@agentproto/model-catalog 0.2.0 → 0.4.0

package/dist/{chunk-OHAMH3ZE.mjs → chunk-FCPS2VGK.mjs}

@@ -2032,6 +2032,57 @@ var OPENROUTER_ROUTES = {
   }
 };
 
+// src/llm/context-windows.generated.ts
+var CONTEXT_WINDOWS = {
+  "allam-2-7b": { contextWindow: 4096, maxOutput: 4096, provider: "groq" },
+  "canopylabs/orpheus-arabic-saudi": { contextWindow: 4e3, maxOutput: 5e4, provider: "groq" },
+  "canopylabs/orpheus-v1-english": { contextWindow: 4e3, maxOutput: 5e4, provider: "groq" },
+  "claude-fable-5": { contextWindow: 1e6, maxOutput: 128e3, displayName: "Claude Fable 5", provider: "anthropic" },
+  "claude-haiku-4-5-20251001": { contextWindow: 2e5, maxOutput: 64e3, displayName: "Claude Haiku 4.5", provider: "anthropic" },
+  "claude-opus-4-1-20250805": { contextWindow: 2e5, maxOutput: 32e3, displayName: "Claude Opus 4.1", provider: "anthropic" },
+  "claude-opus-4-5-20251101": { contextWindow: 2e5, maxOutput: 64e3, displayName: "Claude Opus 4.5", provider: "anthropic" },
+  "claude-opus-4-6": { contextWindow: 1e6, maxOutput: 128e3, displayName: "Claude Opus 4.6", provider: "anthropic" },
+  "claude-opus-4-7": { contextWindow: 1e6, maxOutput: 128e3, displayName: "Claude Opus 4.7", provider: "anthropic" },
+  "claude-opus-4-8": { contextWindow: 1e6, maxOutput: 128e3, displayName: "Claude Opus 4.8", provider: "anthropic" },
+  "claude-sonnet-4-5-20250929": { contextWindow: 1e6, maxOutput: 64e3, displayName: "Claude Sonnet 4.5", provider: "anthropic" },
+  "claude-sonnet-4-6": { contextWindow: 1e6, maxOutput: 128e3, displayName: "Claude Sonnet 4.6", provider: "anthropic" },
+  "claude-sonnet-5": { contextWindow: 1e6, maxOutput: 128e3, displayName: "Claude Sonnet 5", provider: "anthropic" },
+  "grok-4.20-0309-non-reasoning": { contextWindow: 1e6, provider: "xai" },
+  "grok-4.20-0309-reasoning": { contextWindow: 1e6, provider: "xai" },
+  "grok-4.20-multi-agent-0309": { contextWindow: 1e6, provider: "xai" },
+  "grok-4.3": { contextWindow: 1e6, provider: "xai" },
+  "grok-4.5": { contextWindow: 5e5, provider: "xai" },
+  "grok-build-0.1": { contextWindow: 256e3, provider: "xai" },
+  "grok-imagine-image": { contextWindow: 8e3, provider: "xai" },
+  "grok-imagine-image-quality": { contextWindow: 8e3, provider: "xai" },
+  "groq/compound": { contextWindow: 131072, maxOutput: 8192, provider: "groq" },
+  "groq/compound-mini": { contextWindow: 131072, maxOutput: 8192, provider: "groq" },
+  "kimi-k2.5": { contextWindow: 262144, provider: "moonshot" },
+  "kimi-k2.6": { contextWindow: 262144, provider: "moonshot" },
+  "kimi-k2.7-code": { contextWindow: 262144, provider: "moonshot" },
+  "kimi-k2.7-code-highspeed": { contextWindow: 262144, provider: "moonshot" },
+  "kimi-k3": { contextWindow: 1048576, provider: "moonshot" },
+  "llama-3.1-8b-instant": { contextWindow: 131072, maxOutput: 131072, provider: "groq" },
+  "llama-3.3-70b-versatile": { contextWindow: 131072, maxOutput: 32768, provider: "groq" },
+  "meta-llama/llama-4-scout-17b-16e-instruct": { contextWindow: 131072, maxOutput: 8192, provider: "groq" },
+  "meta-llama/llama-prompt-guard-2-22m": { contextWindow: 512, maxOutput: 512, provider: "groq" },
+  "meta-llama/llama-prompt-guard-2-86m": { contextWindow: 512, maxOutput: 512, provider: "groq" },
+  "moonshot-v1-128k": { contextWindow: 131072, provider: "moonshot" },
+  "moonshot-v1-128k-vision-preview": { contextWindow: 131072, provider: "moonshot" },
+  "moonshot-v1-32k": { contextWindow: 32768, provider: "moonshot" },
+  "moonshot-v1-32k-vision-preview": { contextWindow: 32768, provider: "moonshot" },
+  "moonshot-v1-8k": { contextWindow: 8192, provider: "moonshot" },
+  "moonshot-v1-8k-vision-preview": { contextWindow: 8192, provider: "moonshot" },
