@artemiskit/core 0.2.0 → 0.2.3

package/dist/index.js

@@ -11564,8 +11564,7 @@ class LLMGraderEvaluator {
       const result = await context.client.generate({
         prompt,
         model: expected.model,
-        temperature: 0,
-        maxTokens: 200
+        maxTokens: 1000
       });
       const parsed = this.parseGraderResponse(result.text);
       const passed = parsed.score >= expected.threshold;
@@ -11590,9 +11589,17 @@ class LLMGraderEvaluator {
     }
   }
   parseGraderResponse(text) {
-    const jsonMatch = text.match(/\{[\s\S]*?\}/);
+    const cleanedText = text.replace(/```json\s*/gi, "").replace(/```\s*/g, "").trim();
+    const jsonMatch = cleanedText.match(/\{[\s\S]*?\}/);
     if (!jsonMatch) {
-      throw new Error("No JSON found in grader response");
+      const scoreMatch = cleanedText.match(/(?:score[:\s]*)?(\d+\.?\d*)/i);
+      if (scoreMatch) {
+        const score = Number(scoreMatch[1]);
+        if (!Number.isNaN(score) && score >= 0 && score <= 1) {
+          return { score, reason: cleanedText };
+        }
+      }
+      throw new Error(`No JSON found in grader response: ${text.substring(0, 100)}...`);
     }
     try {
       const parsed = JSON.parse(jsonMatch[0]);
@@ -11605,26 +11612,39 @@ class LLMGraderEvaluator {
         reason: parsed.reason
       };
     } catch (error) {
+      const scoreMatch = jsonMatch[0].match(/"score"[:\s]*(\d+\.?\d*)/i);
+      if (scoreMatch) {
+        const score = Number(scoreMatch[1]);
+        if (!Number.isNaN(score) && score >= 0 && score <= 1) {
+          const reasonMatch = jsonMatch[0].match(/"reason"[:\s]*"([^"]+)"/i);
+          return { score, reason: reasonMatch?.[1] };
+        }
+      }
       throw new Error(`Failed to parse grader response: ${error.message}`);
     }
   }
 }
-var GRADER_PROMPT = `You are an evaluator grading an AI response based on a rubric.
+var GRADER_PROMPT = `You are a strict JSON-only evaluator. You grade AI responses based on rubrics.
 
-## RUBRIC
+RUBRIC:
 {{rubric}}
 
-## RESPONSE TO EVALUATE
+RESPONSE TO EVALUATE:
 {{response}}
 
-## INSTRUCTIONS
-Score the response from 0.0 to 1.0 based on the rubric.
-Be objective and consistent in your scoring.
+TASK: Score the response from 0.0 to 1.0 based on the rubric above.
 
-Respond with ONLY a JSON object in this exact format:
-{"score": <number between 0 and 1>, "reason": "<brief explanation of score>"}
+OUTPUT FORMAT: You MUST respond with ONLY this exact JSON structure, nothing else:
+{"score":0.0,"reason":"explanation"}
+
+RULES:
+- Output ONLY valid JSON, no markdown, no code blocks, no extra text
+- "score" must be a number between 0.0 and 1.0
+- "reason" must be a brief string explaining the score
+- Do NOT wrap in \`\`\`json or any formatting
+- Your entire response must be parseable by JSON.parse()
 
-Do not include any other text, markdown, or formatting.`;
+JSON OUTPUT:`;
 
 // src/evaluators/not-contains.ts
 class NotContainsEvaluator {
@@ -13487,6 +13507,7 @@ var ProviderConfigSchema = exports_external.object({
   deploymentName: exports_external.string().optional(),
   apiVersion: exports_external.string().optional(),
   embeddingDeploymentName: exports_external.string().optional(),
+  modelFamily: exports_external.string().optional(),
   underlyingProvider: exports_external.enum(["openai", "azure", "anthropic", "google", "mistral"]).optional()
