContent" | "qwenPreserveThinking" | "requiresThinkingAsText" | "requiresMistralToolIds" | "requiresToolResultName" | "requiresAssistantAfterToolResult" | "requiresAssistantContentForToolCalls" | "stripDeepseekSpecialTokens" | "streamMarkupHealingPattern" | "reasoningDeltasMayBeCumulative" | "emptyLengthFinishIsContextError" | "usesOpenAIToolCallIdLimit" | "promptCacheSessionHeader" | "openRouterRouting" | "isOpenRouterHost" | "supportsStrictMode" | "supportsLongPromptCacheRetention" | "alwaysSendMaxTokens" | "wireModelIdMode" | "vercelGatewayRouting" | "extraBody" | "toolStrictMode" | "toolSchemaFlavor" | "streamIdleTimeoutMs" | "cacheControlFormat" | "thinkingKeep" | "strictResponsesPairing" | "supportsImageDetailOriginal" | "requiresReasoningSuppressionPrompt" | "enableGeminiThinkingLoopGuard" | "whenThinking">> & {
+export type ResolvedOpenAICompat = ResolvedOpenAISharedCompat & Required<Omit<OpenAICompat, "supportsDeveloperRole" | "supportsReasoningEffort" | "reasoningEffortMap" | "supportsReasoningParams" | "thinkingFormat" | "reasoningDisableMode" | "omitReasoningEffort" | "includeEncryptedReasoning" | "filterReasoningHistory" | "disableReasoningOnForcedToolChoice" | "disableReasoningOnToolChoice" | "supportsToolChoice" | "supportsForcedToolChoice" | "supportsNamedToolChoice" | "reasoningContentField" | "requiresReasoningContentForToolCalls" | "requiresReasoningContentForAllAssistantTurns" | "allowsSyntheticReasoningContentForToolCalls" | "replayReasoningContent" | "qwenPreserveThinking" | "requiresThinkingAsText" | "requiresMistralToolIds" | "requiresToolResultName" | "requiresAssistantAfterToolResult" | "requiresAssistantContentForToolCalls" | "stripDeepseekSpecialTokens" | "streamMarkupHealingPattern" | "reasoningDeltasMayBeCumulative" | "emptyLengthFinishIsContextError" | "usesOpenAIToolCallIdLimit" | "promptCacheSessionHeader" | "openRouterRouting" | "isOpenRouterHost" | "supportsStrictMode" | "supportsLongPromptCacheRetention" | "alwaysSendMaxTokens" | "wireModelIdMode" | "vercelGatewayRouting" | "extraBody" | "toolStrictMode" | "toolSchemaFlavor" | "streamIdleTimeoutMs" | "cacheControlFormat" | "thinkingKeep" | "strictResponsesPairing" | "supportsImageDetailOriginal" | "enableGeminiThinkingLoopGuard" | "whenThinking">> & {
     vercelGatewayRouting?: OpenAICompat["vercelGatewayRouting"];
     extraBody?: OpenAICompat["extraBody"];
     cacheControlFormat?: OpenAICompat["cacheControlFormat"];
@@ -482,7 +474,6 @@ export interface ResolvedOpenAIResponsesCompat extends ResolvedOpenAISharedCompa
     supportsLongPromptCacheRetention: boolean;
     strictResponsesPairing: boolean;
     supportsImageDetailOriginal: boolean;
-    requiresReasoningSuppressionPrompt: boolean;
     supportsObfuscationOptOut: boolean;
     streamIdleTimeoutMs?: number;
 }

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "16.3.0",
+	"version": "16.3.2",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -34,12 +34,12 @@
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "^2.12.0",
-		"@oh-my-pi/pi-utils": "16.3.0",
+		"@oh-my-pi/pi-utils": "16.3.2",
 		"arktype": "^2.2.0",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "16.3.0",
+		"@oh-my-pi/pi-ai": "16.3.2",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/openai.ts

