@oh-my-pi/pi-catalog 17.1.6 → 17.1.8

package/CHANGELOG.md

@@ -2,6 +2,27 @@
 
 ## [Unreleased]
 
+## [17.1.8] - 2026-07-28
+
+### Added
+
+- Added `resolveVertexEndpointHost(location)` utility to resolve the correct Vertex AI API endpoint hostnames for global, multi-region, and regional locations.
+
+### Fixed
+
+- Fixed an issue where `calculateCost` under-reported Anthropic cache-write costs by honoring the `usage.cttl` breakdown to correctly price 1-hour retention writes at 2x the base input rate.
+
+## [17.1.7] - 2026-07-27
+
+### Added
+
+- Added support for moonshotai/Kimi-K3 and kimi-k3-fast models
+- Added umans-kimi-k3 prerelease model configuration
+
+### Changed
+
+- Updated pricing and token limits for selected models
+
 ## [17.1.6] - 2026-07-27
 
 ### Added

package/dist/types/hosts.d.ts

@@ -136,6 +136,19 @@ export declare function modelMatchesHost(model: {
     provider: string;
     baseUrl: string;
 }, host: KnownHost): boolean;
+/**
+ * Hostname for a Vertex AI GenerateContent / rawPredict / OpenAI-compat
+ * request for the given location.
+ *
+ * - `global` → global endpoint (`aiplatform.googleapis.com`)
+ * - `eu` / `us` multi-regions → REP endpoints (`aiplatform.{eu|us}.rep.googleapis.com`)
+ * - every other location → regional (`{location}-aiplatform.googleapis.com`)
+ *
+ * Multi-region codes do NOT follow the regional `{location}-aiplatform` pattern;
+ * interpolating them that way yields hosts like `eu-aiplatform.googleapis.com`
+ * that 404.
+ */
+export declare function resolveVertexEndpointHost(location: string): string;
 /** Vertex AI express-mode OpenAI-compatible endpoint (`…/endpoints/openapi`). */
 export declare function isVertexExpressOpenAIUrl(baseUrl: string): boolean;
 /** Vertex AI Anthropic raw-predict endpoints (`:streamRawPredict` / `:rawPredict`). */

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "17.1.6",
+	"version": "17.1.8",
 	"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": "^2.12.1",
-		"@oh-my-pi/pi-utils": "17.1.6",
+		"@oh-my-pi/pi-utils": "17.1.8",
 		"arktype": "2.2.3",
 		"zod": "^4"
 	},
 	"devDependencies": {
-		"@oh-my-pi/pi-ai": "17.1.6",
+		"@oh-my-pi/pi-ai": "17.1.8",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/hosts.ts

@@ -111,6 +111,24 @@ function includesAsciiCaseInsensitive(value: string, lowerNeedle: string): boole
 
 // --- Endpoint-shape predicates (URL path/verb shapes, not vendor hosts) ---
 
+/**
+ * Hostname for a Vertex AI GenerateContent / rawPredict / OpenAI-compat
+ * request for the given location.
+ *
+ * - `global` → global endpoint (`aiplatform.googleapis.com`)
+ * - `eu` / `us` multi-regions → REP endpoints (`aiplatform.{eu|us}.rep.googleapis.com`)
+ * - every other location → regional (`{location}-aiplatform.googleapis.com`)
+ *
+ * Multi-region codes do NOT follow the regional `{location}-aiplatform` pattern;
+ * interpolating them that way yields hosts like `eu-aiplatform.googleapis.com`
+ * that 404.
+ */
+export function resolveVertexEndpointHost(location: string): string {
+	if (location === "global") return "aiplatform.googleapis.com";
+	if (location === "eu" || location === "us") return `aiplatform.${location}.rep.googleapis.com`;
+	return `${location}-aiplatform.googleapis.com`;
+}
+
 /** Vertex AI express-mode OpenAI-compatible endpoint (`…/endpoints/openapi`). */
 export function isVertexExpressOpenAIUrl(baseUrl: string): boolean {
 	return baseUrl.includes("/endpoints/openapi");