 }).optional();
 var BaseExpectedSchema = exports_external.discriminatedUnion("type", [
@@ -14217,6 +14238,319 @@ function nanoid(size = 21) {
   return id;
 }
 
+// src/cost/pricing.ts
+var MODEL_PRICING = {
+  "gpt-5": {
+    promptPer1K: 0.00125,
+    completionPer1K: 0.01,
+    lastUpdated: "2026-01",
+    notes: "400K context window"
+  },
+  "gpt-5.1": {
+    promptPer1K: 0.00125,
+    completionPer1K: 0.01,
+    lastUpdated: "2026-01"
+  },
+  "gpt-5.2": {
+    promptPer1K: 0.00175,
+    completionPer1K: 0.014,
+    lastUpdated: "2026-01"
+  },
+  "gpt-5-mini": {
+    promptPer1K: 0.00025,
+    completionPer1K: 0.002,
+    lastUpdated: "2026-01"
+  },
+  "gpt-5-nano": {
+    promptPer1K: 0.00005,
+    completionPer1K: 0.0004,
+    lastUpdated: "2026-01"
+  },
+  "gpt-4.1": {
+    promptPer1K: 0.002,
+    completionPer1K: 0.008,
+    lastUpdated: "2026-01",
+    notes: "1M context window"
+  },
+  "gpt-4.1-mini": {
+    promptPer1K: 0.0004,
+    completionPer1K: 0.0016,
+    lastUpdated: "2026-01"
+  },
+  "gpt-4.1-nano": {
+    promptPer1K: 0.0001,
+    completionPer1K: 0.0004,
+    lastUpdated: "2026-01"
+  },
+  "gpt-4o": {
+    promptPer1K: 0.0025,
+    completionPer1K: 0.01,
+    lastUpdated: "2026-01",
+    notes: "128K context window"
+  },
+  "gpt-4o-mini": {
+    promptPer1K: 0.00015,
+    completionPer1K: 0.0006,
+    lastUpdated: "2026-01",
+    notes: "128K context window"
+  },
+  o1: {
+    promptPer1K: 0.015,
+    completionPer1K: 0.06,
+    lastUpdated: "2026-01",
+    notes: "Reasoning model - internal thinking tokens billed as output"
+  },
+  o3: {
+    promptPer1K: 0.002,
+    completionPer1K: 0.008,
+    lastUpdated: "2026-01"
+  },
+  "o3-mini": {
+    promptPer1K: 0.0011,
+    completionPer1K: 0.0044,
+    lastUpdated: "2026-01"
+  },
+  "o4-mini": {
+    promptPer1K: 0.0011,
+    completionPer1K: 0.0044,
+    lastUpdated: "2026-01"
+  },
+  "gpt-4-turbo": {
+    promptPer1K: 0.01,
+    completionPer1K: 0.03,
+    lastUpdated: "2026-01"
+  },
+  "gpt-4": {
+    promptPer1K: 0.03,
+    completionPer1K: 0.06,
+    lastUpdated: "2026-01"
+  },
+  "gpt-3.5-turbo": {
+    promptPer1K: 0.0005,
+    completionPer1K: 0.0015,
+    lastUpdated: "2026-01"
+  },
+  "claude-opus-4.5": {
+    promptPer1K: 0.005,
+    completionPer1K: 0.025,
+    lastUpdated: "2026-01",
+    notes: "Most capable Claude model"
+  },
+  "claude-sonnet-4.5": {
+    promptPer1K: 0.003,
+    completionPer1K: 0.015,
+    lastUpdated: "2026-01",
+    notes: "Balanced performance and cost"
+  },
+  "claude-haiku-4.5": {
+    promptPer1K: 0.001,
+    completionPer1K: 0.005,
+    lastUpdated: "2026-01",
+    notes: "Fastest Claude model"
+  },
+  "claude-opus-4": {
+    promptPer1K: 0.015,
+    completionPer1K: 0.075,
+    lastUpdated: "2026-01"
+  },
+  "claude-opus-4.1": {
+    promptPer1K: 0.015,
+    completionPer1K: 0.075,
+    lastUpdated: "2026-01"
+  },
+  "claude-sonnet-4": {
+    promptPer1K: 0.003,
+    completionPer1K: 0.015,
+    lastUpdated: "2026-01"
+  },
+  "claude-sonnet-3.7": {
+    promptPer1K: 0.003,
+    completionPer1K: 0.015,
+    lastUpdated: "2026-01"
+  },
+  "claude-3-7-sonnet": {
+    promptPer1K: 0.003,
+    completionPer1K: 0.015,
+    lastUpdated: "2026-01"
+  },
+  "claude-3-5-sonnet-20241022": {
