@adaptic/lumic-utils 1.0.20 → 1.0.22

package/dist/{index-BVl0tRmx.js → index-DfuMX-MS.js}

@@ -99,6 +99,25 @@ const OPENAI_COMPATIBLE_PROVIDERS = {
  * Supported models with their capabilities
  */
 const SUPPORTED_MODELS = {
+    // ── OpenAI GPT-5.5 series (current flagship) ─────────────────────────
+    'gpt-5.5': {
+        provider: 'openai',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: true,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: false,
+    },
+    'gpt-5.5-pro': {
+        provider: 'openai',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: true,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: false,
+    },
     // ── OpenAI GPT-5.4 series ────────────────────────────────────────────
     'gpt-5.4': {
         provider: 'openai',
@@ -221,6 +240,15 @@ const SUPPORTED_MODELS = {
         isReasoningModel: true,
     },
     // ── Anthropic Claude models ──────────────────────────────────────────
+    'claude-opus-4-7': {
+        provider: 'anthropic',
+        supportsTools: true,
+        supportsJson: false,
+        supportsStructuredOutput: false,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: false,
+    },
     'claude-opus-4-6': {
         provider: 'anthropic',
         supportsTools: true,
@@ -249,6 +277,24 @@ const SUPPORTED_MODELS = {
         isReasoningModel: false,
     },
     // ── Deepseek models ──────────────────────────────────────────────────
+    'deepseek-v4-pro': {
+        provider: 'deepseek',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: false,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: true,
+    },
+    'deepseek-v4-flash': {
+        provider: 'deepseek',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: false,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: true,
+    },
     'deepseek-chat': {
         provider: 'deepseek',
         supportsTools: true,
@@ -268,6 +314,24 @@ const SUPPORTED_MODELS = {
         isReasoningModel: true,
     },
     // ── Kimi (Moonshot) models ───────────────────────────────────────────
+    'kimi-k2.6': {
+        provider: 'kimi',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: false,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: false,
+    },
+    'kimi-k2.5': {
+        provider: 'kimi',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: false,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: false,
+    },
     'kimi-k2-0905-preview': {
         provider: 'kimi',
         supportsTools: true,
@@ -333,6 +397,33 @@ const SUPPORTED_MODELS = {
         isReasoningModel: false,
     },
     // ── xAI Grok models ─────────────────────────────────────────────────
+    'grok-4.3': {
+        provider: 'xai',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: true,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: false,
+    },
+    'grok-4.20': {
+        provider: 'xai',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: true,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: false,
+    },
+    'grok-4.1-fast': {
+        provider: 'xai',
+        supportsTools: true,
+        supportsJson: true,
+        supportsStructuredOutput: true,
+        supportsDeveloperPrompt: true,
+        supportsTemperature: true,
+        isReasoningModel: false,
+    },
     'grok-4': {
         provider: 'xai',
         supportsTools: true,
@@ -411,21 +502,41 @@ const SUPPORTED_MODELS = {
  * Common model aliases that map to canonical model names
  */
 const MODEL_ALIASES = {
-    'deepseek-r1': 'deepseek-reasoner',
-    'deepseek-v3': 'deepseek-chat',
-    'claude-4-opus': 'claude-opus-4-6',
+    // OpenAI aliases
+    'gpt-5.5-flagship': 'gpt-5.5',
+    // Anthropic aliases
+    'claude-opus-4.7': 'claude-opus-4-7',