package/src/models.json

@@ -13155,6 +13155,27 @@
 			"supportsComputerUse": false,
 			"supportsComputerUseConfig": false
 		},
+		"moonshotai/Kimi-K3": {
+			"id": "moonshotai/Kimi-K3",
+			"name": "Kimi K3",
+			"api": "openai-completions",
+			"provider": "baseten",
+			"baseUrl": "https://inference.baseten.co/v1",
+			"reasoning": false,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 3,
+				"output": 15,
+				"cacheRead": 0.3,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 1048576,
+			"supportsComputerUse": false
+		},
 		"nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B": {
 			"id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
 			"name": "Nemotron Ultra",
@@ -17111,6 +17132,40 @@
 			"supportsComputerUse": false,
 			"supportsComputerUseConfig": false
 		},
+		"kimi-k3": {
+			"id": "kimi-k3",
+			"name": "Kimi K3",
+			"api": "openai-completions",
+			"provider": "fireworks",
+			"baseUrl": "https://api.fireworks.ai/inference/v1",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 3,
+				"output": 15,
+				"cacheRead": 0.3,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 131072,
+			"thinking": {
+				"mode": "effort",
+				"efforts": [
+					"low",
+					"high",
+					"max"
+				],
+				"defaultLevel": "max",
+				"effortMap": {
+					"max": "max"
+				},
+				"requiresEffort": true
+			},
+			"supportsComputerUse": false
+		},
 		"minimax-m2.5": {
 			"id": "minimax-m2.5",
 			"name": "MiniMax-M2.5",
@@ -27541,7 +27596,8 @@
 			},
 			"contextWindow": null,
 			"maxTokens": null,
-			"supportsComputerUse": false
+			"supportsComputerUse": false,
+			"supportsComputerUseConfig": false
 		},
 		"liquid/lfm-2-24b-a2b": {
 			"id": "liquid/lfm-2-24b-a2b",
@@ -30167,7 +30223,8 @@
 			},
 			"contextWindow": null,
 			"maxTokens": null,
-			"supportsComputerUse": false
+			"supportsComputerUse": false,
+			"supportsComputerUseConfig": false
 		},
 		"openai/gpt-4o:extended": {
 			"id": "openai/gpt-4o:extended",
@@ -64047,6 +64104,38 @@
 				]
 			}
 		},
+		"kimi-k3": {
+			"id": "kimi-k3",
+			"name": "Kimi K3",
+			"api": "ollama-chat",
+			"provider": "ollama-cloud",
+			"baseUrl": "https://ollama.com",
+			"reasoning": true,
+			"thinking": {
+				"mode": "effort",
+				"efforts": [
+					"low",
+					"high",
+					"max"
+				],
+				"defaultLevel": "max",
+				"requiresEffort": true
+			},
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 3,
+				"output": 15,
+				"cacheRead": 0.3,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 8192,
+			"omitMaxOutputTokens": true,
+			"supportsComputerUse": false
+		},
 		"minimax-m2": {
 			"id": "minimax-m2",
 			"name": "minimax-m2",
@@ -68713,6 +68802,39 @@
 				]
 			}
 		},
+		"kimi-k3": {
+			"id": "kimi-k3",
+			"name": "Kimi K3",
+			"api": "openai-completions",
+			"provider": "opencode-zen",
+			"baseUrl": "https://opencode.ai/zen/v1",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 3,
+				"output": 15,
+				"cacheRead": 0.3,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 131072,
+			"thinking": {
+				"mode": "effort",
+				"efforts": [
+					"low",
+					"high",
+					"max"
+				],
+				"defaultLevel": "max",
+				"effortMap": {
+					"max": "max"
+				},
+				"requiresEffort": true
+			}
+		},
 		"laguna-s-2.1-free": {
 			"id": "laguna-s-2.1-free",
 			"name": "Laguna S 2.1 Free",
@@ -72047,8 +72169,8 @@
 				"image"
 			],
 			"cost": {
-				"input": 0.12,
-				"output": 0.35,
+				"input": 0.14,
+				"output": 0.42,
 				"cacheRead": 0.049999999999999996,
 				"cacheWrite": 0
 			},
@@ -74224,9 +74346,9 @@
 				"text"
 			],
 			"cost": {
-				"input": 0.6,
-				"output": 3.5999999999999996,
-				"cacheRead": 0.19999999999999998,
