@oh-my-pi/pi-catalog 16.0.1 → 16.0.2

package/CHANGELOG.md

@@ -2,6 +2,13 @@
 
 ## [Unreleased]
 
+## [16.0.2] - 2026-06-16
+
+### Fixed
+
+- Fixed Kimi output caps for Umans AI Coding Plan and Venice so discovery metadata cannot use context-sized token ceilings as request caps.
+- Marked Umans Anthropic-compatible models as client-tool escaped so cached and bundled metadata do not expose `web_search` as a provider server tool.
+
 ## [16.0.1] - 2026-06-15
 
 ### Added

package/dist/types/hosts.d.ts

@@ -65,6 +65,10 @@ export declare const KNOWN_HOSTS: {
         readonly providers: readonly ["alibaba-coding-plan"];
         readonly urlMarkers: readonly ["dashscope"];
     };
+    readonly umans: {
+        readonly providers: readonly ["umans"];
+        readonly urlMarkers: readonly ["api.code.umans.ai"];
+    };
     readonly xiaomi: {
         readonly providers: readonly ["xiaomi"];
         readonly providerPrefixes: readonly ["xiaomi-token-plan-"];

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

@@ -172,6 +172,14 @@ export declare function isFireworksKimiK2ModelId(modelId: string): boolean;
  */
 export declare function clampFireworksKimiMaxTokens(modelId: string, candidate: number): number;
 export declare function clampFireworksKimiMaxTokens(modelId: string, candidate: number | null): number | null;
+/**
+ * Kimi K2.7 Code's documented recommended output budget. Some provider
+ * discovery rows report the context-sized `max_completion_tokens` instead.
+ */
+export declare const KIMI_K27_CODE_RECOMMENDED_MAX_TOKENS = 32768;
+export declare function isKimiK27CodeModelId(modelId: string): boolean;
+export declare function clampKimiK27CodeMaxTokens(modelId: string, candidate: number): number;
+export declare function clampKimiK27CodeMaxTokens(modelId: string, candidate: number | null): number | null;
 /**
  * Fireworks DeepSeek V4 accepts effort via `reasoning_effort` but rejects the
  * DeepSeek-native binary `thinking` toggle when both are present.

package/dist/types/types.d.ts

@@ -285,6 +285,13 @@ export interface AnthropicCompat {
      * Default: auto-detected from provider/baseUrl and `model.reasoning`.
      */
     replayUnsignedThinking?: boolean;
+    /**
+     * Prefix Anthropic built-in tool names (`web_search`, `code_execution`, ...)
+     * when they are ordinary client tools. Some Anthropic-compatible gateways
+     * intercept those exact names as server tools and return raw search/result
+     * blocks instead of normal `tool_use` calls.
+     */
+    escapeBuiltinToolNames?: boolean;
 }
 /**
  * OpenRouter provider routing preferences.

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "16.0.1",
+	"version": "16.0.2",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -34,11 +34,11 @@
 	},
 	"dependencies": {
 		"@bufbuild/protobuf": "^2.12.0",
-		"@oh-my-pi/pi-utils": "16.0.1",
+		"@oh-my-pi/pi-utils": "16.0.2",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "16.0.1",
+		"@oh-my-pi/pi-ai": "16.0.2",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/anthropic.ts

@@ -67,6 +67,7 @@ export function buildAnthropicCompat(spec: ModelSpec<"anthropic-messages">): Res
 		// arguments (#2005). Known non-signing hosts (Z.AI, DeepSeek) are also
 		// preserved for compatibility.
 		replayUnsignedThinking: isZai || modelMatchesHost(spec, "deepseekFamily") || (spec.reasoning && !official),
+		escapeBuiltinToolNames: modelMatchesHost(spec, "umans"),
 	};
 	applyCompatOverrides(compat, spec.compat);
 	return compat;

package/src/hosts.ts

@@ -42,6 +42,7 @@ export const KNOWN_HOSTS = {
 	zhipu: { providers: ["zhipu-coding-plan"], urlMarkers: ["open.bigmodel.cn"] },
 	kilo: { providers: ["kilo"], urlMarkers: ["api.kilo.ai"] },
 	alibabaDashscope: { providers: ["alibaba-coding-plan"], urlMarkers: ["dashscope"] },
+	umans: { providers: ["umans"], urlMarkers: ["api.code.umans.ai"] },
 	xiaomi: { providers: ["xiaomi"], providerPrefixes: ["xiaomi-token-plan-"], urlMarkers: ["xiaomimimo.com"] },
 	xai: { providers: ["xai"], urlMarkers: ["api.x.ai"] },
 	mistral: { providers: ["mistral"], urlMarkers: ["mistral.ai"] },