+    promptPer1K: 0.003,
+    completionPer1K: 0.015,
+    lastUpdated: "2026-01"
+  },
+  "claude-3-5-haiku-20241022": {
+    promptPer1K: 0.0008,
+    completionPer1K: 0.004,
+    lastUpdated: "2026-01"
+  },
+  "claude-haiku-3.5": {
+    promptPer1K: 0.0008,
+    completionPer1K: 0.004,
+    lastUpdated: "2026-01"
+  },
+  "claude-3-opus": {
+    promptPer1K: 0.015,
+    completionPer1K: 0.075,
+    lastUpdated: "2026-01"
+  },
+  "claude-3-sonnet": {
+    promptPer1K: 0.003,
+    completionPer1K: 0.015,
+    lastUpdated: "2026-01"
+  },
+  "claude-3-haiku": {
+    promptPer1K: 0.00025,
+    completionPer1K: 0.00125,
+    lastUpdated: "2026-01"
+  },
+  "claude-3.5-sonnet": {
+    promptPer1K: 0.003,
+    completionPer1K: 0.015,
+    lastUpdated: "2026-01"
+  },
+  "claude-3.5-haiku": {
+    promptPer1K: 0.0008,
+    completionPer1K: 0.004,
+    lastUpdated: "2026-01"
+  }
+};
+var DEFAULT_PRICING = {
+  promptPer1K: 0.003,
+  completionPer1K: 0.015,
+  lastUpdated: "2026-01",
+  notes: "Default pricing - verify with provider"
+};
+function getModelPricing(model) {
+  if (MODEL_PRICING[model]) {
+    return MODEL_PRICING[model];
+  }
+  const lowerModel = model.toLowerCase();
+  for (const [key, pricing] of Object.entries(MODEL_PRICING)) {
+    if (key.toLowerCase() === lowerModel) {
+      return pricing;
+    }
+  }
+  if (lowerModel.includes("gpt-5.2")) {
+    return MODEL_PRICING["gpt-5.2"];
+  }
+  if (lowerModel.includes("gpt-5.1")) {
+    return MODEL_PRICING["gpt-5.1"];
+  }
+  if (lowerModel.includes("gpt-5-mini")) {
+    return MODEL_PRICING["gpt-5-mini"];
+  }
+  if (lowerModel.includes("gpt-5-nano")) {
+    return MODEL_PRICING["gpt-5-nano"];
+  }
+  if (lowerModel.includes("gpt-5")) {
+    return MODEL_PRICING["gpt-5"];
+  }
+  if (lowerModel.includes("gpt-4.1-mini")) {
+    return MODEL_PRICING["gpt-4.1-mini"];
+  }
+  if (lowerModel.includes("gpt-4.1-nano")) {
+    return MODEL_PRICING["gpt-4.1-nano"];
+  }
+  if (lowerModel.includes("gpt-4.1")) {
+    return MODEL_PRICING["gpt-4.1"];
+  }
+  if (lowerModel.includes("gpt-4o-mini")) {
+    return MODEL_PRICING["gpt-4o-mini"];
+  }
+  if (lowerModel.includes("gpt-4o")) {
+    return MODEL_PRICING["gpt-4o"];
+  }
+  if (lowerModel.includes("o4-mini")) {
+    return MODEL_PRICING["o4-mini"];
+  }
+  if (lowerModel.includes("o3-mini")) {
+    return MODEL_PRICING["o3-mini"];
+  }
+  if (lowerModel.includes("o3")) {
+    return MODEL_PRICING.o3;
+  }
+  if (lowerModel.includes("o1")) {
+    return MODEL_PRICING.o1;
+  }
+  if (lowerModel.includes("gpt-4-turbo")) {
+    return MODEL_PRICING["gpt-4-turbo"];
+  }
+  if (lowerModel.includes("gpt-4")) {
+    return MODEL_PRICING["gpt-4"];
+  }
+  if (lowerModel.includes("gpt-3.5")) {
+    return MODEL_PRICING["gpt-3.5-turbo"];
+  }
+  if (lowerModel.includes("opus-4.5") || lowerModel.includes("opus-4-5")) {
+    return MODEL_PRICING["claude-opus-4.5"];
+  }
+  if (lowerModel.includes("sonnet-4.5") || lowerModel.includes("sonnet-4-5")) {
+    return MODEL_PRICING["claude-sonnet-4.5"];
+  }
+  if (lowerModel.includes("haiku-4.5") || lowerModel.includes("haiku-4-5")) {
+    return MODEL_PRICING["claude-haiku-4.5"];
+  }
+  if (lowerModel.includes("opus-4.1") || lowerModel.includes("opus-4-1")) {
+    return MODEL_PRICING["claude-opus-4.1"];
+  }
+  if (lowerModel.includes("opus-4")) {
