agent-duelist 0.1.0

package/dist/index.cjs

@@ -0,0 +1,2207 @@
+"use strict";
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  // If the importer is in node compatibility mode or this is not an ESM
+  // file that has been converted to a CommonJS file using a Babel-
+  // compatible transform (i.e. "__esModule" has not been set), then set
+  // "default" to the CommonJS "module.exports" for node compatibility.
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// src/index.ts
+var index_exports = {};
+__export(index_exports, {
+  anthropic: () => anthropic,
+  azureOpenai: () => azureOpenai,
+  consoleReporter: () => consoleReporter,
+  defineArena: () => defineArena,
+  gemini: () => gemini,
+  jsonReporter: () => jsonReporter,
+  openai: () => openai,
+  openaiCompatible: () => openaiCompatible,
+  registerPricing: () => registerPricing
+});
+module.exports = __toCommonJS(index_exports);
+
+// src/scorers/latency.ts
+var MIN_MS = 500;
+var MAX_MS = 1e4;
+var latencyScorer = ({ result }) => {
+  const clamped = Math.max(MIN_MS, Math.min(MAX_MS, result.latencyMs));
+  const value = 1 - (clamped - MIN_MS) / (MAX_MS - MIN_MS);
+  return {
+    name: "latency",
+    value: Math.round(value * 100) / 100,
+    details: { ms: result.latencyMs }
+  };
+};
+
+// src/pricing/catalog.json
+var catalog_default = {
+  _meta: {
+    source: "OpenRouter API \u2014 all providers (https://openrouter.ai/api/v1/models)",
+    updatedAt: "2026-02-28",
+    unit: "USD per token"
+  },
+  models: {
+    "ai21/jamba-large-1.7": {
+      inputPerToken: 2e-6,
+      outputPerToken: 8e-6
+    },
+    "aion-labs/aion-1.0": {
+      inputPerToken: 4e-6,
+      outputPerToken: 8e-6
+    },
+    "aion-labs/aion-1.0-mini": {
+      inputPerToken: 7e-7,
+      outputPerToken: 14e-7
+    },
+    "aion-labs/aion-2.0": {
+      inputPerToken: 8e-7,
+      outputPerToken: 16e-7
+    },
+    "aion-labs/aion-rp-llama-3.1-8b": {
+      inputPerToken: 8e-7,
+      outputPerToken: 16e-7
+    },
+    "alfredpros/codellama-7b-instruct-solidity": {
+      inputPerToken: 8e-7,
+      outputPerToken: 12e-7
+    },
+    "alibaba/tongyi-deepresearch-30b-a3b": {
+      inputPerToken: 9e-8,
+      outputPerToken: 45e-8
+    },
+    "allenai/molmo-2-8b": {
+      inputPerToken: 2e-7,
+      outputPerToken: 2e-7
+    },
+    "allenai/olmo-2-0325-32b-instruct": {
+      inputPerToken: 5e-8,
+      outputPerToken: 2e-7
+    },
+    "allenai/olmo-3-32b-think": {
+      inputPerToken: 15e-8,
+      outputPerToken: 5e-7
+    },
+    "allenai/olmo-3-7b-instruct": {
+      inputPerToken: 1e-7,
+      outputPerToken: 2e-7
+    },
+    "allenai/olmo-3-7b-think": {
+      inputPerToken: 12e-8,
+      outputPerToken: 2e-7
+    },
+    "allenai/olmo-3.1-32b-instruct": {
+      inputPerToken: 2e-7,
+      outputPerToken: 6e-7
+    },
+    "alpindale/goliath-120b": {
+      inputPerToken: 375e-8,
+      outputPerToken: 75e-7
+    },
+    "amazon/nova-2-lite-v1": {
+      inputPerToken: 3e-7,
+      outputPerToken: 25e-7
+    },
+    "amazon/nova-lite-v1": {
+      inputPerToken: 6e-8,
+      outputPerToken: 24e-8
+    },
+    "amazon/nova-micro-v1": {
+      inputPerToken: 35e-9,
+      outputPerToken: 14e-8
+    },
+    "amazon/nova-premier-v1": {
+      inputPerToken: 25e-7,
+      outputPerToken: 125e-7
+    },
+    "amazon/nova-pro-v1": {
+      inputPerToken: 8e-7,
+      outputPerToken: 32e-7
+    },
+    "anthracite-org/magnum-v4-72b": {
+      inputPerToken: 3e-6,
+      outputPerToken: 5e-6
+    },
+    "anthropic/claude-3-haiku": {
+      inputPerToken: 25e-8,
+      outputPerToken: 125e-8
+    },
+    "anthropic/claude-3.5-haiku": {
+      inputPerToken: 8e-7,
+      outputPerToken: 4e-6
+    },
+    "anthropic/claude-3.5-sonnet": {
+      inputPerToken: 6e-6,
+      outputPerToken: 3e-5
+    },
+    "anthropic/claude-3.7-sonnet": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "anthropic/claude-3.7-sonnet:thinking": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "anthropic/claude-haiku-4.5": {
+      inputPerToken: 1e-6,
+      outputPerToken: 5e-6
+    },
+    "anthropic/claude-opus-4": {
+      inputPerToken: 15e-6,
+      outputPerToken: 75e-6
+    },
+    "anthropic/claude-opus-4.1": {
+      inputPerToken: 15e-6,
+      outputPerToken: 75e-6
+    },
+    "anthropic/claude-opus-4.5": {
+      inputPerToken: 5e-6,
+      outputPerToken: 25e-6
+    },
+    "anthropic/claude-opus-4.6": {
+      inputPerToken: 5e-6,
+      outputPerToken: 25e-6
+    },
+    "anthropic/claude-sonnet-4": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "anthropic/claude-sonnet-4.5": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "anthropic/claude-sonnet-4.6": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "arcee-ai/coder-large": {
+      inputPerToken: 5e-7,
+      outputPerToken: 8e-7
+    },
+    "arcee-ai/maestro-reasoning": {
+      inputPerToken: 9e-7,
+      outputPerToken: 33e-7
+    },
+    "arcee-ai/spotlight": {
+      inputPerToken: 18e-8,
+      outputPerToken: 18e-8
+    },
+    "arcee-ai/trinity-mini": {
+      inputPerToken: 45e-9,
+      outputPerToken: 15e-8
+    },
+    "arcee-ai/virtuoso-large": {
+      inputPerToken: 75e-8,
+      outputPerToken: 12e-7
+    },
+    "baidu/ernie-4.5-21b-a3b": {
+      inputPerToken: 7e-8,
+      outputPerToken: 28e-8
+    },
+    "baidu/ernie-4.5-21b-a3b-thinking": {
+      inputPerToken: 7e-8,
+      outputPerToken: 28e-8
+    },
+    "baidu/ernie-4.5-300b-a47b": {
+      inputPerToken: 28e-8,
+      outputPerToken: 11e-7
+    },
+    "baidu/ernie-4.5-vl-28b-a3b": {
+      inputPerToken: 14e-8,
+      outputPerToken: 56e-8
+    },
+    "baidu/ernie-4.5-vl-424b-a47b": {
+      inputPerToken: 42e-8,
+      outputPerToken: 125e-8
+    },
+    "bytedance/seed-1.6": {
+      inputPerToken: 25e-8,
+      outputPerToken: 2e-6
+    },
+    "bytedance/seed-1.6-flash": {
+      inputPerToken: 75e-9,
+      outputPerToken: 3e-7
+    },
+    "bytedance/seed-2.0-mini": {
+      inputPerToken: 1e-7,
+      outputPerToken: 4e-7
+    },
+    "bytedance/ui-tars-1.5-7b": {
+      inputPerToken: 1e-7,
+      outputPerToken: 2e-7
+    },
+    "cohere/command-a": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "cohere/command-r-08-2024": {
+      inputPerToken: 15e-8,
+      outputPerToken: 6e-7
+    },
+    "cohere/command-r-plus-08-2024": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "cohere/command-r7b-12-2024": {
+      inputPerToken: 375e-10,
+      outputPerToken: 15e-8
+    },
+    "deepcogito/cogito-v2.1-671b": {
+      inputPerToken: 125e-8,
+      outputPerToken: 125e-8
+    },
+    "deepseek/deepseek-chat": {
+      inputPerToken: 32e-8,
+      outputPerToken: 89e-8
+    },
+    "deepseek/deepseek-chat-v3-0324": {
+      inputPerToken: 2e-7,
+      outputPerToken: 77e-8
+    },
+    "deepseek/deepseek-chat-v3.1": {
+      inputPerToken: 15e-8,
+      outputPerToken: 75e-8
+    },
+    "deepseek/deepseek-r1": {
+      inputPerToken: 7e-7,
+      outputPerToken: 25e-7
+    },
+    "deepseek/deepseek-r1-0528": {
+      inputPerToken: 45e-8,
+      outputPerToken: 215e-8
+    },
+    "deepseek/deepseek-r1-distill-llama-70b": {
+      inputPerToken: 7e-7,
+      outputPerToken: 8e-7
+    },
+    "deepseek/deepseek-r1-distill-qwen-32b": {
+      inputPerToken: 29e-8,
+      outputPerToken: 29e-8
+    },
+    "deepseek/deepseek-v3": {
+      inputPerToken: 3e-7,
+      outputPerToken: 88e-8
+    },
+    "deepseek/deepseek-v3.1-terminus": {