package/src/models.json

@@ -60012,8 +60012,8 @@
 				"text"
 			],
 			"cost": {
-				"input": 0.09,
-				"output": 0.18,
+				"input": 0.098,
+				"output": 0.196,
 				"cacheRead": 0.02,
 				"cacheWrite": 0
 			},
@@ -65288,7 +65288,7 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144,
+			"maxTokens": 81920,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -65311,12 +65311,12 @@
 				"image"
 			],
 			"cost": {
-				"input": 0.39,
-				"output": 2.34,
+				"input": 0.385,
+				"output": 2.4499999999999997,
 				"cacheRead": 0.195,
 				"cacheWrite": 0
 			},
-			"contextWindow": 262144,
+			"contextWindow": 256000,
 			"maxTokens": 65536,
 			"thinking": {
 				"mode": "effort",
@@ -65932,13 +65932,13 @@
 				"text"
 			],
 			"cost": {
-				"input": 0.063,
-				"output": 0.21,
-				"cacheRead": 0.020999999999999998,
+				"input": 0.06599999999999999,
+				"output": 0.26,
+				"cacheRead": 0.029,
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 64000,
+			"maxTokens": 262144,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -68906,7 +68906,7 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144,
+			"maxTokens": 32768,
 			"thinking": {
 				"mode": "budget",
 				"efforts": [
@@ -68916,6 +68916,9 @@
 					"high",
 					"xhigh"
 				]
+			},
+			"compat": {
+				"escapeBuiltinToolNames": true
 			}
 		},
 		"umans-flash": {
@@ -68936,7 +68939,7 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144,
+			"maxTokens": 32768,
 			"thinking": {
 				"mode": "budget",
 				"efforts": [
@@ -68946,11 +68949,14 @@
 					"high",
 					"xhigh"
 				]
+			},
+			"compat": {
+				"escapeBuiltinToolNames": true
 			}
 		},
 		"umans-glm-5.1": {
 			"id": "umans-glm-5.1",
-			"name": "GLM 5.1",
+			"name": "Umans GLM 5.1",
 			"api": "anthropic-messages",
 			"provider": "umans",
 			"baseUrl": "https://api.code.umans.ai",
@@ -68965,7 +68971,7 @@
 				"cacheRead": 0,
 				"cacheWrite": 0
 			},
-			"contextWindow": 204800,
+			"contextWindow": 202752,
 			"maxTokens": 131072,
 			"thinking": {
 				"mode": "budget",
@@ -68976,11 +68982,14 @@
 					"high",
 					"xhigh"
 				]
+			},
+			"compat": {
+				"escapeBuiltinToolNames": true
 			}
 		},
 		"umans-kimi-k2.6": {
 			"id": "umans-kimi-k2.6",
-			"name": "Kimi K2.6",
+			"name": "Umans Kimi K2.6",
 			"api": "anthropic-messages",
 			"provider": "umans",
 			"baseUrl": "https://api.code.umans.ai",
@@ -68996,7 +69005,7 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144,
+			"maxTokens": 32768,
 			"thinking": {
 				"mode": "budget",
 				"efforts": [
@@ -69006,11 +69015,14 @@
 					"high",
 					"xhigh"
 				]
+			},
+			"compat": {
+				"escapeBuiltinToolNames": true
 			}
 		},
 		"umans-kimi-k2.7": {
 			"id": "umans-kimi-k2.7",
-			"name": "Kimi K2.7 Code",
+			"name": "Umans Kimi K2.7 Code",
 			"api": "anthropic-messages",
 			"provider": "umans",
 			"baseUrl": "https://api.code.umans.ai",
@@ -69026,7 +69038,7 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144,
+			"maxTokens": 32768,
 			"thinking": {
 				"mode": "budget",
 				"efforts": [
@@ -69037,11 +69049,14 @@
 					"xhigh"
 				],
 				"requiresEffort": true
+			},
+			"compat": {
+				"escapeBuiltinToolNames": true
 			}
 		},
 		"umans-qwen3.6-35b-a3b": {
 			"id": "umans-qwen3.6-35b-a3b",
-			"name": "Qwen3.6 35B A3B",
+			"name": "Umans Qwen3.6 35B A3B",
 			"api": "anthropic-messages",
 			"provider": "umans",
 			"baseUrl": "https://api.code.umans.ai",
@@ -69057,7 +69072,7 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144,
+			"maxTokens": 32768,
 			"thinking": {
 				"mode": "budget",
 				"efforts": [
@@ -69067,6 +69082,9 @@
 					"high",
 					"xhigh"
 				]
+			},
+			"compat": {
+				"escapeBuiltinToolNames": true
 			}
 		}
 	},
@@ -70539,6 +70557,28 @@
 				]
 			}
 		},
+		"kimi-k2-7-code": {
+			"id": "kimi-k2-7-code",
+			"name": "kimi-k2-7-code",
+			"api": "openai-completions",
+			"provider": "venice",
+			"baseUrl": "https://api.venice.ai/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 256000,
+			"maxTokens": 32768,
+			"compat": {
+				"supportsUsageInStreaming": false
+			}
+		},
 		"kimi-k2-thinking": {
 			"id": "kimi-k2-thinking",
 			"name": "Kimi K2 Thinking",
@@ -72288,7 +72328,8 @@
 				"supportsForcedToolChoice": true,
 				"supportsSamplingParams": true,
 				"requiresToolResultId": false,
-				"replayUnsignedThinking": false
+				"replayUnsignedThinking": false,
+				"escapeBuiltinToolNames": false
 			}
 		},
 		"alibaba/qwen3-max-preview": {