+    return MODEL_PRICING["claude-opus-4"];
+  }
+  if (lowerModel.includes("sonnet-4")) {
+    return MODEL_PRICING["claude-sonnet-4"];
+  }
+  if (lowerModel.includes("sonnet-3.7") || lowerModel.includes("sonnet-3-7")) {
+    return MODEL_PRICING["claude-sonnet-3.7"];
+  }
+  if (lowerModel.includes("claude-3-5-sonnet") || lowerModel.includes("claude-3.5-sonnet")) {
+    return MODEL_PRICING["claude-3.5-sonnet"];
+  }
+  if (lowerModel.includes("claude-3-5-haiku") || lowerModel.includes("claude-3.5-haiku")) {
+    return MODEL_PRICING["claude-3.5-haiku"];
+  }
+  if (lowerModel.includes("claude-3-opus")) {
+    return MODEL_PRICING["claude-3-opus"];
+  }
+  if (lowerModel.includes("claude-3-sonnet")) {
+    return MODEL_PRICING["claude-3-sonnet"];
+  }
+  if (lowerModel.includes("claude-3-haiku")) {
+    return MODEL_PRICING["claude-3-haiku"];
+  }
+  if (lowerModel.includes("claude")) {
+    return MODEL_PRICING["claude-sonnet-4.5"];
+  }
+  return DEFAULT_PRICING;
+}
+function estimateCost(promptTokens, completionTokens, model) {
+  const pricing = getModelPricing(model);
+  const promptCostUsd = promptTokens / 1000 * pricing.promptPer1K;
+  const completionCostUsd = completionTokens / 1000 * pricing.completionPer1K;
+  const totalUsd = promptCostUsd + completionCostUsd;
+  return {
+    totalUsd,
+    promptCostUsd,
+    completionCostUsd,
+    model,
+    pricing
+  };
+}
+function formatCost(costUsd) {
+  if (costUsd < 0.01) {
+    return `$${(costUsd * 100).toFixed(4)} cents`;
+  }
+  if (costUsd < 1) {
+    return `$${costUsd.toFixed(4)}`;
+  }
+  return `$${costUsd.toFixed(2)}`;
+}
+function listKnownModels() {
+  return Object.entries(MODEL_PRICING).map(([model, pricing]) => ({
+    model,
+    pricing
+  }));
+}
+
 // src/provenance/environment.ts
 function getEnvironmentInfo() {
   return {
@@ -14275,7 +14609,8 @@ function createRunManifest(options) {
     runReason,
     redaction
   } = options;
-  const metrics = calculateMetrics(cases);
+  const modelForCost = resolvedConfig?.model || config.model;
+  const metrics = calculateMetrics(cases, modelForCost);
   const git = getGitInfo();
   const environment = getEnvironmentInfo();
   return {
@@ -14299,7 +14634,7 @@ function createRunManifest(options) {
     redaction
   };
 }
-function calculateMetrics(cases) {
+function calculateMetrics(cases, model) {
   const passedCases = cases.filter((c) => c.ok);
   const latencies = cases.map((c) => c.latencyMs).sort((a, b) => a - b);
   const medianLatency = latencies.length > 0 ? latencies[Math.floor(latencies.length / 2)] : 0;
@@ -14307,6 +14642,21 @@ function calculateMetrics(cases) {
   const p95Latency = latencies.length > 0 ? latencies[p95Index] : 0;
   const totalPromptTokens = cases.reduce((sum, c) => sum + c.tokens.prompt, 0);
   const totalCompletionTokens = cases.reduce((sum, c) => sum + c.tokens.completion, 0);
+  let cost;
+  if (model && (totalPromptTokens > 0 || totalCompletionTokens > 0)) {
+    const costEstimate = estimateCost(totalPromptTokens, totalCompletionTokens, model);
+    const pricing = getModelPricing(model);
+    cost = {
+      total_usd: costEstimate.totalUsd,
+      prompt_cost_usd: costEstimate.promptCostUsd,
+      completion_cost_usd: costEstimate.completionCostUsd,
+      model: costEstimate.model,
+      pricing: {