+      inputPerToken: 21e-8,
+      outputPerToken: 79e-8
+    },
+    "deepseek/deepseek-v3.1-terminus:exacto": {
+      inputPerToken: 21e-8,
+      outputPerToken: 79e-8
+    },
+    "deepseek/deepseek-v3.2": {
+      inputPerToken: 25e-8,
+      outputPerToken: 4e-7
+    },
+    "deepseek/deepseek-v3.2-exp": {
+      inputPerToken: 27e-8,
+      outputPerToken: 41e-8
+    },
+    "deepseek/deepseek-v3.2-speciale": {
+      inputPerToken: 4e-7,
+      outputPerToken: 12e-7
+    },
+    "eleutherai/llemma_7b": {
+      inputPerToken: 8e-7,
+      outputPerToken: 12e-7
+    },
+    "essentialai/rnj-1-instruct": {
+      inputPerToken: 15e-8,
+      outputPerToken: 15e-8
+    },
+    "google/gemini-2.0-flash": {
+      inputPerToken: 1e-7,
+      outputPerToken: 4e-7
+    },
+    "google/gemini-2.0-flash-001": {
+      inputPerToken: 1e-7,
+      outputPerToken: 4e-7
+    },
+    "google/gemini-2.0-flash-lite-001": {
+      inputPerToken: 75e-9,
+      outputPerToken: 3e-7
+    },
+    "google/gemini-2.5-flash": {
+      inputPerToken: 3e-7,
+      outputPerToken: 25e-7
+    },
+    "google/gemini-2.5-flash-image": {
+      inputPerToken: 3e-7,
+      outputPerToken: 25e-7
+    },
+    "google/gemini-2.5-flash-lite": {
+      inputPerToken: 1e-7,
+      outputPerToken: 4e-7
+    },
+    "google/gemini-2.5-flash-lite-preview-09-2025": {
+      inputPerToken: 1e-7,
+      outputPerToken: 4e-7
+    },
+    "google/gemini-2.5-pro": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "google/gemini-2.5-pro-preview": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "google/gemini-2.5-pro-preview-05-06": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "google/gemini-3-flash-preview": {
+      inputPerToken: 5e-7,
+      outputPerToken: 3e-6
+    },
+    "google/gemini-3-pro-image-preview": {
+      inputPerToken: 2e-6,
+      outputPerToken: 12e-6
+    },
+    "google/gemini-3-pro-preview": {
+      inputPerToken: 2e-6,
+      outputPerToken: 12e-6
+    },
+    "google/gemini-3.1-flash-image-preview": {
+      inputPerToken: 25e-8,
+      outputPerToken: 15e-7
+    },
+    "google/gemini-3.1-pro-preview": {
+      inputPerToken: 2e-6,
+      outputPerToken: 12e-6
+    },
+    "google/gemini-3.1-pro-preview-customtools": {
+      inputPerToken: 2e-6,
+      outputPerToken: 12e-6
+    },
+    "google/gemma-2-27b-it": {
+      inputPerToken: 65e-8,
+      outputPerToken: 65e-8
+    },
+    "google/gemma-2-9b-it": {
+      inputPerToken: 3e-8,
+      outputPerToken: 9e-8
+    },
+    "google/gemma-3-12b-it": {
+      inputPerToken: 4e-8,
+      outputPerToken: 13e-8
+    },
+    "google/gemma-3-27b-it": {
+      inputPerToken: 4e-8,
+      outputPerToken: 15e-8
+    },
+    "google/gemma-3-4b-it": {
+      inputPerToken: 4e-8,
+      outputPerToken: 8e-8
+    },
+    "google/gemma-3n-e4b-it": {
+      inputPerToken: 2e-8,
+      outputPerToken: 4e-8
+    },
+    "gryphe/mythomax-l2-13b": {
+      inputPerToken: 6e-8,
+      outputPerToken: 6e-8
+    },
+    "ibm-granite/granite-4.0-h-micro": {
+      inputPerToken: 17e-9,
+      outputPerToken: 11e-8
+    },
+    "inception/mercury": {
+      inputPerToken: 25e-8,
+      outputPerToken: 1e-6
+    },
+    "inception/mercury-coder": {
+      inputPerToken: 25e-8,
+      outputPerToken: 1e-6
+    },
+    "inflection/inflection-3-pi": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "inflection/inflection-3-productivity": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "kwaipilot/kat-coder-pro": {
+      inputPerToken: 207e-9,
+      outputPerToken: 828e-9
+    },
+    "liquid/lfm-2-24b-a2b": {
+      inputPerToken: 3e-8,
+      outputPerToken: 12e-8
+    },
+    "liquid/lfm-2.2-6b": {
+      inputPerToken: 1e-8,
+      outputPerToken: 2e-8
+    },
+    "liquid/lfm2-8b-a1b": {
+      inputPerToken: 1e-8,
+      outputPerToken: 2e-8
+    },
+    "mancer/weaver": {
+      inputPerToken: 75e-8,
+      outputPerToken: 1e-6
+    },
+    "meituan/longcat-flash-chat": {
+      inputPerToken: 2e-7,
+      outputPerToken: 8e-7
+    },
+    "meta/llama-3-70b-instruct": {
+      inputPerToken: 51e-8,
+      outputPerToken: 74e-8
+    },
+    "meta/llama-3-8b-instruct": {
+      inputPerToken: 3e-8,
+      outputPerToken: 4e-8
+    },
+    "meta/llama-3.1-405b": {
+      inputPerToken: 4e-6,
+      outputPerToken: 4e-6
+    },
+    "meta/llama-3.1-405b-instruct": {
+      inputPerToken: 4e-6,
+      outputPerToken: 4e-6
+    },
+    "meta/llama-3.1-70b-instruct": {
+      inputPerToken: 4e-7,
+      outputPerToken: 4e-7
+    },
+    "meta/llama-3.1-8b-instruct": {
+      inputPerToken: 2e-8,
+      outputPerToken: 5e-8
+    },
+    "meta/llama-3.2-11b-vision-instruct": {
+      inputPerToken: 49e-9,
+      outputPerToken: 49e-9
+    },
+    "meta/llama-3.2-1b-instruct": {
+      inputPerToken: 27e-9,
+      outputPerToken: 2e-7
+    },
+    "meta/llama-3.2-3b-instruct": {
+      inputPerToken: 2e-8,
+      outputPerToken: 2e-8
+    },
+    "meta/llama-3.3-70b": {
+      inputPerToken: 12e-8,
+      outputPerToken: 3e-7
+    },
+    "meta/llama-3.3-70b-instruct": {
+      inputPerToken: 1e-7,
+      outputPerToken: 32e-8
+    },
+    "meta/llama-4-maverick": {
+      inputPerToken: 15e-8,
+      outputPerToken: 6e-7
+    },
+    "meta/llama-4-scout": {
+      inputPerToken: 8e-8,
+      outputPerToken: 3e-7
+    },
+    "meta/llama-guard-2-8b": {
+      inputPerToken: 2e-7,
+      outputPerToken: 2e-7
+    },
+    "meta/llama-guard-3-8b": {
+      inputPerToken: 2e-8,
+      outputPerToken: 6e-8
+    },
+    "meta/llama-guard-4-12b": {
+      inputPerToken: 18e-8,
+      outputPerToken: 18e-8
+    },
+    "microsoft/phi-4": {
+      inputPerToken: 6e-8,
+      outputPerToken: 14e-8
+    },
+    "microsoft/wizardlm-2-8x22b": {
+      inputPerToken: 62e-8,
+      outputPerToken: 62e-8
+    },
+    "minimax/minimax-01": {
+      inputPerToken: 2e-7,
+      outputPerToken: 11e-7
+    },
+    "minimax/minimax-m1": {
+      inputPerToken: 4e-7,
+      outputPerToken: 22e-7
+    },
+    "minimax/minimax-m2": {
+      inputPerToken: 255e-9,
+      outputPerToken: 1e-6
+    },
+    "minimax/minimax-m2-her": {
+      inputPerToken: 3e-7,
+      outputPerToken: 12e-7
+    },
+    "minimax/minimax-m2.1": {
+      inputPerToken: 27e-8,
+      outputPerToken: 95e-8
+    },
+    "minimax/minimax-m2.5": {
+      inputPerToken: 295e-9,
+      outputPerToken: 12e-7
+    },
+    "mistral/codestral-2508": {
+      inputPerToken: 3e-7,
+      outputPerToken: 9e-7
+    },
+    "mistral/devstral-2512": {
+      inputPerToken: 4e-7,
+      outputPerToken: 2e-6
+    },
+    "mistral/devstral-medium": {
+      inputPerToken: 4e-7,
+      outputPerToken: 2e-6
+    },
+    "mistral/devstral-small": {
+      inputPerToken: 1e-7,
+      outputPerToken: 3e-7
+    },
+    "mistral/ministral-14b-2512": {
+      inputPerToken: 2e-7,
+      outputPerToken: 2e-7
+    },
+    "mistral/ministral-3b-2512": {
+      inputPerToken: 1e-7,
+      outputPerToken: 1e-7
+    },
+    "mistral/ministral-8b-2512": {
+      inputPerToken: 15e-8,
+      outputPerToken: 15e-8
+    },
+    "mistral/mistral-7b-instruct": {
+      inputPerToken: 2e-7,
+      outputPerToken: 2e-7
+    },
+    "mistral/mistral-7b-instruct-v0.1": {
+      inputPerToken: 11e-8,
+      outputPerToken: 19e-8
+    },
+    "mistral/mistral-7b-instruct-v0.3": {
+      inputPerToken: 2e-7,
+      outputPerToken: 2e-7
+    },
+    "mistral/mistral-large": {