+    'claude-opus-4.6': 'claude-opus-4-6',
+    'claude-sonnet-4.6': 'claude-sonnet-4-6',
+    'claude-haiku-4.5': 'claude-haiku-4-5',
+    'claude-4-opus': 'claude-opus-4-7',
     'claude-4-sonnet': 'claude-sonnet-4-6',
     'claude-4-haiku': 'claude-haiku-4-5',
+    'claude-4.7-opus': 'claude-opus-4-7',
+    'claude-4.6-opus': 'claude-opus-4-6',
+    // Deepseek aliases (legacy names route to V4 Flash thinking modes per DeepSeek migration policy)
+    'deepseek-r1': 'deepseek-reasoner',
+    'deepseek-v3': 'deepseek-chat',
+    'deepseek-v4': 'deepseek-v4-pro',
+    // xAI aliases
     'grok-4-0709': 'grok-4',
+    'grok-3-fast': 'grok-4-fast-non-reasoning',
+    'grok-3-mini-fast': 'grok-3-mini',
+    'grok-4-3': 'grok-4.3',
+    'grok-4-20': 'grok-4.20',
+    // Gemini aliases
     'gemini-pro': 'gemini-3.1-pro-preview',
     'gemini-flash': 'gemini-3-flash-preview',
-    'kimi-k2.5': 'kimi-k2-0905-preview',
-    'moonshot-v1-auto': 'kimi-k2-0905-preview',
+    'gemini-3.1-pro': 'gemini-3.1-pro-preview',
+    'gemini-3-flash': 'gemini-3-flash-preview',
+    'gemini-3.1-flash-lite': 'gemini-3.1-flash-lite-preview',
+    // Kimi aliases
+    'moonshot-v1-auto': 'kimi-k2.6',
+    'kimi-k3': 'kimi-k2.6',
+    // Qwen aliases
     'qwen-plus': 'qwen3.5-plus',
     'qwen-max': 'qwen3-max',
     'qwen-turbo': 'qwen3.5-flash',
-    'grok-3-fast': 'grok-4-fast-non-reasoning',
-    'grok-3-mini-fast': 'grok-3-mini',
 };
 /**
  * Type guard to check if a string is a valid supported model
@@ -455,12 +566,12 @@ function getModelProvider(model) {
  * Default model tiers per provider for use with LLM_MODEL_MINI/NORMAL/ADVANCED env vars
  */
 const PROVIDER_DEFAULT_MODELS = {
-    openai: { mini: 'gpt-5.4-nano', normal: 'gpt-5.4-mini', advanced: 'gpt-5.4' },
-    anthropic: { mini: 'claude-haiku-4-5', normal: 'claude-sonnet-4-6', advanced: 'claude-opus-4-6' },
-    deepseek: { mini: 'deepseek-chat', normal: 'deepseek-chat', advanced: 'deepseek-reasoner' },
-    kimi: { mini: 'kimi-k2-0905-preview', normal: 'kimi-k2-turbo-preview', advanced: 'kimi-k2-thinking-turbo' },
+    openai: { mini: 'gpt-5.4-nano', normal: 'gpt-5.4-mini', advanced: 'gpt-5.5' },
+    anthropic: { mini: 'claude-haiku-4-5', normal: 'claude-sonnet-4-6', advanced: 'claude-opus-4-7' },
+    deepseek: { mini: 'deepseek-v4-flash', normal: 'deepseek-v4-flash', advanced: 'deepseek-v4-pro' },
+    kimi: { mini: 'kimi-k2-0905-preview', normal: 'kimi-k2.5', advanced: 'kimi-k2.6' },
     qwen: { mini: 'qwen3.5-flash', normal: 'qwen3.5-plus', advanced: 'qwen3-max' },
-    xai: { mini: 'grok-3-mini', normal: 'grok-4-fast-non-reasoning', advanced: 'grok-4' },
+    xai: { mini: 'grok-4.1-fast', normal: 'grok-4-fast-non-reasoning', advanced: 'grok-4.3' },
     gemini: { mini: 'gemini-3.1-flash-lite-preview', normal: 'gemini-3-flash-preview', advanced: 'gemini-3.1-pro-preview' },
 };
 
@@ -768,62 +879,98 @@ const DEFAULT_DEVELOPER_PROMPT = `
     Present complete, high-confidence, final answers only. Do not rephrase to be more brief or omit parts of answers.
     Respond only with final content (e.g. code, a json or yaml object, a formatted string, or a markdown document) and nothing else. Do not reply with a preamble, introduction, or conclusion.
 `;
-/** Token costs in USD per token. Last updated Mar 2026. */
+/**
+ * Token costs in USD per token. Last updated May 2026.
+ *
+ * `cacheHitCost` reflects OpenAI's cached-input billing rate (~50% of the
+ * standard input rate per OpenAI's prompt caching documentation). When set,
+ * `calculateCost` splits prompt tokens into cached vs non-cached buckets and
+ * applies the discount; when omitted, cached tokens are billed at full input
+ * rate (a silent ~50% cost overstatement for cache-friendly workloads).