+        prompt_per_1k: pricing.promptPer1K,
+        completion_per_1k: pricing.completionPer1K
+      }
+    };
+  }
   return {
     success_rate: cases.length > 0 ? passedCases.length / cases.length : 0,
     total_cases: cases.length,
@@ -14316,7 +14666,8 @@ function calculateMetrics(cases) {
     p95_latency_ms: p95Latency,
     total_tokens: totalPromptTokens + totalCompletionTokens,
     total_prompt_tokens: totalPromptTokens,
-    total_completion_tokens: totalCompletionTokens
+    total_completion_tokens: totalCompletionTokens,
+    cost
   };
 }
 function detectCIEnvironment() {
@@ -14487,14 +14838,26 @@ function getSuccessRate(manifest) {
   }
   return manifest.metrics.success_rate;
 }
+function getEstimatedCost(manifest) {
+  const type = getManifestType(manifest);
+  if (type === "stress") {
+    return manifest.metrics.cost?.estimated_total_usd;
+  }
+  if (type === "run") {
+    return manifest.metrics.cost?.total_usd;
+  }
+  return;
+}
 function getScenario(manifest) {
   return manifest.config.scenario;
 }
 
 class LocalStorageAdapter {
   basePath;
+  baselinesPath;
   constructor(basePath = "./artemis-runs") {
     this.basePath = resolve2(basePath);
+    this.baselinesPath = join2(this.basePath, ".artemis", "baselines.json");
   }
   async save(manifest) {
     const dir = join2(this.basePath, manifest.project);
@@ -14554,13 +14917,17 @@ class LocalStorageAdapter {
           if (options?.scenario && getScenario(manifest) !== options.scenario) {
             continue;
           }
-          results.push({
+          const item = {
             runId: manifest.run_id,
             scenario: getScenario(manifest),
             successRate: getSuccessRate(manifest),
             createdAt: manifest.start_time,
             type: manifestType
-          });
+          };
+          if (options?.includeCost) {
+            item.estimatedCostUsd = getEstimatedCost(manifest);
+          }
+          results.push(item);
         } catch {}
       }
     }
@@ -14613,6 +14980,89 @@ class LocalStorageAdapter {
       return [];
     }
   }
+  async loadBaselinesFile() {
+    try {
+      const content = await readFile2(this.baselinesPath, "utf-8");
+      return JSON.parse(content);
+    } catch {
+      return { version: "1.0", baselines: {} };
+    }
+  }
+  async saveBaselinesFile(data) {
+    const dir = join2(this.basePath, ".artemis");
+    await mkdir(dir, { recursive: true });
+    await writeFile(this.baselinesPath, JSON.stringify(data, null, 2));
+  }
+  async setBaseline(scenario, runId, tag) {
+    const manifest = await this.loadRun(runId);
+    const scenarioName = scenario || getScenario(manifest);
+    const baseline = {
+      scenario: scenarioName,
+      runId,
+      createdAt: new Date().toISOString(),
+      metrics: {
+        successRate: manifest.metrics.success_rate,
+        medianLatencyMs: manifest.metrics.median_latency_ms,
+        totalTokens: manifest.metrics.total_tokens,
+        passedCases: manifest.metrics.passed_cases,
+        failedCases: manifest.metrics.failed_cases,
+        totalCases: manifest.metrics.total_cases
+      },
+      tag
+    };
+    const data = await this.loadBaselinesFile();
+    data.baselines[scenarioName] = baseline;
+    await this.saveBaselinesFile(data);
+    return baseline;
+  }