@@ -593,8 +593,6 @@ interface OpenAIResponsesSpecLike {
  * as well as URL — bundled `azure` models carry no baseUrl (the deployment host
  * is per-resource, resolved at runtime) — while OpenAI/Copilot developer-role
  * and prompt-cache detection stay URL-keyed, as the historical call sites were.
- * The GPT-5 reasoning-suppression prompt keys on the model name, matching the
- * historical request-time check.
  */
 export function buildOpenAIResponsesCompat(spec: OpenAIResponsesSpecLike): ResolvedOpenAIResponsesCompat {
 	const baseUrl = spec.baseUrl ?? "";
@@ -626,7 +624,6 @@ export function buildOpenAIResponsesCompat(spec: OpenAIResponsesSpecLike): Resol
 		// or base-URL host (mirroring the Anthropic compat builder) so a model
 		// pointed at the Copilot host under a different provider id still clamps.
 		supportsImageDetailOriginal: !modelMatchesHost({ provider: spec.provider, baseUrl }, "githubCopilot"),
-		requiresReasoningSuppressionPrompt: spec.name.toLowerCase().startsWith("gpt-5"),
 		reasoningEffortMap: {},
 		supportsReasoningParams: true,
 		thinkingFormat,
@@ -693,7 +690,6 @@ function pickResponsesOnly(compat: ResolvedOpenAIResponsesCompat): ResponsesOnly
 		supportsLongPromptCacheRetention: compat.supportsLongPromptCacheRetention,
 		strictResponsesPairing: compat.strictResponsesPairing,
 		supportsImageDetailOriginal: compat.supportsImageDetailOriginal,
-		requiresReasoningSuppressionPrompt: compat.requiresReasoningSuppressionPrompt,
 		supportsObfuscationOptOut: compat.supportsObfuscationOptOut,
 	} satisfies ResponsesOnlyCompat;
 }

package/src/provider-models/descriptors.ts

@@ -453,7 +453,8 @@ export const CATALOG_PROVIDERS = [
 		defaultModel: "anthropic/claude-opus-4.8",
 		envVars: ["ZENMUX_API_KEY"],
 		createModelManagerOptions: (config: ModelManagerConfig) => zenmuxModelManagerOptions(config),
-		catalogDiscovery: { label: "ZenMux" },
+		allowUnauthenticated: true,
+		catalogDiscovery: { label: "ZenMux", allowUnauthenticated: true },
 	},
 	{
 		id: "zhipu-coding-plan",

package/src/provider-models/openai-compat.ts

@@ -2035,7 +2035,11 @@ function normalizeZenMuxOpenAiBaseUrl(baseUrl?: string): string {
 	if (!value) {
 		return ZENMUX_OPENAI_BASE_URL;
 	}
-	return value.endsWith("/") ? value.slice(0, -1) : value;
+	const normalized = value.endsWith("/") ? value.slice(0, -1) : value;
+	if (normalized.endsWith("/api/anthropic")) {
+		return normalized.replace("/api/anthropic", "/api/v1");
+	}
+	return normalized;
 }
 