+				"input": 0.5,
+				"output": 2.2,
+				"cacheRead": 0.09999999999999999,
 				"cacheWrite": 0
 			},
 			"contextWindow": 512288,
@@ -75036,11 +75158,11 @@
 			"cost": {
 				"input": 1.25,
 				"output": 10,
-				"cacheRead": 0.125,
+				"cacheRead": 0.13,
 				"cacheWrite": 0
 			},
 			"contextWindow": 128000,
-			"maxTokens": 16384,
+			"maxTokens": 32000,
 			"supportsComputerUse": false,
 			"supportsComputerUseConfig": false
 		},
@@ -75503,10 +75625,10 @@
 				"image"
 			],
 			"cost": {
-				"input": 1,
-				"output": 6,
-				"cacheRead": 0.09999999999999999,
-				"cacheWrite": 1.25
+				"input": 0.5,
+				"output": 3,
+				"cacheRead": 0.049999999999999996,
+				"cacheWrite": 0.625
 			},
 			"contextWindow": 1050000,
 			"maxTokens": 128000,
@@ -75535,10 +75657,10 @@
 				"image"
 			],
 			"cost": {
-				"input": 1,
-				"output": 6,
-				"cacheRead": 0.09999999999999999,
-				"cacheWrite": 1.25
+				"input": 0.5,
+				"output": 3,
+				"cacheRead": 0.049999999999999996,
+				"cacheWrite": 0.625
 			},
 			"contextWindow": 1050000,
 			"maxTokens": 128000,
@@ -75631,10 +75753,10 @@
 				"image"
 			],
 			"cost": {
-				"input": 2.5,
-				"output": 15,
-				"cacheRead": 0.25,
-				"cacheWrite": 3.125
+				"input": 1.25,
+				"output": 7.5,
+				"cacheRead": 0.125,
+				"cacheWrite": 1.5625
 			},
 			"contextWindow": 1050000,
 			"maxTokens": 128000,
@@ -75663,10 +75785,10 @@
 				"image"
 			],
 			"cost": {
-				"input": 2.5,
-				"output": 15,
-				"cacheRead": 0.25,
-				"cacheWrite": 3.125
+				"input": 1.25,
+				"output": 7.5,
+				"cacheRead": 0.125,
+				"cacheWrite": 1.5625
 			},
 			"contextWindow": 1050000,
 			"maxTokens": 128000,
@@ -79676,9 +79798,9 @@
 				"text"
 			],
 			"cost": {
-				"input": 0.6692,
-				"output": 2.1032,
-				"cacheRead": 0.12428,
+				"input": 0.7714000000000001,
+				"output": 2.4244,
+				"cacheRead": 0.14326,
 				"cacheWrite": 0
 			},
 			"contextWindow": 1048576,
@@ -81152,6 +81274,40 @@
 				"escapeBuiltinToolNames": true
 			}
 		},
+		"umans-kimi-k3": {
+			"id": "umans-kimi-k3",
+			"name": "Umans Kimi K3 (prerelease)",
+			"api": "anthropic-messages",
+			"provider": "umans",
+			"baseUrl": "https://api.code.umans.ai",
+			"reasoning": true,
+			"thinking": {
+				"mode": "budget",
+				"efforts": [
+					"minimal",
+					"low",
+					"medium",
+					"high",
+					"xhigh"
+				]
+			},
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 131071,
+			"supportsComputerUse": false,
+			"compat": {
+				"escapeBuiltinToolNames": true
+			}
+		},
 		"umans-qwen3.6-35b-a3b": {
 			"id": "umans-qwen3.6-35b-a3b",
 			"name": "Umans Qwen3.6 35B A3B",
@@ -88444,6 +88600,37 @@
 			},
 			"supportsComputerUse": false
 		},
+		"moonshotai/kimi-k3-fast": {
+			"id": "moonshotai/kimi-k3-fast",
+			"name": "Kimi K3 Fast",
+			"api": "anthropic-messages",
+			"provider": "vercel-ai-gateway",
+			"baseUrl": "https://ai-gateway.vercel.sh",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 4.5,
+				"output": 22.5,
+				"cacheRead": 0.44999999999999996,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1000000,
+			"maxTokens": 131072,
+			"thinking": {
+				"mode": "budget",
+				"efforts": [
+					"low",
+					"high",
+					"max"
+				],
+				"defaultLevel": "max",
+				"requiresEffort": true
+			},
+			"supportsComputerUse": false
+		},
 		"nvidia/nemotron-3-nano-30b-a3b": {
 			"id": "nvidia/nemotron-3-nano-30b-a3b",
 			"name": "Nemotron 3 Nano 30B A3B",
@@ -91404,9 +91591,9 @@
 				"text"
 			],
 			"cost": {
-				"input": 1.75,
-				"output": 5.5,
-				"cacheRead": 0.325,
+				"input": 1.575,
+				"output": 4.95,
+				"cacheRead": 0.2875,
 				"cacheWrite": 0
 			},
 			"contextWindow": 1048576,