+  async getBaseline(scenario) {
+    const data = await this.loadBaselinesFile();
+    return data.baselines[scenario] || null;
+  }
+  async getBaselineByRunId(runId) {
+    const data = await this.loadBaselinesFile();
+    const baselines = Object.values(data.baselines);
+    return baselines.find((b) => b.runId === runId) || null;
+  }
+  async listBaselines() {
+    const data = await this.loadBaselinesFile();
+    return Object.values(data.baselines).sort((a, b) => new Date(b.createdAt).getTime() - new Date(a.createdAt).getTime());
+  }
+  async removeBaseline(scenario) {
+    const data = await this.loadBaselinesFile();
+    if (data.baselines[scenario]) {
+      delete data.baselines[scenario];
+      await this.saveBaselinesFile(data);
+      return true;
+    }
+    return false;
+  }
+  async removeBaselineByRunId(runId) {
+    const data = await this.loadBaselinesFile();
+    const entry = Object.entries(data.baselines).find(([_, b]) => b.runId === runId);
+    if (entry) {
+      delete data.baselines[entry[0]];
+      await this.saveBaselinesFile(data);
+      return true;
+    }
+    return false;
+  }
+  async compareToBaseline(runId, regressionThreshold = 0.05) {
+    const currentManifest = await this.loadRun(runId);
+    const scenario = getScenario(currentManifest);
+    const baseline = await this.getBaseline(scenario);
+    if (!baseline) {
+      return null;
+    }
+    const comparison = await this.compare(baseline.runId, runId);
+    const hasRegression = comparison.delta.successRate < -regressionThreshold;
+    return {
+      baseline,
+      comparison,
+      hasRegression,
+      regressionThreshold
+    };
+  }
 }
 
 // ../../node_modules/.bun/tslib@2.8.1/node_modules/tslib/modules/index.js
@@ -24301,193 +24751,6 @@ class Logger {
   }
 }
 var logger = new Logger("artemis");
-// src/cost/pricing.ts
-var MODEL_PRICING = {
-  "gpt-4": {
-    promptPer1K: 0.03,
-    completionPer1K: 0.06,
-    lastUpdated: "2024-01"
-  },
-  "gpt-4-32k": {
-    promptPer1K: 0.06,
-    completionPer1K: 0.12,
-    lastUpdated: "2024-01"
-  },
-  "gpt-4-turbo": {
-    promptPer1K: 0.01,
-    completionPer1K: 0.03,
-    lastUpdated: "2024-01"
-  },
-  "gpt-4-turbo-preview": {
-    promptPer1K: 0.01,
-    completionPer1K: 0.03,
-    lastUpdated: "2024-01"
-  },
-  "gpt-4o": {
-    promptPer1K: 0.005,
-    completionPer1K: 0.015,
-    lastUpdated: "2024-05"
-  },
-  "gpt-4o-mini": {
-    promptPer1K: 0.00015,
-    completionPer1K: 0.0006,
-    lastUpdated: "2024-07"
-  },
-  "gpt-3.5-turbo": {
-    promptPer1K: 0.0005,
-    completionPer1K: 0.0015,
-    lastUpdated: "2024-01"
-  },
-  "gpt-3.5-turbo-16k": {
-    promptPer1K: 0.003,
-    completionPer1K: 0.004,
-    lastUpdated: "2024-01"
-  },
-  "claude-3-opus-20240229": {
-    promptPer1K: 0.015,
-    completionPer1K: 0.075,
-    lastUpdated: "2024-03"
-  },
-  "claude-3-sonnet-20240229": {
-    promptPer1K: 0.003,
-    completionPer1K: 0.015,
-    lastUpdated: "2024-03"
-  },
-  "claude-3-haiku-20240307": {
-    promptPer1K: 0.00025,
-    completionPer1K: 0.00125,
-    lastUpdated: "2024-03"
-  },
-  "claude-3-5-sonnet-20240620": {
-    promptPer1K: 0.003,
-    completionPer1K: 0.015,
-    lastUpdated: "2024-06"
-  },
-  "claude-3-5-sonnet-20241022": {
-    promptPer1K: 0.003,
-    completionPer1K: 0.015,