+      inputPerToken: 2e-6,
+      outputPerToken: 6e-6
+    },
+    "mistral/mistral-large-2407": {
+      inputPerToken: 2e-6,
+      outputPerToken: 6e-6
+    },
+    "mistral/mistral-large-2411": {
+      inputPerToken: 2e-6,
+      outputPerToken: 6e-6
+    },
+    "mistral/mistral-large-2512": {
+      inputPerToken: 5e-7,
+      outputPerToken: 15e-7
+    },
+    "mistral/mistral-medium-3": {
+      inputPerToken: 4e-7,
+      outputPerToken: 2e-6
+    },
+    "mistral/mistral-medium-3.1": {
+      inputPerToken: 4e-7,
+      outputPerToken: 2e-6
+    },
+    "mistral/mistral-nemo": {
+      inputPerToken: 2e-8,
+      outputPerToken: 4e-8
+    },
+    "mistral/mistral-saba": {
+      inputPerToken: 2e-7,
+      outputPerToken: 6e-7
+    },
+    "mistral/mistral-small": {
+      inputPerToken: 1e-7,
+      outputPerToken: 3e-7
+    },
+    "mistral/mistral-small-24b-instruct-2501": {
+      inputPerToken: 5e-8,
+      outputPerToken: 8e-8
+    },
+    "mistral/mistral-small-3.1-24b-instruct": {
+      inputPerToken: 35e-8,
+      outputPerToken: 56e-8
+    },
+    "mistral/mistral-small-3.2-24b-instruct": {
+      inputPerToken: 6e-8,
+      outputPerToken: 18e-8
+    },
+    "mistral/mistral-small-creative": {
+      inputPerToken: 1e-7,
+      outputPerToken: 3e-7
+    },
+    "mistral/mixtral-8x22b-instruct": {
+      inputPerToken: 2e-6,
+      outputPerToken: 6e-6
+    },
+    "mistral/mixtral-8x7b-instruct": {
+      inputPerToken: 54e-8,
+      outputPerToken: 54e-8
+    },
+    "mistral/pixtral-large-2411": {
+      inputPerToken: 2e-6,
+      outputPerToken: 6e-6
+    },
+    "mistral/voxtral-small-24b-2507": {
+      inputPerToken: 1e-7,
+      outputPerToken: 3e-7
+    },
+    "moonshotai/kimi-k2": {
+      inputPerToken: 55e-8,
+      outputPerToken: 22e-7
+    },
+    "moonshotai/kimi-k2-0905": {
+      inputPerToken: 4e-7,
+      outputPerToken: 2e-6
+    },
+    "moonshotai/kimi-k2-0905:exacto": {
+      inputPerToken: 6e-7,
+      outputPerToken: 25e-7
+    },
+    "moonshotai/kimi-k2-thinking": {
+      inputPerToken: 47e-8,
+      outputPerToken: 2e-6
+    },
+    "moonshotai/kimi-k2.5": {
+      inputPerToken: 45e-8,
+      outputPerToken: 22e-7
+    },
+    "morph/morph-v3-fast": {
+      inputPerToken: 8e-7,
+      outputPerToken: 12e-7
+    },
+    "morph/morph-v3-large": {
+      inputPerToken: 9e-7,
+      outputPerToken: 19e-7
+    },
+    "neversleep/llama-3.1-lumimaid-8b": {
+      inputPerToken: 9e-8,
+      outputPerToken: 6e-7
+    },
+    "neversleep/noromaid-20b": {
+      inputPerToken: 1e-6,
+      outputPerToken: 175e-8
+    },
+    "nex-agi/deepseek-v3.1-nex-n1": {
+      inputPerToken: 27e-8,
+      outputPerToken: 1e-6
+    },
+    "nousresearch/hermes-2-pro-llama-3-8b": {
+      inputPerToken: 14e-8,
+      outputPerToken: 14e-8
+    },
+    "nousresearch/hermes-3-llama-3.1-405b": {
+      inputPerToken: 1e-6,
+      outputPerToken: 1e-6
+    },
+    "nousresearch/hermes-3-llama-3.1-70b": {
+      inputPerToken: 3e-7,
+      outputPerToken: 3e-7
+    },
+    "nousresearch/hermes-4-405b": {
+      inputPerToken: 1e-6,
+      outputPerToken: 3e-6
+    },
+    "nousresearch/hermes-4-70b": {
+      inputPerToken: 13e-8,
+      outputPerToken: 4e-7
+    },
+    "nvidia/llama-3.1-nemotron-70b-instruct": {
+      inputPerToken: 12e-7,
+      outputPerToken: 12e-7
+    },
+    "nvidia/llama-3.3-nemotron-super-49b-v1.5": {
+      inputPerToken: 1e-7,
+      outputPerToken: 4e-7
+    },
+    "nvidia/nemotron-3-nano-30b-a3b": {
+      inputPerToken: 5e-8,
+      outputPerToken: 2e-7
+    },
+    "nvidia/nemotron-nano-12b-v2-vl": {
+      inputPerToken: 2e-7,
+      outputPerToken: 6e-7
+    },
+    "nvidia/nemotron-nano-9b-v2": {
+      inputPerToken: 4e-8,
+      outputPerToken: 16e-8
+    },
+    "openai/gpt-3.5-turbo": {
+      inputPerToken: 5e-7,
+      outputPerToken: 15e-7
+    },
+    "openai/gpt-3.5-turbo-0613": {
+      inputPerToken: 1e-6,
+      outputPerToken: 2e-6
+    },
+    "openai/gpt-3.5-turbo-16k": {
+      inputPerToken: 3e-6,
+      outputPerToken: 4e-6
+    },
+    "openai/gpt-3.5-turbo-instruct": {
+      inputPerToken: 15e-7,
+      outputPerToken: 2e-6
+    },
+    "openai/gpt-4": {
+      inputPerToken: 3e-5,
+      outputPerToken: 6e-5
+    },
+    "openai/gpt-4-0314": {
+      inputPerToken: 3e-5,
+      outputPerToken: 6e-5
+    },
+    "openai/gpt-4-1106-preview": {
+      inputPerToken: 1e-5,
+      outputPerToken: 3e-5
+    },
+    "openai/gpt-4-turbo": {
+      inputPerToken: 1e-5,
+      outputPerToken: 3e-5
+    },
+    "openai/gpt-4-turbo-preview": {
+      inputPerToken: 1e-5,
+      outputPerToken: 3e-5
+    },
+    "openai/gpt-4.1": {
+      inputPerToken: 2e-6,
+      outputPerToken: 8e-6
+    },
+    "openai/gpt-4.1-mini": {
+      inputPerToken: 4e-7,
+      outputPerToken: 16e-7
+    },
+    "openai/gpt-4.1-nano": {
+      inputPerToken: 1e-7,
+      outputPerToken: 4e-7
+    },
+    "openai/gpt-4o": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-4o-2024-05-13": {
+      inputPerToken: 5e-6,
+      outputPerToken: 15e-6
+    },
+    "openai/gpt-4o-2024-08-06": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-4o-2024-11-20": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-4o-audio-preview": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-4o-mini": {
+      inputPerToken: 15e-8,
+      outputPerToken: 6e-7
+    },
+    "openai/gpt-4o-mini-2024-07-18": {
+      inputPerToken: 15e-8,
+      outputPerToken: 6e-7
+    },
+    "openai/gpt-4o-mini-search-preview": {
+      inputPerToken: 15e-8,
+      outputPerToken: 6e-7
+    },
+    "openai/gpt-4o-search-preview": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-4o:extended": {
+      inputPerToken: 6e-6,
+      outputPerToken: 18e-6
+    },
+    "openai/gpt-5": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-5-chat": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-5-codex": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-5-image": {
+      inputPerToken: 1e-5,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-5-image-mini": {
+      inputPerToken: 25e-7,
+      outputPerToken: 2e-6
+    },
+    "openai/gpt-5-mini": {
+      inputPerToken: 25e-8,
+      outputPerToken: 2e-6
+    },
+    "openai/gpt-5-nano": {
+      inputPerToken: 5e-8,
+      outputPerToken: 4e-7
+    },
+    "openai/gpt-5-pro": {
+      inputPerToken: 15e-6,
+      outputPerToken: 12e-5
+    },
+    "openai/gpt-5.1": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-5.1-chat": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-5.1-codex": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-5.1-codex-max": {
+      inputPerToken: 125e-8,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-5.1-codex-mini": {
+      inputPerToken: 25e-8,
+      outputPerToken: 2e-6
+    },
+    "openai/gpt-5.2": {
+      inputPerToken: 175e-8,
+      outputPerToken: 14e-6
+    },
+    "openai/gpt-5.2-chat": {
+      inputPerToken: 175e-8,
+      outputPerToken: 14e-6
+    },
+    "openai/gpt-5.2-codex": {
+      inputPerToken: 175e-8,
+      outputPerToken: 14e-6
+    },
+    "openai/gpt-5.2-pro": {
+      inputPerToken: 21e-6,
+      outputPerToken: 168e-6
+    },
+    "openai/gpt-5.3-codex": {
+      inputPerToken: 175e-8,
+      outputPerToken: 14e-6
+    },
+    "openai/gpt-audio": {
+      inputPerToken: 25e-7,
+      outputPerToken: 1e-5
+    },
+    "openai/gpt-audio-mini": {
+      inputPerToken: 6e-7,