+ */
 const openAiModelCosts = {
+    'gpt-5.5': {
+        inputCost: 5 / 1_000_000,
+        cacheHitCost: 2.5 / 1_000_000,
+        outputCost: 30 / 1_000_000,
+    },
+    'gpt-5.5-pro': {
+        inputCost: 30 / 1_000_000,
+        cacheHitCost: 15 / 1_000_000,
+        outputCost: 180 / 1_000_000,
+    },
     'gpt-5.4': {
         inputCost: 2.5 / 1_000_000,
+        cacheHitCost: 1.25 / 1_000_000,
         outputCost: 15 / 1_000_000,
     },
     'gpt-5.4-mini': {
         inputCost: 0.75 / 1_000_000,
+        cacheHitCost: 0.375 / 1_000_000,
         outputCost: 4.5 / 1_000_000,
     },
     'gpt-5.4-nano': {
         inputCost: 0.2 / 1_000_000,
+        cacheHitCost: 0.1 / 1_000_000,
         outputCost: 1.25 / 1_000_000,
     },
     'gpt-5': {
         inputCost: 2.5 / 1_000_000,
+        cacheHitCost: 1.25 / 1_000_000,
         outputCost: 10 / 1_000_000,
     },
     'gpt-5-mini': {
         inputCost: 0.15 / 1_000_000,
+        cacheHitCost: 0.075 / 1_000_000,
         outputCost: 0.6 / 1_000_000,
     },
     'o1-mini': {
         inputCost: 1.1 / 1_000_000,
+        cacheHitCost: 0.55 / 1_000_000,
         outputCost: 4.4 / 1_000_000,
     },
     'o1': {
         inputCost: 15 / 1_000_000,
+        cacheHitCost: 7.5 / 1_000_000,
         outputCost: 60 / 1_000_000,
     },
     'o3-mini': {
         inputCost: 1.1 / 1_000_000,
+        cacheHitCost: 0.55 / 1_000_000,
         outputCost: 4.4 / 1_000_000,
     },
     'o3': {
         inputCost: 2 / 1_000_000,
+        cacheHitCost: 1 / 1_000_000,
         outputCost: 8 / 1_000_000,
     },
     'gpt-4.1': {
         inputCost: 2 / 1_000_000,
+        cacheHitCost: 1 / 1_000_000,
         outputCost: 8 / 1_000_000,
     },
     'gpt-4.1-mini': {
         inputCost: 0.4 / 1_000_000,
+        cacheHitCost: 0.2 / 1_000_000,
         outputCost: 1.6 / 1_000_000,
     },
     'gpt-4.1-nano': {
         inputCost: 0.1 / 1_000_000,
+        cacheHitCost: 0.05 / 1_000_000,
         outputCost: 0.4 / 1_000_000,
     },
     'o4-mini': {
         inputCost: 1.1 / 1_000_000,
+        cacheHitCost: 0.55 / 1_000_000,
         outputCost: 4.4 / 1_000_000,
     },
 };
 const anthropicModelCosts = {
+    'claude-opus-4-7': {
+        inputCost: 5 / 1_000_000,
+        outputCost: 25 / 1_000_000,
+        cacheHitCost: 0.5 / 1_000_000,
+    },
     'claude-opus-4-6': {
         inputCost: 15 / 1_000_000,
         outputCost: 75 / 1_000_000,
@@ -841,6 +988,16 @@ const anthropicModelCosts = {
     },
 };
 const deepseekModelCosts = {
+    'deepseek-v4-pro': {
+        inputCost: 0.55 / 1_000_000,
+        cacheHitCost: 0.14 / 1_000_000,
+        outputCost: 2.19 / 1_000_000,
+    },
+    'deepseek-v4-flash': {
+        inputCost: 0.27 / 1_000_000,
+        cacheHitCost: 0.07 / 1_000_000,
+        outputCost: 1.1 / 1_000_000,
+    },
     'deepseek-chat': {
         inputCost: 0.27 / 1_000_000,
         cacheHitCost: 0.07 / 1_000_000,
@@ -853,6 +1010,14 @@ const deepseekModelCosts = {
     },
 };
 const kimiModelCosts = {
+    'kimi-k2.6': {
+        inputCost: 0.74 / 1_000_000,
+        outputCost: 3.49 / 1_000_000,
+    },
+    'kimi-k2.5': {
+        inputCost: 0.44 / 1_000_000,
+        outputCost: 2 / 1_000_000,
+    },
     'kimi-k2-0905-preview': {
         inputCost: 0.15 / 1_000_000,
         outputCost: 0.6 / 1_000_000,
@@ -885,6 +1050,21 @@ const qwenModelCosts = {
     },
 };
 const xaiModelCosts = {
+    'grok-4.3': {
+        inputCost: 1.25 / 1_000_000,
+        outputCost: 2.5 / 1_000_000,
+        cacheHitCost: 0.31 / 1_000_000,
+    },
+    'grok-4.20': {
+        inputCost: 2 / 1_000_000,
+        outputCost: 6 / 1_000_000,
+        cacheHitCost: 0.5 / 1_000_000,
+    },
+    'grok-4.1-fast': {
+        inputCost: 0.2 / 1_000_000,
+        outputCost: 0.5 / 1_000_000,
+        cacheHitCost: 0.05 / 1_000_000,
+    },
     'grok-4': {
         inputCost: 3 / 1_000_000,
         outputCost: 15 / 1_000_000,