-    lastUpdated: "2024-10"
-  },
-  "claude-3-5-haiku-20241022": {
-    promptPer1K: 0.0008,
-    completionPer1K: 0.004,
-    lastUpdated: "2024-10"
-  },
-  "claude-3-opus": {
-    promptPer1K: 0.015,
-    completionPer1K: 0.075,
-    lastUpdated: "2024-03"
-  },
-  "claude-3-sonnet": {
-    promptPer1K: 0.003,
-    completionPer1K: 0.015,
-    lastUpdated: "2024-03"
-  },
-  "claude-3-haiku": {
-    promptPer1K: 0.00025,
-    completionPer1K: 0.00125,
-    lastUpdated: "2024-03"
-  },
-  "claude-3.5-sonnet": {
-    promptPer1K: 0.003,
-    completionPer1K: 0.015,
-    lastUpdated: "2024-10"
-  },
-  "claude-3.5-haiku": {
-    promptPer1K: 0.0008,
-    completionPer1K: 0.004,
-    lastUpdated: "2024-10"
-  },
-  "claude-2": {
-    promptPer1K: 0.008,
-    completionPer1K: 0.024,
-    lastUpdated: "2024-01"
-  },
-  "claude-instant-1": {
-    promptPer1K: 0.0008,
-    completionPer1K: 0.0024,
-    lastUpdated: "2024-01"
-  }
-};
-var DEFAULT_PRICING = {
-  promptPer1K: 0.01,
-  completionPer1K: 0.03,
-  lastUpdated: "2024-01",
-  notes: "Default pricing - verify with provider"
-};
-function getModelPricing(model) {
-  if (MODEL_PRICING[model]) {
-    return MODEL_PRICING[model];
-  }
-  const lowerModel = model.toLowerCase();
-  for (const [key, pricing] of Object.entries(MODEL_PRICING)) {
-    if (key.toLowerCase() === lowerModel) {
-      return pricing;
-    }
-  }
-  if (lowerModel.includes("gpt-4o-mini")) {
-    return MODEL_PRICING["gpt-4o-mini"];
-  }
-  if (lowerModel.includes("gpt-4o")) {
-    return MODEL_PRICING["gpt-4o"];
-  }
-  if (lowerModel.includes("gpt-4-turbo")) {
-    return MODEL_PRICING["gpt-4-turbo"];
-  }
-  if (lowerModel.includes("gpt-4")) {
-    return MODEL_PRICING["gpt-4"];
-  }
-  if (lowerModel.includes("gpt-3.5")) {
-    return MODEL_PRICING["gpt-3.5-turbo"];
-  }
-  if (lowerModel.includes("claude-3-5-sonnet") || lowerModel.includes("claude-3.5-sonnet")) {
-    return MODEL_PRICING["claude-3.5-sonnet"];
-  }
-  if (lowerModel.includes("claude-3-5-haiku") || lowerModel.includes("claude-3.5-haiku")) {
-    return MODEL_PRICING["claude-3.5-haiku"];
-  }
-  if (lowerModel.includes("claude-3-opus")) {
-    return MODEL_PRICING["claude-3-opus"];
-  }
-  if (lowerModel.includes("claude-3-sonnet")) {
-    return MODEL_PRICING["claude-3-sonnet"];
-  }
-  if (lowerModel.includes("claude-3-haiku")) {
-    return MODEL_PRICING["claude-3-haiku"];
-  }
-  if (lowerModel.includes("claude")) {
-    return MODEL_PRICING["claude-2"];
-  }
-  return DEFAULT_PRICING;
-}
-function estimateCost(promptTokens, completionTokens, model) {
-  const pricing = getModelPricing(model);
-  const promptCostUsd = promptTokens / 1000 * pricing.promptPer1K;
-  const completionCostUsd = completionTokens / 1000 * pricing.completionPer1K;
-  const totalUsd = promptCostUsd + completionCostUsd;
-  return {
-    totalUsd,
-    promptCostUsd,
-    completionCostUsd,
-    model,
-    pricing
-  };
-}
-function formatCost(costUsd) {
-  if (costUsd < 0.01) {
-    return `$${(costUsd * 100).toFixed(4)} cents`;
-  }
-  if (costUsd < 1) {
-    return `$${costUsd.toFixed(4)}`;
-  }
-  return `$${costUsd.toFixed(2)}`;
-}
-function listKnownModels() {
-  return Object.entries(MODEL_PRICING).map(([model, pricing]) => ({
-    model,
-    pricing
-  }));
-}
 export {
   wrapError,
   validateScenario,