 function toZenMuxAnthropicBaseUrl(openAiBaseUrl: string): string {
@@ -2103,37 +2107,35 @@ export function zenmuxModelManagerOptions(config?: ZenMuxModelManagerConfig): Mo
 	const anthropicBaseUrl = toZenMuxAnthropicBaseUrl(openAiBaseUrl);
 	return {
 		providerId: "zenmux",
-		...(apiKey && {
-			fetchDynamicModels: () =>
-				fetchOpenAICompatibleModels<Api>({
-					api: "openai-completions",
-					provider: "zenmux",
-					baseUrl: openAiBaseUrl,
-					apiKey,
-					mapModel: (entry, defaults) => {
-						const pricings = isRecord(entry.pricings) ? entry.pricings : undefined;
-						const capabilities = isRecord(entry.capabilities) ? entry.capabilities : undefined;
-						const isAnthropicModel = isZenMuxAnthropicModel(entry, defaults.id);
-						return {
-							...defaults,
-							name: toModelName(entry.display_name, defaults.name),
-							api: isAnthropicModel ? "anthropic-messages" : "openai-completions",
-							baseUrl: isAnthropicModel ? anthropicBaseUrl : openAiBaseUrl,
-							reasoning: capabilities?.reasoning === true || defaults.reasoning,
-							input: toInputCapabilities(entry.input_modalities),
-							cost: {
-								input: getZenMuxPricingValue(pricings, "prompt"),
-								output: getZenMuxPricingValue(pricings, "completion"),
-								cacheRead: getZenMuxPricingValue(pricings, "input_cache_read"),
-								cacheWrite: getZenMuxCacheWritePrice(pricings),
-							},
-							contextWindow: toPositiveNumber(entry.context_length, defaults.contextWindow),
-							maxTokens: toPositiveNumber(entry.max_completion_tokens, defaults.maxTokens),
-						};
-					},
-					fetch: config?.fetch,
-				}),
-		}),
+		fetchDynamicModels: () =>
+			fetchOpenAICompatibleModels<Api>({
+				api: "openai-completions",
+				provider: "zenmux",
+				baseUrl: openAiBaseUrl,
+				apiKey,
+				mapModel: (entry, defaults) => {
+					const pricings = isRecord(entry.pricings) ? entry.pricings : undefined;
+					const capabilities = isRecord(entry.capabilities) ? entry.capabilities : undefined;
+					const isAnthropicModel = isZenMuxAnthropicModel(entry, defaults.id);
+					return {
+						...defaults,
+						name: toModelName(entry.display_name, defaults.name),
+						api: isAnthropicModel ? "anthropic-messages" : "openai-completions",
+						baseUrl: isAnthropicModel ? anthropicBaseUrl : openAiBaseUrl,
+						reasoning: capabilities?.reasoning === true || defaults.reasoning,
+						input: toInputCapabilities(entry.input_modalities),
+						cost: {
+							input: getZenMuxPricingValue(pricings, "prompt"),
+							output: getZenMuxPricingValue(pricings, "completion"),
+							cacheRead: getZenMuxPricingValue(pricings, "input_cache_read"),
+							cacheWrite: getZenMuxCacheWritePrice(pricings),
+						},
+						contextWindow: toPositiveNumber(entry.context_length, defaults.contextWindow),
+						maxTokens: toPositiveNumber(entry.max_completion_tokens, defaults.maxTokens),
+					};
+				},
+				fetch: config?.fetch,
+			}),
 	};
 }
 

package/src/types.ts

@@ -324,14 +324,6 @@ export interface OpenAICompat {
 	strictResponsesPairing?: boolean;
 	/** Whether the Responses API accepts the `detail: "original"` image hint. Default: auto-detected (false for GitHub Copilot, which rejects it with a 400). */
 	supportsImageDetailOriginal?: boolean;
-	/**
-	 * Append a trailing no-reasoning developer item when the caller did not
-	 * request reasoning, suppressing default reasoning on models that cannot
-	 * disable it via request params (Responses APIs only; see
-	 * https://community.openai.com/t/need-reasoning-false-option-for-gpt-5/1351588/7).
-	 * The prompt must not look like an execution or tool budget. Default: auto-detected (GPT-5-family model names).
-	 */
-	requiresReasoningSuppressionPrompt?: boolean;
 	/** Whether streamed reasoning deltas for the same field may repeat the full cumulative text snapshot. Default: false. */
 	reasoningDeltasMayBeCumulative?: boolean;
 	/** Strip leaked DeepSeek chat-template special tokens from visible content deltas. Default: auto-detected. */
@@ -557,7 +549,6 @@ export type ResolvedOpenAICompat = ResolvedOpenAISharedCompat &
 			| "thinkingKeep"
 			| "strictResponsesPairing"
 			| "supportsImageDetailOriginal"
-			| "requiresReasoningSuppressionPrompt"
 			| "enableGeminiThinkingLoopGuard"
 			| "whenThinking"
 		>
@@ -581,7 +572,6 @@ export interface ResolvedOpenAIResponsesCompat extends ResolvedOpenAISharedCompa
 	supportsLongPromptCacheRetention: boolean;
 	strictResponsesPairing: boolean;
 	supportsImageDetailOriginal: boolean;
-	requiresReasoningSuppressionPrompt: boolean;
 	supportsObfuscationOptOut: boolean;
 	streamIdleTimeoutMs?: number;
 }