+  "moonshot-v1-auto": { contextWindow: 131072, provider: "moonshot" },
+  "openai/gpt-oss-120b": { contextWindow: 131072, maxOutput: 65536, provider: "groq" },
+  "openai/gpt-oss-20b": { contextWindow: 131072, maxOutput: 65536, provider: "groq" },
+  "openai/gpt-oss-safeguard-20b": { contextWindow: 131072, maxOutput: 65536, provider: "groq" },
+  "qwen/qwen3-32b": { contextWindow: 131072, maxOutput: 40960, provider: "groq" },
+  "qwen/qwen3.6-27b": { contextWindow: 131072, maxOutput: 32768, provider: "groq" },
+  "whisper-large-v3": { contextWindow: 448, maxOutput: 448, provider: "groq" },
+  "whisper-large-v3-turbo": { contextWindow: 448, maxOutput: 448, provider: "groq" }
+};
+
 // src/llm/catalog.ts
 var LLM_PRICING_CATALOG = {
   // ── Anthropic ──────────────────────────────────────────────────────────
@@ -2063,6 +2114,35 @@ var LLM_PRICING_CATALOG = {
     vendor: "anthropic",
     provider: "anthropic"
   },
+  // Latest-generation Anthropic ids the claude-code / claude-sdk adapters
+  // advertise. Registered so `agentproto models` reports a $/Mtok figure and
+  // a concrete provider — before these rows they resolved to provider
+  // `unknown` (no pricing entry) and showed as runnable:false. Prices per 1M
+  // tokens from platform.claude.com/docs/en/pricing.
+  "claude-opus-4-8": {
+    inputPer1M: 5,
+    outputPer1M: 25,
+    cacheReadMultiplier: 0.1,
+    cacheWriteMultiplier: 1.25,
+    vendor: "anthropic",
+    provider: "anthropic"
+  },
+  "claude-sonnet-5": {
+    inputPer1M: 3,
+    outputPer1M: 15,
+    cacheReadMultiplier: 0.1,
+    cacheWriteMultiplier: 1.25,
+    vendor: "anthropic",
+    provider: "anthropic"
+  },
+  "claude-fable-5": {
+    inputPer1M: 10,
+    outputPer1M: 50,
+    cacheReadMultiplier: 0.1,
+    cacheWriteMultiplier: 1.25,
+    vendor: "anthropic",
+    provider: "anthropic"
+  },
   // ── Google ─────────────────────────────────────────────────────────────
   // Source: https://ai.google.dev/gemini-api/docs/pricing (fetched 2026-05-29)
   // — Flash text/image/video input $0.30/1M, output $2.50/1M, context-cache
@@ -2289,6 +2369,202 @@ var LLM_PRICING_CATALOG = {
     vendor: "openai",
     provider: "openai"
   },
+  // ── OpenAI GPT-5.6 series ──────────────────────────────────────────────
+  // Source: OpenRouter pricing (fetched 2026-07-16)
+  "gpt-5.6-luna": {
+    // $1.00 in / $6.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 1,
+    outputPer1M: 6,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.6-luna-pro": {
+    // $1.00 in / $6.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 1,
+    outputPer1M: 6,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.6-terra": {
+    // $2.50 in / $15.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 2.5,
+    outputPer1M: 15,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.6-terra-pro": {
+    // $2.50 in / $15.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 2.5,
+    outputPer1M: 15,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.6-sol": {
+    // $5.00 in / $30.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 5,
+    outputPer1M: 30,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.6-sol-pro": {
+    // $5.00 in / $30.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 5,
+    outputPer1M: 30,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  // ── OpenAI GPT-5 min/nano (lite variants) ──────────────────────────────
+  "gpt-5-mini": {
+    // $0.25 in / $2.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 0.25,
+    outputPer1M: 2,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5-nano": {
+    // $0.05 in / $0.40 out per 1M. Cache read 0.1x.