+      outputPerToken: 24e-7
+    },
+    "openai/gpt-oss-120b": {
+      inputPerToken: 39e-9,
+      outputPerToken: 19e-8
+    },
+    "openai/gpt-oss-120b:exacto": {
+      inputPerToken: 39e-9,
+      outputPerToken: 19e-8
+    },
+    "openai/gpt-oss-20b": {
+      inputPerToken: 3e-8,
+      outputPerToken: 14e-8
+    },
+    "openai/gpt-oss-safeguard-20b": {
+      inputPerToken: 75e-9,
+      outputPerToken: 3e-7
+    },
+    "openai/o1": {
+      inputPerToken: 15e-6,
+      outputPerToken: 6e-5
+    },
+    "openai/o1-pro": {
+      inputPerToken: 15e-5,
+      outputPerToken: 6e-4
+    },
+    "openai/o3": {
+      inputPerToken: 2e-6,
+      outputPerToken: 8e-6
+    },
+    "openai/o3-deep-research": {
+      inputPerToken: 1e-5,
+      outputPerToken: 4e-5
+    },
+    "openai/o3-mini": {
+      inputPerToken: 11e-7,
+      outputPerToken: 44e-7
+    },
+    "openai/o3-mini-high": {
+      inputPerToken: 11e-7,
+      outputPerToken: 44e-7
+    },
+    "openai/o3-pro": {
+      inputPerToken: 2e-5,
+      outputPerToken: 8e-5
+    },
+    "openai/o4-mini": {
+      inputPerToken: 11e-7,
+      outputPerToken: 44e-7
+    },
+    "openai/o4-mini-deep-research": {
+      inputPerToken: 2e-6,
+      outputPerToken: 8e-6
+    },
+    "openai/o4-mini-high": {
+      inputPerToken: 11e-7,
+      outputPerToken: 44e-7
+    },
+    "opengvlab/internvl3-78b": {
+      inputPerToken: 15e-8,
+      outputPerToken: 6e-7
+    },
+    "perplexity/sonar": {
+      inputPerToken: 1e-6,
+      outputPerToken: 1e-6
+    },
+    "perplexity/sonar-deep-research": {
+      inputPerToken: 2e-6,
+      outputPerToken: 8e-6
+    },
+    "perplexity/sonar-pro": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "perplexity/sonar-pro-search": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "perplexity/sonar-reasoning-pro": {
+      inputPerToken: 2e-6,
+      outputPerToken: 8e-6
+    },
+    "prime-intellect/intellect-3": {
+      inputPerToken: 2e-7,
+      outputPerToken: 11e-7
+    },
+    "qwen/qwen-2.5-72b-instruct": {
+      inputPerToken: 12e-8,
+      outputPerToken: 39e-8
+    },
+    "qwen/qwen-2.5-7b-instruct": {
+      inputPerToken: 4e-8,
+      outputPerToken: 1e-7
+    },
+    "qwen/qwen-2.5-coder-32b-instruct": {
+      inputPerToken: 20000000000000002e-23,
+      outputPerToken: 20000000000000002e-23
+    },
+    "qwen/qwen-2.5-vl-7b-instruct": {
+      inputPerToken: 20000000000000002e-23,
+      outputPerToken: 20000000000000002e-23
+    },
+    "qwen/qwen-max": {
+      inputPerToken: 16e-7,
+      outputPerToken: 64e-7
+    },
+    "qwen/qwen-plus": {
+      inputPerToken: 4e-7,
+      outputPerToken: 12e-7
+    },
+    "qwen/qwen-plus-2025-07-28": {
+      inputPerToken: 4e-7,
+      outputPerToken: 12e-7
+    },
+    "qwen/qwen-plus-2025-07-28:thinking": {
+      inputPerToken: 4e-7,
+      outputPerToken: 12e-7
+    },
+    "qwen/qwen-turbo": {
+      inputPerToken: 5e-8,
+      outputPerToken: 2e-7
+    },
+    "qwen/qwen-vl-max": {
+      inputPerToken: 8e-7,
+      outputPerToken: 32e-7
+    },
+    "qwen/qwen-vl-plus": {
+      inputPerToken: 21e-8,
+      outputPerToken: 63e-8
+    },
+    "qwen/qwen2.5-coder-7b-instruct": {
+      inputPerToken: 3e-8,
+      outputPerToken: 9e-8
+    },
+    "qwen/qwen2.5-vl-32b-instruct": {
+      inputPerToken: 2e-7,
+      outputPerToken: 6e-7
+    },
+    "qwen/qwen2.5-vl-72b-instruct": {
+      inputPerToken: 8e-7,
+      outputPerToken: 8e-7
+    },
+    "qwen/qwen3-14b": {
+      inputPerToken: 6e-8,
+      outputPerToken: 24e-8
+    },
+    "qwen/qwen3-235b-a22b": {
+      inputPerToken: 455e-9,
+      outputPerToken: 182e-8
+    },
+    "qwen/qwen3-235b-a22b-2507": {
+      inputPerToken: 71e-9,
+      outputPerToken: 1e-7
+    },
+    "qwen/qwen3-30b-a3b": {
+      inputPerToken: 8e-8,
+      outputPerToken: 28e-8
+    },
+    "qwen/qwen3-30b-a3b-instruct-2507": {
+      inputPerToken: 9e-8,
+      outputPerToken: 3e-7
+    },
+    "qwen/qwen3-30b-a3b-thinking-2507": {
+      inputPerToken: 51e-9,
+      outputPerToken: 34e-8
+    },
+    "qwen/qwen3-32b": {
+      inputPerToken: 8e-8,
+      outputPerToken: 24e-8
+    },
+    "qwen/qwen3-8b": {
+      inputPerToken: 5e-8,
+      outputPerToken: 4e-7
+    },
+    "qwen/qwen3-coder": {
+      inputPerToken: 22e-8,
+      outputPerToken: 1e-6
+    },
+    "qwen/qwen3-coder-30b-a3b-instruct": {
+      inputPerToken: 7e-8,
+      outputPerToken: 27e-8
+    },
+    "qwen/qwen3-coder-flash": {
+      inputPerToken: 3e-7,
+      outputPerToken: 15e-7
+    },
+    "qwen/qwen3-coder-next": {
+      inputPerToken: 12e-8,
+      outputPerToken: 75e-8
+    },
+    "qwen/qwen3-coder-plus": {
+      inputPerToken: 1e-6,
+      outputPerToken: 5e-6
+    },
+    "qwen/qwen3-coder:exacto": {
+      inputPerToken: 22e-8,
+      outputPerToken: 18e-7
+    },
+    "qwen/qwen3-max": {
+      inputPerToken: 12e-7,
+      outputPerToken: 6e-6
+    },
+    "qwen/qwen3-max-thinking": {
+      inputPerToken: 12e-7,
+      outputPerToken: 6e-6
+    },
+    "qwen/qwen3-next-80b-a3b-instruct": {
+      inputPerToken: 9e-8,
+      outputPerToken: 11e-7
+    },
+    "qwen/qwen3-next-80b-a3b-thinking": {
+      inputPerToken: 15e-8,
+      outputPerToken: 12e-7
+    },
+    "qwen/qwen3-vl-235b-a22b-instruct": {
+      inputPerToken: 2e-7,
+      outputPerToken: 88e-8
+    },
+    "qwen/qwen3-vl-30b-a3b-instruct": {
+      inputPerToken: 13e-8,
+      outputPerToken: 52e-8
+    },
+    "qwen/qwen3-vl-32b-instruct": {
+      inputPerToken: 104e-9,
+      outputPerToken: 416e-9
+    },
+    "qwen/qwen3-vl-8b-instruct": {
+      inputPerToken: 8e-8,
+      outputPerToken: 5e-7
+    },
+    "qwen/qwen3-vl-8b-thinking": {
+      inputPerToken: 117e-9,
+      outputPerToken: 1365e-9
+    },
+    "qwen/qwen3.5-122b-a10b": {
+      inputPerToken: 4e-7,
+      outputPerToken: 32e-7
+    },
+    "qwen/qwen3.5-27b": {
+      inputPerToken: 3e-7,
+      outputPerToken: 24e-7
+    },
+    "qwen/qwen3.5-35b-a3b": {
+      inputPerToken: 25e-8,
+      outputPerToken: 2e-6
+    },
+    "qwen/qwen3.5-397b-a17b": {
+      inputPerToken: 55e-8,
+      outputPerToken: 35e-7
+    },
+    "qwen/qwen3.5-flash-02-23": {
+      inputPerToken: 1e-7,
+      outputPerToken: 4e-7
+    },
+    "qwen/qwen3.5-plus-02-15": {
+      inputPerToken: 4e-7,
+      outputPerToken: 24e-7
+    },
+    "qwen/qwq-32b": {
+      inputPerToken: 15e-8,
+      outputPerToken: 4e-7
+    },
+    "raifle/sorcererlm-8x22b": {
+      inputPerToken: 45e-7,
+      outputPerToken: 45e-7
+    },
+    "relace/relace-apply-3": {
+      inputPerToken: 85e-8,
+      outputPerToken: 125e-8
+    },
+    "relace/relace-search": {
+      inputPerToken: 1e-6,
+      outputPerToken: 3e-6
+    },
+    "sao10k/l3-euryale-70b": {
+      inputPerToken: 148e-8,
+      outputPerToken: 148e-8
+    },
+    "sao10k/l3-lunaris-8b": {
+      inputPerToken: 4e-8,
+      outputPerToken: 5e-8
+    },
+    "sao10k/l3.1-70b-hanami-x1": {
+      inputPerToken: 3e-6,
+      outputPerToken: 3e-6
+    },
+    "sao10k/l3.1-euryale-70b": {
+      inputPerToken: 65e-8,
+      outputPerToken: 75e-8
+    },
+    "sao10k/l3.3-euryale-70b": {
+      inputPerToken: 65e-8,
+      outputPerToken: 75e-8
+    },
+    "stepfun/step-3.5-flash": {
+      inputPerToken: 1e-7,
+      outputPerToken: 3e-7
+    },
+    "switchpoint/router": {
+      inputPerToken: 85e-8,
+      outputPerToken: 34e-7
+    },
+    "tencent/hunyuan-a13b-instruct": {