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

@@ -658,13 +658,17 @@ function mapUmansModelInfo(
 		api: "anthropic-messages",
 		provider: "umans",
 		baseUrl,
+		compat: { ...reference?.compat, escapeBuiltinToolNames: true },
 		reasoning: thinking !== undefined,
 		...(thinking ? { thinking } : {}),
 		input: umansSupportsVision(capabilities.supports_vision) ? ["text", "image"] : ["text"],
 		...(supportsTools === false ? { supportsTools: false } : {}),
 		cost: reference?.cost ?? { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
 		contextWindow: toPositiveNumber(capabilities.context_window, reference?.contextWindow ?? null),
-		maxTokens: toPositiveNumber(capabilities.max_completion_tokens, reference?.maxTokens ?? null),
+		maxTokens: toPositiveNumber(
+			capabilities.recommended_max_tokens,
+			toPositiveNumber(capabilities.max_completion_tokens, reference?.maxTokens ?? null),
+		),
 	};
 }
 
@@ -1232,6 +1236,23 @@ export function clampFireworksKimiMaxTokens(modelId: string, candidate: number |
 	return isFireworksKimiK2ModelId(modelId) ? Math.min(candidate, FIREWORKS_KIMI_MAX_TOKENS) : candidate;
 }
 
+/**
+ * Kimi K2.7 Code's documented recommended output budget. Some provider
+ * discovery rows report the context-sized `max_completion_tokens` instead.
+ */
+export const KIMI_K27_CODE_RECOMMENDED_MAX_TOKENS = 32_768;
+
+export function isKimiK27CodeModelId(modelId: string): boolean {
+	return /(?:^|\/)kimi[-._]?k2(?:[._-]?|p)7[-._]?code$/i.test(modelId);
+}
+
+export function clampKimiK27CodeMaxTokens(modelId: string, candidate: number): number;
+export function clampKimiK27CodeMaxTokens(modelId: string, candidate: number | null): number | null;
+export function clampKimiK27CodeMaxTokens(modelId: string, candidate: number | null): number | null {
+	if (candidate === null) return null;
+	return isKimiK27CodeModelId(modelId) ? Math.min(candidate, KIMI_K27_CODE_RECOMMENDED_MAX_TOKENS) : candidate;
+}
+
 /**
  * Fireworks DeepSeek V4 accepts effort via `reasoning_effort` but rejects the
  * DeepSeek-native binary `thinking` toggle when both are present.
@@ -2276,6 +2297,7 @@ export function veniceModelManagerOptions(
 					const model = mapWithBundledReference(entry, defaults, reference);
 					return {
 						...model,
+						maxTokens: clampKimiK27CodeMaxTokens(defaults.id, model.maxTokens),
 						compat: { ...model.compat, supportsUsageInStreaming: false },
 					};
 				},
@@ -3544,7 +3566,12 @@ const MODELS_DEV_PROVIDER_DESCRIPTORS_SPECIALIZED: readonly ModelsDevProviderDes
 	// --- Synthetic ---
 	openAiCompletionsDescriptor("synthetic", "synthetic", "https://api.synthetic.new/openai/v1"),
 	// --- Venice AI ---
-	openAiCompletionsDescriptor("venice", "venice", "https://api.venice.ai/api/v1"),
+	openAiCompletionsDescriptor("venice", "venice", "https://api.venice.ai/api/v1", {
+		transformModel: model => {
+			const maxTokens = clampKimiK27CodeMaxTokens(model.id, model.maxTokens);
+			return maxTokens === model.maxTokens ? model : { ...model, maxTokens };
+		},
+	}),
 	// --- Ollama Cloud ---
 	simpleModelsDevDescriptor("ollama-cloud", "ollama-cloud", "ollama-chat", "https://ollama.com"),
 	// --- Xiaomi Token Plan ---

package/src/types.ts

@@ -311,6 +311,13 @@ export interface AnthropicCompat {
 	 * Default: auto-detected from provider/baseUrl and `model.reasoning`.
 	 */
 	replayUnsignedThinking?: boolean;
+	/**
+	 * Prefix Anthropic built-in tool names (`web_search`, `code_execution`, ...)
+	 * when they are ordinary client tools. Some Anthropic-compatible gateways
+	 * intercept those exact names as server tools and return raw search/result
+	 * blocks instead of normal `tool_use` calls.
+	 */
+	escapeBuiltinToolNames?: boolean;
 }
 
 /**