+    inputPer1M: 0.05,
+    outputPer1M: 0.4,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  // ── OpenAI GPT-5 codex (coding specialist) ─────────────────────────────
+  "gpt-5-codex": {
+    // $1.25 in / $10.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 1.25,
+    outputPer1M: 10,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  // ── OpenAI GPT-5 pro (reasoning-enhanced) ──────────────────────────────
+  "gpt-5-pro": {
+    // $15.00 in / $120.00 out per 1M. No cache discount.
+    inputPer1M: 15,
+    outputPer1M: 120,
+    vendor: "openai",
+    provider: "openai"
+  },
+  // ── OpenAI GPT-5 image generation ──────────────────────────────────────
+  "gpt-5-image": {
+    // $10.00 in / $10.00 out (+ $40/1M image output). Cache read 0.125x.
+    inputPer1M: 10,
+    outputPer1M: 10,
+    cacheReadMultiplier: 0.125,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5-image-mini": {
+    // $2.50 in / $2.00 out (+ $8/1M image output). Cache read 0.1x.
+    inputPer1M: 2.5,
+    outputPer1M: 2,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  // ── OpenAI GPT audio (multimodal) ──────────────────────────────────────
+  "gpt-audio": {
+    // $2.50 in / $10.00 out (+ audio in $32/1M, audio out $64/1M).
+    inputPer1M: 2.5,
+    outputPer1M: 10,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-audio-mini": {
+    // $0.60 in / $2.40 out (+ audio in $0.60/1M, audio out $2.40/1M).
+    inputPer1M: 0.6,
+    outputPer1M: 2.4,
+    vendor: "openai",
+    provider: "openai"
+  },
+  // ── OpenAI o-series (reasoning models) ─────────────────────────────────
+  // Extended reasoning variants with cache support.
+  "o3-pro": {
+    // $20.00 in / $80.00 out per 1M. No cache discount.
+    inputPer1M: 20,
+    outputPer1M: 80,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "o3-deep-research": {
+    // $10.00 in / $40.00 out per 1M. Cache read 0.25x.
+    inputPer1M: 10,
+    outputPer1M: 40,
+    cacheReadMultiplier: 0.25,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "o4-mini-high": {
+    // $1.10 in / $4.40 out per 1M. Cache read 0.25x.
+    inputPer1M: 1.1,
+    outputPer1M: 4.4,
+    cacheReadMultiplier: 0.25,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "o4-mini-deep-research": {
+    // $2.00 in / $8.00 out per 1M. Cache read 0.25x.
+    inputPer1M: 2,
+    outputPer1M: 8,
+    cacheReadMultiplier: 0.25,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "o1-pro": {
+    // $150.00 in / $600.00 out per 1M. No cache discount.