+      inputPerToken: 14e-8,
+      outputPerToken: 57e-8
+    },
+    "thedrummer/cydonia-24b-v4.1": {
+      inputPerToken: 3e-7,
+      outputPerToken: 5e-7
+    },
+    "thedrummer/rocinante-12b": {
+      inputPerToken: 17e-8,
+      outputPerToken: 43e-8
+    },
+    "thedrummer/skyfall-36b-v2": {
+      inputPerToken: 55e-8,
+      outputPerToken: 8e-7
+    },
+    "thedrummer/unslopnemo-12b": {
+      inputPerToken: 4e-7,
+      outputPerToken: 4e-7
+    },
+    "tngtech/deepseek-r1t2-chimera": {
+      inputPerToken: 25e-8,
+      outputPerToken: 85e-8
+    },
+    "undi95/remm-slerp-l2-13b": {
+      inputPerToken: 45e-8,
+      outputPerToken: 65e-8
+    },
+    "writer/palmyra-x5": {
+      inputPerToken: 6e-7,
+      outputPerToken: 6e-6
+    },
+    "xai/grok-3": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "xai/grok-3-beta": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "xai/grok-3-mini": {
+      inputPerToken: 3e-7,
+      outputPerToken: 5e-7
+    },
+    "xai/grok-3-mini-beta": {
+      inputPerToken: 3e-7,
+      outputPerToken: 5e-7
+    },
+    "xai/grok-4": {
+      inputPerToken: 3e-6,
+      outputPerToken: 15e-6
+    },
+    "xai/grok-4-fast": {
+      inputPerToken: 2e-7,
+      outputPerToken: 5e-7
+    },
+    "xai/grok-4.1-fast": {
+      inputPerToken: 2e-7,
+      outputPerToken: 5e-7
+    },
+    "xai/grok-code-fast-1": {
+      inputPerToken: 2e-7,
+      outputPerToken: 15e-7
+    },
+    "xiaomi/mimo-v2-flash": {
+      inputPerToken: 9e-8,
+      outputPerToken: 29e-8
+    },
+    "z-ai/glm-4-32b": {
+      inputPerToken: 1e-7,
+      outputPerToken: 1e-7
+    },
+    "z-ai/glm-4.5": {
+      inputPerToken: 55e-8,
+      outputPerToken: 2e-6
+    },
+    "z-ai/glm-4.5-air": {
+      inputPerToken: 13e-8,
+      outputPerToken: 85e-8
+    },
+    "z-ai/glm-4.5v": {
+      inputPerToken: 6e-7,
+      outputPerToken: 18e-7
+    },
+    "z-ai/glm-4.6": {
+      inputPerToken: 35e-8,
+      outputPerToken: 171e-8
+    },
+    "z-ai/glm-4.6:exacto": {
+      inputPerToken: 44e-8,
+      outputPerToken: 176e-8
+    },
+    "z-ai/glm-4.6v": {
+      inputPerToken: 3e-7,
+      outputPerToken: 9e-7
+    },
+    "z-ai/glm-4.7": {
+      inputPerToken: 3e-7,
+      outputPerToken: 14e-7
+    },
+    "z-ai/glm-4.7-flash": {
+      inputPerToken: 6e-8,
+      outputPerToken: 4e-7
+    },
+    "z-ai/glm-5": {
+      inputPerToken: 95e-8,
+      outputPerToken: 255e-8
+    }
+  }
+};
+
+// src/pricing/lookup.ts
+var models = catalog_default.models;
+var modelNameIndex = /* @__PURE__ */ new Map();
+for (const key of Object.keys(models)) {
+  const name = key.split("/").slice(1).join("/");
+  if (name && !modelNameIndex.has(name)) {
+    modelNameIndex.set(name, key);
+  }
+}
+function lookupPricing(providerId) {
+  if (models[providerId]) return models[providerId];
+  const model = providerId.split("/").slice(1).join("/");
+  if (!model) return void 0;
+  const asOpenai = `openai/${model}`;
+  if (models[asOpenai]) return models[asOpenai];
+  const crossKey = modelNameIndex.get(model);
+  if (crossKey) return models[crossKey];
+  return void 0;
+}
+function registerPricing(providerId, pricing) {
+  models[providerId] = pricing;
+}
+function estimateCost(pricing, promptTokens, completionTokens) {
+  return pricing.inputPerToken * promptTokens + pricing.outputPerToken * completionTokens;
+}
+
+// src/scorers/cost.ts
+var costScorer = ({ result }, providerId) => {
+  const promptTokens = result.usage?.promptTokens ?? 0;
+  const completionTokens = result.usage?.completionTokens ?? 0;
+  const totalTokens = promptTokens + completionTokens;
+  const pricing = lookupPricing(providerId);
+  if (!pricing) {
+    return {
+      name: "cost",
+      value: -1,
+      details: {
+        estimatedUsd: null,
+        promptTokens,
+        completionTokens,
+        totalTokens,
+        note: "No pricing data available for this model"
+      }
+    };
+  }
+  const usd = estimateCost(pricing, promptTokens, completionTokens);
+  return {
+    name: "cost",
+    value: usd,
+    details: {
+      estimatedUsd: usd,
+      promptTokens,
+      completionTokens,
+      totalTokens
+    }
+  };
+};
+
+// src/scorers/correctness.ts
+var correctnessScorer = ({ task, result }) => {
+  if (task.expected === void 0) {
+    return { name: "correctness", value: 0.5, details: { reason: "no expected value" } };
+  }
+  const match = deepEqual(task.expected, result.output);
+  return {
+    name: "correctness",
+    value: match ? 1 : 0,
+    details: { expected: task.expected, actual: result.output }
+  };
+};
+function deepEqual(a, b) {
+  if (a === b) return true;
+  if (typeof a === "string" && typeof b === "string") {
+    return a.trim().toLowerCase() === b.trim().toLowerCase();
+  }
+  if (typeof a !== typeof b) return false;
+  if (a === null || b === null) return a === b;
+  if (Array.isArray(a) && Array.isArray(b)) {
+    if (a.length !== b.length) return false;
+    return a.every((val, i) => deepEqual(val, b[i]));
+  }
+  if (typeof a === "object" && typeof b === "object") {
+    const objA = a;
+    const objB = b;
+    const keysA = Object.keys(objA);
+    const keysB = Object.keys(objB);
+    if (keysA.length !== keysB.length) return false;
+    return keysA.every((key) => key in objB && deepEqual(objA[key], objB[key]));
+  }
+  return a === b;
+}
+
+// src/scorers/schema-correctness.ts
+var schemaCorrectnessScorer = ({ task, result }) => {
+  if (!task.schema) {
+    return { name: "schema-correctness", value: -1, details: { reason: "no schema defined" } };
+  }
+  let data = result.output;
+  if (typeof data === "string") {
+    try {
+      data = JSON.parse(data);
+    } catch {
+      return {
+        name: "schema-correctness",
+        value: 0,
+        details: { reason: "output is not valid JSON" }
+      };
+    }
+  }
+  const parsed = task.schema.safeParse(data);
+  return {
+    name: "schema-correctness",
+    value: parsed.success ? 1 : 0,
+    details: parsed.success ? { valid: true } : { valid: false, errors: parsed.error.issues.map((i) => i.message) }
+  };
+};
+
+// src/scorers/fuzzy-similarity.ts
+var fuzzySimilarityScorer = ({ task, result }) => {
+  if (task.expected === void 0) {
+    return { name: "fuzzy-similarity", value: -1, details: { reason: "no expected value" } };
+  }
+  const a = stringify(task.expected);
+  const b = stringify(result.output);
+  const similarity = jaccardSimilarity(tokenize(a), tokenize(b));
+  return {
+    name: "fuzzy-similarity",
+    value: Math.round(similarity * 100) / 100,
+    details: { method: "jaccard", expectedTokens: tokenize(a).size, actualTokens: tokenize(b).size }
+  };
+};
+function stringify(value) {
+  if (typeof value === "string") return value.toLowerCase();
+  return JSON.stringify(value).toLowerCase();
+}
+function tokenize(text) {
+  return new Set(text.match(/\w+/g) ?? []);
+}
+function jaccardSimilarity(a, b) {
+  if (a.size === 0 && b.size === 0) return 1;
+  let intersection = 0;
+  for (const token of a) {
+    if (b.has(token)) intersection++;
+  }
+  const union = a.size + b.size - intersection;
+  return union === 0 ? 1 : intersection / union;
+}
+
+// src/scorers/llm-judge.ts
+var import_openai = __toESM(require("openai"), 1);
+var JUDGE_PROMPT = `You are a strict scoring judge. Evaluate the actual output against the expected output on three criteria. Score each from 0.0 to 1.0 using the full range (not just 0, 0.5, 1).