@@ -911,16 +1091,16 @@ const xaiModelCosts = {
 };
 const geminiModelCosts = {
     'gemini-3.1-pro-preview': {
-        inputCost: 1.25 / 1_000_000,
-        outputCost: 10 / 1_000_000,
+        inputCost: 2 / 1_000_000,
+        outputCost: 12 / 1_000_000,
     },
     'gemini-3-flash-preview': {
-        inputCost: 0.15 / 1_000_000,
-        outputCost: 0.6 / 1_000_000,
+        inputCost: 0.5 / 1_000_000,
+        outputCost: 3 / 1_000_000,
     },
     'gemini-3.1-flash-lite-preview': {
-        inputCost: 0.075 / 1_000_000,
-        outputCost: 0.3 / 1_000_000,
+        inputCost: 0.25 / 1_000_000,
+        outputCost: 1.5 / 1_000_000,
     },
 };
 /**
@@ -1894,7 +2074,10 @@ class LLMCostTracker {
             timestamp: Date.now(),
         };
         this.usageRecords.push(record);
-        getLumicLogger().info(`LLM cost tracked: ${provider}/${model} - $${cost.toFixed(6)}`, { provider, model, inputTokens, outputTokens, cost });
+        // Emit cachedTokens and reasoningTokens explicitly so operators can
+        // verify cache effectiveness from logs alone (the prior log shape only
+        // surfaced inputTokens and outputTokens, hiding the cache discount).
+        getLumicLogger().info(`LLM cost tracked: ${provider}/${model} - $${cost.toFixed(6)}`, { provider, model, inputTokens, cachedTokens: cacheHitTokens, outputTokens, reasoningTokens, cost });
     }
     /**
      * Records usage from an image generation call.
@@ -1975,11 +2158,13 @@ class LLMCostTracker {
         const images = this.getImageCosts();
         let totalCost = 0;
         let totalInputTokens = 0;
+        let totalCacheHitTokens = 0;
         let totalOutputTokens = 0;
         let totalReasoningTokens = 0;
         for (const summary of Object.values(byModel)) {
             totalCost += summary.totalCost;
             totalInputTokens += summary.totalInputTokens;
+            totalCacheHitTokens += summary.totalCacheHitTokens;
             totalOutputTokens += summary.totalOutputTokens;
             totalReasoningTokens += summary.totalReasoningTokens;
         }
@@ -1996,6 +2181,7 @@ class LLMCostTracker {
             totalCost,
             totalCalls: this.usageRecords.length + this.imageRecords.length,
             totalInputTokens,
+            totalCacheHitTokens,
             totalOutputTokens,
             totalReasoningTokens,
             byModel,
@@ -2018,7 +2204,9 @@ class LLMCostTracker {
             cost: `$${m.totalCost.toFixed(6)}`,
             calls: m.callCount,
             inputTokens: m.totalInputTokens,
+            cachedTokens: m.totalCacheHitTokens,
             outputTokens: m.totalOutputTokens,
+            reasoningTokens: m.totalReasoningTokens,
         }));
         const images = Object.values(summary.images).map((img) => ({
             model: img.model,
@@ -22821,11 +23009,11 @@ let poolConfig = DEFAULT_POOL_CONFIG;
 async function loadApolloModules() {
     if (typeof window === "undefined" || process.env.AWS_EXECUTION_ENV) {
         // Server-side (or Lambda): load the CommonJS‑based implementation.
-        return (await Promise.resolve().then(function () { return require('./apollo-client.server-Djh4v__C.js'); }));
+        return (await Promise.resolve().then(function () { return require('./apollo-client.server-BTHAQqYl.js'); }));
     }
     else {
         // Client-side: load the ESM‑based implementation.
-        return (await Promise.resolve().then(function () { return require('./apollo-client.client-DVsbR05r.js'); }));
+        return (await Promise.resolve().then(function () { return require('./apollo-client.client-YzlpL7kf.js'); }));
     }
 }
 /**
@@ -81536,4 +81724,4 @@ exports.withCorrelationId = withCorrelationId;
 exports.withMetrics = withMetrics;
 exports.withRateLimit = withRateLimit;
 exports.withRetry = withRetry;
-//# sourceMappingURL=index-BVl0tRmx.js.map
+//# sourceMappingURL=index-DfuMX-MS.js.map