+    inputPer1M: 150,
+    outputPer1M: 600,
+    vendor: "openai",
+    provider: "openai"
+  },
+  // ── OpenAI GPT-5.1/5.2 codex (code specialist) ─────────────────────────
+  "gpt-5.1-codex": {
+    // $1.25 in / $10.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 1.25,
+    outputPer1M: 10,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.1-codex-mini": {
+    // $0.25 in / $2.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 0.25,
+    outputPer1M: 2,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.1-codex-max": {
+    // $1.25 in / $10.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 1.25,
+    outputPer1M: 10,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.2-codex": {
+    // $1.75 in / $14.00 out per 1M. Cache read 0.1x.
+    inputPer1M: 1.75,
+    outputPer1M: 14,
+    cacheReadMultiplier: 0.1,
+    vendor: "openai",
+    provider: "openai"
+  },
+  "gpt-5.2-pro": {
+    // $21.00 in / $168.00 out per 1M. No cache discount.
+    inputPer1M: 21,
+    outputPer1M: 168,
+    vendor: "openai",
+    provider: "openai"
+  },
   // ── MiniMax (direct SDK) ───────────────────────────────────────────────
   // Bare ids match what the MiniMax SDK expects on the wire (PascalCase
   // with optional minor version). OpenRouter equivalents under
@@ -2376,6 +2652,52 @@ var LLM_PRICING_CATALOG = {
     vendor: "moonshot",
     provider: "moonshot"
   },
+  // ── xAI (direct SDK + proxy) ─────────────────────────────────────────
+  // Source: https://docs.x.ai/docs/models (fetched 2026-07-14)
+  // Pricing from API: prompt_text_token_price / completion_text_token_price
+  // in units per 1 token. Converted to $ per 1M tokens.
+  "grok-4.5": {
+    // $2.00 in / $6.00 out per 1M (flagship).
+    inputPer1M: 2,
+    outputPer1M: 6,
+    vendor: "xai",
+    provider: "xai"
+  },
+  "grok-4.20": {
+    // $1.25 in / $2.50 out per 1M (non-reasoning default).
+    inputPer1M: 1.25,
+    outputPer1M: 2.5,
+    vendor: "xai",
+    provider: "xai"
+  },
+  "grok-4.20-reasoning": {
+    // Same pricing as grok-4.20, reasoning variant.
+    inputPer1M: 1.25,
+    outputPer1M: 2.5,
+    vendor: "xai",
+    provider: "xai"
+  },
+  "grok-4.20-multi-agent": {
+    // Same pricing as grok-4.20.
+    inputPer1M: 1.25,
+    outputPer1M: 2.5,
+    vendor: "xai",
+    provider: "xai"
+  },
+  "grok-4.3": {
+    // $1.25 in / $2.50 out per 1M (alias: grok-latest).
+    inputPer1M: 1.25,
+    outputPer1M: 2.5,
+    vendor: "xai",
+    provider: "xai"
+  },
+  "grok-build-0.1": {
+    // $1.00 in / $2.00 out per 1M (code/fast variant, 256k context).
+    inputPer1M: 1,
+    outputPer1M: 2,
+    vendor: "xai",
+    provider: "xai"
+  },
   // ── Mistral (direct SDK) ──────────────────────────────────────────────
   "mistral-large-latest": {
     inputPer1M: 0.5,
@@ -2461,7 +2783,9 @@ var MODEL_ALIASES = {
   "moonshotai/kimi-k2.6": "kimi-k2.6",
   // Mistral
   "mistralai/mistral-large-latest": "mistral-large-latest",
-  "mistralai/mistral-small-latest": "mistral-small-latest"
+  "mistralai/mistral-small-latest": "mistral-small-latest",
+  // xAI — alias grok-latest to grok-4.3 (current default).