+
+Criteria:
+1. Accuracy \u2014 are the facts, entities, and claims correct? Penalize hallucinations or wrong details.
+2. Completeness \u2014 does it capture all key information from the expected output? Penalize missing points.
+3. Conciseness \u2014 is it free of unnecessary filler, repetition, or tangential content? Penalize verbosity.
+
+Respond with ONLY this exact format \u2014 three lines, no other text:
+accuracy: <number>
+completeness: <number>
+conciseness: <number>
+
+Task: {task}
+Expected: {expected}
+Actual: {actual}`;
+function resolveJudgeClient(configModel) {
+  const model = configModel ?? process.env.DUELIST_JUDGE_MODEL ?? "gpt-4o-mini";
+  if (model.startsWith("gemini") && process.env.GOOGLE_API_KEY) {
+    return {
+      client: new import_openai.default({
+        apiKey: process.env.GOOGLE_API_KEY,
+        baseURL: "https://generativelanguage.googleapis.com/v1beta/openai/"
+      }),
+      model
+    };
+  }
+  if (!process.env.OPENAI_API_KEY && process.env.AZURE_OPENAI_API_KEY) {
+    return {
+      client: new import_openai.AzureOpenAI({
+        apiKey: process.env.AZURE_OPENAI_API_KEY,
+        endpoint: process.env.AZURE_OPENAI_ENDPOINT,
+        apiVersion: process.env.AZURE_OPENAI_API_VERSION ?? "2024-12-01-preview",
+        deployment: model
+      }),
+      model
+    };
+  }
+  const apiKey = process.env.OPENAI_API_KEY;
+  if (!apiKey) return void 0;
+  return { client: new import_openai.default({ apiKey }), model };
+}
+function createLlmJudgeScorer(judgeModel) {
+  let cached = void 0;
+  return async ({ task, result }) => {
+    if (task.expected === void 0) {
+      return { name: "llm-judge-correctness", value: -1, details: { reason: "no expected value" } };
+    }
+    if (cached === void 0) {
+      cached = resolveJudgeClient(judgeModel) ?? null;
+    }
+    if (!cached) {
+      return {
+        name: "llm-judge-correctness",
+        value: -1,
+        details: { reason: "no API key available for judge model" }
+      };
+    }
+    const { client, model } = cached;
+    const prompt = JUDGE_PROMPT.replace("{task}", task.prompt).replace("{expected}", JSON.stringify(task.expected)).replace("{actual}", JSON.stringify(result.output));
+    try {
+      const response = await client.chat.completions.create({
+        model,
+        messages: [{ role: "user", content: prompt }],
+        temperature: 0,
+        max_tokens: 2048
+      });
+      const content = response.choices[0]?.message?.content?.trim() ?? "";
+      const parsed = {};
+      for (const line of content.split("\n")) {
+        const match = line.match(/^(accuracy|completeness|conciseness)\s*:\s*([\d.]+)/i);
+        if (match) parsed[match[1].toLowerCase()] = parseFloat(match[2]);
+      }
+      const accuracy = parsed.accuracy;
+      const completeness = parsed.completeness;
+      const conciseness = parsed.conciseness;
+      if (accuracy == null || completeness == null || conciseness == null || [accuracy, completeness, conciseness].some((s) => isNaN(s) || s < 0 || s > 1)) {
+        return {
+          name: "llm-judge-correctness",
+          value: -1,
+          details: { reason: `judge returned unparseable scores: "${content}"`, model }
+        };
+      }
+      const composite = Math.round((accuracy + completeness + conciseness) / 3 * 100) / 100;
+      return {
+        name: "llm-judge-correctness",
+        value: composite,
+        details: { model, accuracy, completeness, conciseness }
+      };
+    } catch (err) {
+      return {
+        name: "llm-judge-correctness",
+        value: -1,
+        details: { reason: `judge call failed: ${err instanceof Error ? err.message : String(err)}` }
+      };
+    }
+  };
+}
+
+// src/scorers/tool-usage.ts
+var toolUsageScorer = ({ task, result }) => {
+  const expectedToolName = task.tools?.[0]?.name;
+  if (!expectedToolName) {
+    return { name: "tool-usage", value: -1, details: { reason: "no tools configured on task" } };
+  }
+  const usedTool = result.toolCalls?.some((c) => c.name === expectedToolName) ?? false;
+  return {
+    name: "tool-usage",
+    value: usedTool ? 1 : 0,
+    details: { expectedToolName, usedTool, toolCalls: result.toolCalls ?? [] }
+  };
+};
+
+// src/scorers/index.ts
+var staticScorers = {
+  latency: latencyScorer,
+  cost: costScorer,
+  correctness: correctnessScorer,
+  "schema-correctness": schemaCorrectnessScorer,
+  "fuzzy-similarity": fuzzySimilarityScorer,
+  "tool-usage": toolUsageScorer
+};
+function resolveScorers(names, judgeModel) {
+  return names.map((name) => {
+    if (name === "llm-judge-correctness") {
+      return createLlmJudgeScorer(judgeModel);
+    }
+    const scorer = staticScorers[name];
+    if (!scorer) {
+      throw new Error(`Unknown scorer: "${name}"`);
+    }
+    return scorer;
+  });
+}
+
+// src/runner.ts
+async function runBenchmarks(options) {
+  const { providers, tasks, scorers, runs, onResult } = options;
+  const results = [];
+  for (const task of tasks) {
+    for (const provider of providers) {
+      for (let run = 1; run <= runs; run++) {
+        let result;
+        try {
+          const taskResult = await provider.run({
+            prompt: task.prompt,
+            schema: task.schema,
+            tools: task.tools
+          });
+          const scores = await Promise.all(
+            scorers.map((scorer) => scorer({ task, result: taskResult }, provider.id))
+          );
+          result = {
+            providerId: provider.id,
+            taskName: task.name,
+            run,
+            scores,
+            raw: {
+              output: taskResult.output,
+              latencyMs: taskResult.latencyMs,
+              usage: taskResult.usage,
+              toolCalls: taskResult.toolCalls
+            }
+          };
+        } catch (err) {
+          const message = err instanceof Error ? err.message : String(err);
+          result = {
+            providerId: provider.id,
+            taskName: task.name,
+            run,
+            scores: [],
+            error: message,
+            raw: { output: "", latencyMs: 0 }
+          };
+        }
+        results.push(result);
+        onResult?.(result);
+      }
+    }
+  }
+  return results;
+}
+
+// src/reporter/console.ts
+var reset = "\x1B[0m";
+var boldCode = "\x1B[1m";
+var dimCode = "\x1B[2m";
+var green = "\x1B[32m";
+var red = "\x1B[31m";
+var yellow = "\x1B[33m";
+var cyan = "\x1B[36m";
+function bold(s) {
+  return `${boldCode}${s}${reset}`;
+}
+function dim(s) {
+  return `${dimCode}${s}${reset}`;
+}
+function colorScore(value) {
+  const pct = Math.round(value * 100);
+  const str = `${pct}%`;
+  if (value >= 0.8) return `${green}${str}${reset}`;
+  if (value >= 0.5) return `${yellow}${str}${reset}`;
+  return `${red}${str}${reset}`;
+}
+function consoleReporter(results) {
+  if (results.length === 0) {
+    console.log("\nNo results to display.\n");
+    return;
+  }
+  const tasks = [...new Set(results.map((r) => r.taskName))];
+  const providers = [...new Set(results.map((r) => r.providerId))];
+  const scorerNames = [...new Set(results.flatMap((r) => r.scores.map((s) => s.name)))];
+  const hasCost = scorerNames.includes("cost");
+  const hasErrors = results.some((r) => r.error);
+  const runsPerCell = Math.max(...results.map((r) => r.run));
+  const runLabel = runsPerCell > 1 ? ` (${runsPerCell} runs each)` : "";
+  console.log("");
+  console.log(`  ${bold(`\u2B21 Agent Duelist Results${runLabel}`)}`);
+  console.log(`  ${dim("\u2500".repeat(70))}`);
+  console.log("");
+  for (const task of tasks) {
+    console.log(`  ${bold(`Task: ${task}`)}`);
+    const cols = [{ label: "Provider", width: 22, align: "left" }];
+    for (const name of scorerNames) {
+      if (name === "latency") cols.push({ label: "Latency", width: 10, align: "right" });
+      else if (name === "cost") {
+        cols.push({ label: "Cost", width: 12, align: "right" });
+        cols.push({ label: "Tokens", width: 9, align: "right" });
+      } else if (name === "correctness") cols.push({ label: "Match", width: 8, align: "right" });
+      else if (name === "schema-correctness") cols.push({ label: "Schema", width: 8, align: "right" });
+      else if (name === "fuzzy-similarity") cols.push({ label: "Fuzzy", width: 8, align: "right" });
+      else if (name === "llm-judge-correctness") cols.push({ label: "Judge", width: 8, align: "right" });
+      else if (name === "tool-usage") cols.push({ label: "Tool", width: 8, align: "right" });
+      else cols.push({ label: name, width: 10, align: "right" });
+    }
+    if (hasErrors) cols.push({ label: "Status", width: 8, align: "left" });
+    const totalWidth = cols.reduce((sum, c) => sum + c.width + 2, 0);