package/src/models.ts

@@ -48,10 +48,38 @@ export function calculateCost<TApi extends Api>(model: Model<TApi>, usage: Usage
 	usage.cost.input = (model.cost.input / 1000000) * (usage.input + (orchestration?.input ?? 0));
 	usage.cost.output = (model.cost.output / 1000000) * (usage.output + (orchestration?.output ?? 0));
 	usage.cost.cacheRead = (model.cost.cacheRead / 1000000) * (usage.cacheRead + (orchestration?.cacheRead ?? 0));
-	usage.cost.cacheWrite = (model.cost.cacheWrite / 1000000) * usage.cacheWrite;
+	usage.cost.cacheWrite = cacheWriteCost(model, usage);
 	usage.cost.total = usage.cost.input + usage.cost.output + usage.cost.cacheRead + usage.cost.cacheWrite;
 	return usage.cost;
 }
+
+/**
+ * Price cache-write tokens, honoring the TTL breakdown when the provider reports one.
+ *
+ * `model.cost.cacheWrite` is the 5-minute write rate (Anthropic bills 5m writes at
+ * 1.25x base input). When `usage.cttl` is present the write mixes 5m and 1h
+ * breakpoints — omp defaults to 1h retention on first-party Anthropic, and 1h writes
+ * bill at 2x base input — so each component is priced at its own rate instead of the
+ * flat 5m rate. Deriving 1h from `input * 2` (Anthropic's published multiplier) is
+ * model-independent and stays correct even for legacy entries whose stored
+ * `cacheWrite` scalar drifts from 1.25x input. Providers that omit `cttl`
+ * (everyone but Anthropic) keep the flat-rate calculation.
+ *
+ * The breakdown is documented to sum to `usage.cacheWrite`, but the two are written
+ * from independent wire fields (`cache_creation` vs `cache_creation_input_tokens`),
+ * so any unattributed remainder is priced at the flat rate instead of being dropped:
+ * a partial or stale breakdown must never make write tokens free.
+ */
+function cacheWriteCost<TApi extends Api>(model: Model<TApi>, usage: Usage): number {
+	const rate5m = model.cost.cacheWrite / 1000000;
+	const cttl = usage.cttl;
+	if (!cttl) return rate5m * usage.cacheWrite;
+	const fiveMinute = cttl.ephemeral5m ?? 0;
+	const oneHour = cttl.ephemeral1h ?? 0;
+	const residual = Math.max(0, usage.cacheWrite - fiveMinute - oneHour);
+	return rate5m * (fiveMinute + residual) + ((model.cost.input * 2) / 1000000) * oneHour;
+}
+
 /**
  * Check if two models are equal by comparing both their id and provider.
  * Returns false if either model is null or undefined.