+  "grok-latest": "grok-4.3"
 };
 var PRICING_BY_ID = LLM_PRICING_CATALOG;
 var ALIAS_BY_ID = MODEL_ALIASES;
@@ -2483,6 +2807,14 @@ function resolveAlias(modelId) {
   }
   return modelId;
 }
+function resolveContextWindow(modelId) {
+  if (CONTEXT_WINDOWS[modelId]) return CONTEXT_WINDOWS[modelId];
+  const datedMatch = Object.keys(CONTEXT_WINDOWS).filter((key) => key.startsWith(`${modelId}-2`)).sort((a, b) => a.length - b.length)[0];
+  if (datedMatch) return CONTEXT_WINDOWS[datedMatch];
+  const alias = ALIAS_BY_ID[modelId];
+  if (alias && CONTEXT_WINDOWS[alias]) return CONTEXT_WINDOWS[alias];
+  return void 0;
+}
 function getModelProvider(modelId) {
   return resolvePricing(modelId)?.provider;
 }
@@ -2581,49 +2913,6 @@ function getCacheStats(modelId, usage, opts) {
   };
 }
 
-// src/llm/defaults.ts
-var WORKFLOW_FAST_MODEL = "claude-sonnet-4-6";
-var WORKFLOW_PREMIUM_MODEL = "claude-opus-4-6";
-var WORKFLOW_TINY_MODEL = "claude-haiku-4-5-20251001";
-var OPERATOR_AGENT_MODELS = [
-  "claude-opus-4-5",
-  "claude-sonnet-4-5",
-  "gemini-3.1-pro-preview"
-];
-var OPERATOR_DEFAULT_PREMIUM = "claude-opus-4-5";
-var OPERATOR_DEFAULT_STANDARD = "claude-sonnet-4-5";
-var OPERATOR_DEFAULT_GEMINI = "gemini-3.1-pro-preview";
-var LLM_TIERS = {
-  /** Cheap-and-quick. Single-model fallback. */
-  nano: ["mistral-small-creative-latest"],
-  /** Low-cost, agentic-capable. Direct alt-providers. */
-  mini: [
-    "MiniMax-M2.5",
-    "MiniMax-M2.1",
-    "MiniMax-M2",
-    "mistral-large-latest"
-  ],
-  /** Medium cost, fast inference + good general perf. OpenRouter-routed. */
-  lite: [
-    "x-ai/grok-4-fast",
-    "xiaomi/mimo-v2-flash",
-    "google/gemini-3-flash-preview"
-  ],
-  /** High quality, premium features. Mix of direct + OpenRouter. */
-  plus: [
-    "claude-sonnet-4-5",
-    "gpt-4o",
-    "kimi-k2.5",
-    "deepseek/deepseek-v3.2"
-  ],
-  /** Top-tier, best performance. */
-  pro: [
-    "claude-opus-4-5",
-    "google/gemini-3.1-pro-preview",
-    "kimi-k2-thinking"
-  ]
-};
-
-export { DEFAULT_PRICING, LLM_PRICING_CATALOG, LLM_TIERS, MODEL_ALIASES, OPERATOR_AGENT_MODELS, OPERATOR_DEFAULT_GEMINI, OPERATOR_DEFAULT_PREMIUM, OPERATOR_DEFAULT_STANDARD, WORKFLOW_FAST_MODEL, WORKFLOW_PREMIUM_MODEL, WORKFLOW_TINY_MODEL, calculateLLMCreditCost, getCacheStats, getModelProvider, resolveAlias, resolveModelRoute, resolvePricing };
-//# sourceMappingURL=chunk-OHAMH3ZE.mjs.map
-//# sourceMappingURL=chunk-OHAMH3ZE.mjs.map
+export { DEFAULT_PRICING, LLM_PRICING_CATALOG, MODEL_ALIASES, OPENROUTER_ROUTES, calculateLLMCreditCost, getCacheStats, getModelProvider, resolveAlias, resolveContextWindow, resolveModelRoute, resolvePricing };
+//# sourceMappingURL=chunk-FCPS2VGK.mjs.map
+//# sourceMappingURL=chunk-FCPS2VGK.mjs.map