+    console.log(`  ${dim(cols.map((c) => pad(c.label, c.width + 2, c.align)).join(""))}`);
+    console.log(`  ${dim("\u2500".repeat(totalWidth))}`);
+    for (const provider of providers) {
+      const taskResults = results.filter(
+        (r) => r.taskName === task && r.providerId === provider
+      );
+      const errorResults2 = taskResults.filter((r) => r.error);
+      const successResults = taskResults.filter((r) => !r.error);
+      if (successResults.length === 0 && errorResults2.length > 0) {
+        const cells2 = [pad(provider, 24, "left")];
+        for (const name of scorerNames) {
+          if (name === "cost") {
+            cells2.push(pad("\u2014", 14, "right"));
+            cells2.push(pad("\u2014", 11, "right"));
+          } else cells2.push(pad("\u2014", cols.find((c) => c.label !== "Provider").width + 2, "right"));
+        }
+        if (hasErrors) cells2.push(`  ${red}FAIL${reset}`);
+        console.log(`  ${cells2.join("")}`);
+        continue;
+      }
+      const avgScores = averageScores(successResults);
+      const avgDetails = averageDetails(successResults);
+      const latencyMs = average(successResults.map((r) => r.raw.latencyMs));
+      const cells = [pad(provider, 24, "left")];
+      for (const name of scorerNames) {
+        if (name === "latency") {
+          cells.push(pad(latencyMs !== void 0 ? `${Math.round(latencyMs)}ms` : "\u2014", 12, "right"));
+        } else if (name === "cost") {
+          cells.push(pad(formatCost(avgDetails.costUsd), 14, "right"));
+          cells.push(pad(avgDetails.totalTokens !== void 0 ? `${avgDetails.totalTokens}` : "\u2014", 11, "right"));
+        } else {
+          const val = avgScores[name];
+          if (val === void 0) cells.push(pad("\u2014", 10, "right"));
+          else cells.push(pad(colorScore(val), 10 + colorLen(colorScore(val)), "right"));
+        }
+      }
+      if (hasErrors) {
+        const failCount = errorResults2.length;
+        cells.push(failCount > 0 ? `  ${yellow}${failCount} err${reset}` : `  ${green}OK${reset}`);
+      }
+      console.log(`  ${cells.join("")}`);
+    }
+    console.log("");
+  }
+  printSummary(results, providers);
+  const errorResults = results.filter((r) => r.error);
+  if (errorResults.length > 0) {
+    console.log(`  ${bold("Errors")}`);
+    console.log(`  ${dim("\u2500".repeat(70))}`);
+    const seen = /* @__PURE__ */ new Set();
+    for (const r of errorResults) {
+      const key = `${r.providerId}::${r.error}`;
+      if (seen.has(key)) continue;
+      seen.add(key);
+      const count = errorResults.filter((e) => e.providerId === r.providerId && e.error === r.error).length;
+      const suffix = count > 1 ? ` (\xD7${count})` : "";
+      console.log(`  ${red}\u2717${reset} ${r.providerId}: ${r.error}${suffix}`);
+      const hint = apiKeyHint(r.providerId, r.error ?? "");
+      if (hint) console.log(`    ${dim(hint)}`);
+    }
+    console.log("");
+  }
+  if (hasCost) {
+    console.log(dim(`  Costs estimated from OpenRouter pricing catalog. Run npx tsx scripts/update-pricing.ts to refresh.`));
+    console.log("");
+  }
+}
+function printSummary(results, providers) {
+  const successResults = results.filter((r) => !r.error);
+  if (successResults.length === 0) return;
+  console.log(`  ${dim("\u2500".repeat(70))}`);
+  console.log(`  ${bold("Summary")}`);
+  console.log("");
+  const single = providers.length === 1;
+  const correctnessKey = successResults.some((r) => r.scores.some((s) => s.name === "llm-judge-correctness" && s.value >= 0)) ? "llm-judge-correctness" : "correctness";
+  const byCorrectness = rankProviders(successResults, providers, correctnessKey);
+  if (byCorrectness) {
+    const label = single ? "Avg correctness" : `Most correct: ${bold(byCorrectness.id)} ${dim(providerLabel(byCorrectness.id))}`;
+    console.log(`  ${cyan}\u25C6${reset} ${label} (avg ${colorScore(byCorrectness.avg)})`);
+  }
+  const byLatency = providers.map((id) => {
+    const runs = successResults.filter((r) => r.providerId === id);
+    const avg = average(runs.map((r) => r.raw.latencyMs));
+    return { id, avg: avg ?? Infinity };
+  }).sort((a, b) => a.avg - b.avg)[0];
+  if (byLatency && byLatency.avg !== Infinity) {
+    const label = single ? "Avg latency" : `Fastest: ${bold(byLatency.id)} ${dim(providerLabel(byLatency.id))}`;
+    console.log(`  ${cyan}\u25C6${reset} ${label} (avg ${Math.round(byLatency.avg)}ms)`);
+  }
+  const byCost = providers.map((id) => {
+    const runs = successResults.filter((r) => r.providerId === id);
+    const costs = runs.map((r) => {
+      const s = r.scores.find((s2) => s2.name === "cost");
+      return s && s.value >= 0 ? s.value : void 0;
+    }).filter((c) => c !== void 0);
+    const avg = costs.length > 0 ? costs.reduce((a, b) => a + b, 0) / costs.length : void 0;
+    return { id, avg };
+  }).filter((p) => p.avg !== void 0).sort((a, b) => a.avg - b.avg)[0];
+  if (byCost?.avg !== void 0) {
+    const label = single ? "Avg cost" : `Cheapest: ${bold(byCost.id)} ${dim(providerLabel(byCost.id))}`;
+    console.log(`  ${cyan}\u25C6${reset} ${label} (avg ${formatCost(byCost.avg)})`);
+  }
+  console.log("");
+}
+function rankProviders(results, providers, scorerName) {
+  const ranked = providers.map((id) => {
+    const runs = results.filter((r) => r.providerId === id);
+    const scores = runs.flatMap((r) => r.scores.filter((s) => s.name === scorerName && s.value >= 0)).map((s) => s.value);
+    const avg = scores.length > 0 ? scores.reduce((a, b) => a + b, 0) / scores.length : void 0;
+    return { id, avg };
+  }).filter((p) => p.avg !== void 0).sort((a, b) => b.avg - a.avg);
+  return ranked[0] ? { id: ranked[0].id, avg: ranked[0].avg } : void 0;
+}
+function averageScores(results) {
+  const sums = {};
+  const counts = {};
+  for (const result of results) {
+    for (const score of result.scores) {
+      if (score.value < 0) continue;
+      sums[score.name] = (sums[score.name] ?? 0) + score.value;
+      counts[score.name] = (counts[score.name] ?? 0) + 1;
+    }
+  }
+  const avgs = {};
+  for (const name of Object.keys(sums)) {
+    avgs[name] = sums[name] / counts[name];
+  }
+  return avgs;
+}
+function averageDetails(results) {
+  let costSum = 0;
+  let costCount = 0;
+  let tokenSum = 0;
+  let tokenCount = 0;
+  for (const result of results) {
+    const costScore = result.scores.find((s) => s.name === "cost");
+    const details = costScore?.details;
+    if (details?.estimatedUsd != null) {
+      costSum += details.estimatedUsd;
+      costCount++;
+    }
+    if (details?.totalTokens != null) {
+      tokenSum += details.totalTokens;
+      tokenCount++;
+    }
+  }
+  return {
+    costUsd: costCount > 0 ? costSum / costCount : void 0,
+    totalTokens: tokenCount > 0 ? Math.round(tokenSum / tokenCount) : void 0
+  };
+}
+function average(nums) {
+  if (nums.length === 0) return void 0;
+  return nums.reduce((a, b) => a + b, 0) / nums.length;
+}
+function formatCost(usd) {
+  if (usd === void 0) return "\u2014";
+  if (usd === 0) return "$0.00";
+  if (usd >= 0.01) return `~$${usd.toFixed(2)}`;
+  const digits = Math.max(4, -Math.floor(Math.log10(usd)) + 1);
+  return `~$${usd.toFixed(digits).replace(/0+$/, "")}`;
+}
+function pad(str, width, align) {
+  if (align === "right") return str.padStart(width);
+  return str.padEnd(width);
+}
+function colorLen(str) {
+  const stripped = str.replace(/\x1b\[[0-9;]*m/g, "");
+  return str.length - stripped.length;
+}
+function apiKeyHint(providerId, error) {
+  const lower = error.toLowerCase();
+  const isAuthError = lower.includes("api key") || lower.includes("401") || lower.includes("unauthorized") || lower.includes("authentication") || lower.includes("incorrect api key") || lower.includes("apikey");
+  if (!isAuthError) return void 0;
+  const prefix = providerId.split("/")[0];
+  switch (prefix) {
+    case "openai":
+      return "Set: export OPENAI_API_KEY=sk-...";
+    case "azure":
+      return "Set: export AZURE_OPENAI_API_KEY=... and AZURE_OPENAI_ENDPOINT=...";
+    case "anthropic":
+      return "Set: export ANTHROPIC_API_KEY=sk-ant-...";
+    case "google":
+      return "Set: export GOOGLE_API_KEY=...";
+    default:
+      return `Check the API key for ${providerId}`;
+  }
+}
+function providerLabel(providerId) {
+  const prefix = providerId.split("/")[0];
+  switch (prefix) {
+    case "azure":
+      return "(OpenAI via Azure)";
+    case "openai":
+      return "(OpenAI)";
+    case "anthropic":
+      return "(Anthropic)";
+    case "google":
+      return "(Google)";
+    case "mistral":
+      return "(Mistral)";
+    case "meta":
+      return "(Meta)";
+    case "deepseek":
+      return "(DeepSeek)";
+    case "cohere":
+      return "(Cohere)";
+    case "qwen":
+      return "(Qwen)";
+    case "xai":
+      return "(xAI)";
+    case "minimax":
+      return "(MiniMax)";
+    case "moonshot":
+      return "(Moonshot / Kimi)";
+    case "perplexity":
+      return "(Perplexity)";
+    case "amazon":
+      return "(Amazon)";
+    case "nvidia":
+      return "(NVIDIA)";
+    case "microsoft":
+      return "(Microsoft)";
+    case "ai21":
+      return "(AI21 Labs)";
+    case "bytedance":
+      return "(ByteDance)";
+    case "together":
+      return "(Together AI)";
+    case "fireworks":
+      return "(Fireworks AI)";
+    case "groq":
+      return "(Groq)";
+    case "cerebras":
+      return "(Cerebras)";
+    default:
+      return `(${prefix})`;
+  }
+}
+
+// src/reporter/json.ts
+function jsonReporter(results) {
+  return JSON.stringify(
+    {
+      timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+      summary: buildSummary(results),
+      results
+    },
+    null,
+    2
+  );
+}
+function buildSummary(results) {
+  const tasks = [...new Set(results.map((r) => r.taskName))];
+  const providers = [...new Set(results.map((r) => r.providerId))];
+  return {
+    totalBenchmarks: results.length,
+    tasks: tasks.length,
+    providers: providers.length,
+    providerIds: providers,
+    taskNames: tasks
+  };
+}
+
+// src/arena.ts
+function defineArena(config) {
+  if (config.providers.length === 0) {
+    throw new Error("At least one provider is required");
+  }
+  if (config.tasks.length === 0) {
+    throw new Error("At least one task is required");
+  }
+  const scorerNames = config.scorers ?? ["latency", "cost", "correctness"];
+  const scorerFns = resolveScorers(scorerNames, config.judgeModel);
+  const runs = config.runs ?? 1;
+  return {
+    config,
+    async run(options) {
+      return runBenchmarks({
+        providers: config.providers,
+        tasks: config.tasks,
+        scorers: scorerFns,
+        runs,
+        onResult: options?.onResult
+      });
+    }
+  };
+}
+
+// src/providers/openai.ts
+var import_openai2 = __toESM(require("openai"), 1);
+var import_zod_to_json_schema = require("zod-to-json-schema");
+function openai(model, options) {
+  const client = new import_openai2.default({
+    apiKey: options?.apiKey ?? process.env.OPENAI_API_KEY,
+    baseURL: options?.baseURL
+  });
+  return makeProvider(`openai/${model}`, "OpenAI", model, client, model);
+}
+function openaiCompatible(options) {
+  const apiKey = options.apiKey ?? (options.apiKeyEnv ? process.env[options.apiKeyEnv] : void 0) ?? "no-key";
+  const client = new import_openai2.default({
+    apiKey,
+    baseURL: options.baseURL
+  });
+  if (options.free) {
+    registerPricing(options.id, { inputPerToken: 0, outputPerToken: 0 });
+  }
+  return makeProvider(options.id, options.name, options.model, client, options.model, options.stripThinking);
+}
+function azureOpenai(model, options) {
+  const deployment = options?.deployment ?? model;
+  const client = new import_openai2.AzureOpenAI({
+    apiKey: options?.apiKey ?? process.env.AZURE_OPENAI_API_KEY,
+    endpoint: options?.endpoint ?? process.env.AZURE_OPENAI_ENDPOINT,
+    apiVersion: options?.apiVersion ?? process.env.AZURE_OPENAI_API_VERSION ?? "2024-12-01-preview",
+    deployment
+  });
+  return makeProvider(`azure/${model}`, "Azure OpenAI", model, client, deployment);
+}
+function makeProvider(id, name, model, client, requestModel, stripThinking) {
+  return {
+    id,
+    name,
+    model,
+    async run(input) {
+      const start = Date.now();
+      const params = {
+        model: requestModel,
+        messages: [{ role: "user", content: input.prompt }]
+      };
+      if (input.schema) {
+        params.response_format = { type: "json_object" };
+        params.messages = [
+          { role: "system", content: "Respond with valid JSON matching the requested schema." },
+          ...params.messages
+        ];
+      }
+      if (input.tools?.length) {
+        params.tools = input.tools.map(toolDefToOpenAI);
+        params.tool_choice = "auto";
+      }
+      const response = await client.chat.completions.create(params);
+      let totalPromptTokens = response.usage?.prompt_tokens ?? 0;
+      let totalCompletionTokens = response.usage?.completion_tokens ?? 0;
+      const choice = response.choices[0];
+      const toolCallsRaw = choice?.message?.tool_calls;
+      const collectedToolCalls = [];
+      let finalResponse = response;
+      if (toolCallsRaw?.length && input.tools?.length) {
+        const toolMessages = [
+          ...params.messages,
+          choice.message
+        ];
+        for (const tc of toolCallsRaw) {
+          const toolDef = input.tools.find((t) => t.name === tc.function.name);
+          let args;
+          try {
+            args = JSON.parse(tc.function.arguments);
+          } catch {
+            args = tc.function.arguments;
+          }
+          let result;
+          if (toolDef?.handler) {
+            result = await toolDef.handler(args);
+          }
+          collectedToolCalls.push({ name: tc.function.name, arguments: args, result });
+          toolMessages.push({
+            role: "tool",
+            tool_call_id: tc.id,
+            content: JSON.stringify(result ?? {})
+          });
+        }
+        const followUp = await client.chat.completions.create({
+          model: requestModel,
+          messages: toolMessages
+        });
+        totalPromptTokens += followUp.usage?.prompt_tokens ?? 0;
+        totalCompletionTokens += followUp.usage?.completion_tokens ?? 0;
+        finalResponse = followUp;
+      }
+      const latencyMs = Date.now() - start;
+      const finalChoice = finalResponse.choices[0];
+      let rawContent = finalChoice?.message?.content ?? "";
+      if (stripThinking) {
+        rawContent = rawContent.replace(/<think>[\s\S]*?<\/think>\s*/, "");
+      }
+      let output = rawContent;
+      if (input.schema) {
+        try {
+          output = JSON.parse(rawContent);
+        } catch {
+        }
+      }
+      return {
+        output,
+        usage: {
+          promptTokens: totalPromptTokens || void 0,
+          completionTokens: totalCompletionTokens || void 0
+        },
+        latencyMs,
+        raw: finalResponse,
+        toolCalls: collectedToolCalls.length > 0 ? collectedToolCalls : void 0
+      };
+    }
+  };
+}
+function toolDefToOpenAI(tool) {
+  return {
+    type: "function",
+    function: {
+      name: tool.name,
+      description: tool.description,
+      parameters: (0, import_zod_to_json_schema.zodToJsonSchema)(tool.parameters, { target: "openAi" })
+    }
+  };
+}
+
+// src/providers/anthropic.ts
+var import_sdk = __toESM(require("@anthropic-ai/sdk"), 1);
+function anthropic(model, options) {
+  const client = new import_sdk.default({
+    apiKey: options?.apiKey ?? process.env.ANTHROPIC_API_KEY
+  });
+  const maxTokens = options?.maxTokens ?? 1024;
+  return {
+    id: `anthropic/${model}`,
+    name: "Anthropic",
+    model,
+    async run(input) {
+      const start = Date.now();
+      const systemMessage = input.schema ? "Respond with valid JSON matching the requested schema." : void 0;
+      const response = await client.messages.create({
+        model,
+        max_tokens: maxTokens,
+        system: systemMessage,
+        messages: [{ role: "user", content: input.prompt }]
+      });
+      const latencyMs = Date.now() - start;
+      const textBlock = response.content.find((b) => b.type === "text");
+      const rawContent = textBlock?.type === "text" ? textBlock.text : "";
+      let output = rawContent;
+      if (input.schema) {
+        try {
+          output = JSON.parse(rawContent);
+        } catch {
+        }
+      }
+      return {
+        output,
+        usage: {
+          promptTokens: response.usage.input_tokens,
+          completionTokens: response.usage.output_tokens
+        },
+        latencyMs,
+        raw: response
+      };
+    }
+  };
+}
+
+// src/providers/gemini.ts
+var import_openai3 = __toESM(require("openai"), 1);
+function gemini(model, options) {
+  const apiKey = options?.apiKey ?? process.env.GOOGLE_API_KEY;
+  if (!apiKey) {
+    throw new Error(
+      `Missing API key for google/${model}. Set GOOGLE_API_KEY or pass apiKey option.`
+    );
+  }
+  const client = new import_openai3.default({
+    apiKey,
+    baseURL: "https://generativelanguage.googleapis.com/v1beta/openai/"
+  });
+  return makeProvider(`google/${model}`, "Google AI", model, client, model);
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  anthropic,
+  azureOpenai,
+  consoleReporter,
+  defineArena,
+  gemini,
+  jsonReporter,
+  openai,
+  openaiCompatible,
+  registerPricing
+});
+//# sourceMappingURL=index.cjs.map