vibezcheck 0.1.0

package/dist/ai-sdk/index.mjs

@@ -0,0 +1,944 @@
+// src/meter/client.ts
+import Stripe from "stripe";
+
+// src/meter/batcher.ts
+var MeterBatcher = class {
+  queue = [];
+  timer = null;
+  isFlushing = false;
+  maxBatchSize;
+  flushIntervalMs;
+  stripeClient;
+  eventName;
+  onUsageCallback;
+  onErrorCallback;
+  debug;
+  // In-memory ledger for local stats
+  totalRequests = 0;
+  totalTokens = 0;
+  totalInputTokens = 0;
+  totalOutputTokens = 0;
+  totalReasoningTokens = 0;
+  totalCostUSD = 0;
+  byModel = {};
+  constructor(options = {}) {
+    this.stripeClient = options.stripe;
+    this.eventName = options.eventName || "token-billing-tokens";
+    this.maxBatchSize = options.batching?.maxBatchSize ?? 50;
+    this.flushIntervalMs = options.batching?.flushIntervalMs ?? 50;
+    this.onUsageCallback = options.onUsage;
+    this.onErrorCallback = options.onError;
+    this.debug = options.debug ?? false;
+  }
+  /**
+   * Enqueue a usage event for batch dispatching
+   */
+  enqueue(event) {
+    this.recordInLedger(event);
+    if (this.onUsageCallback) {
+      try {
+        const res = this.onUsageCallback(event);
+        if (res instanceof Promise) {
+          res.catch((err) => {
+            if (this.debug) console.error("[vibezcheck] Error in onUsage callback:", err);
+          });
+        }
+      } catch (err) {
+        if (this.debug) console.error("[vibezcheck] Error in onUsage callback:", err);
+      }
+    }
+    if (!this.stripeClient) {
+      if (this.debug) {
+        console.log(
+          `[vibezcheck:local] \u{1F4CA} ${event.model} | Tokens: ${event.usage.totalTokens} | Cost: $${event.cost.totalUSD.toFixed(6)}`
+        );
+      }
+      return;
+    }
+    this.queue.push(event);
+    if (this.queue.length >= this.maxBatchSize) {
+      this.flush().catch((err) => {
+        if (this.debug) console.error("[vibezcheck] Batch flush error:", err);
+      });
+    } else if (!this.timer) {
+      this.timer = setTimeout(() => {
+        this.timer = null;
+        this.flush().catch((err) => {
+          if (this.debug) console.error("[vibezcheck] Debounce flush error:", err);
+        });
+      }, this.flushIntervalMs);
+    }
+  }
+  /**
+   * Immediately flush all queued events to Stripe
+   */
+  async flush() {
+    if (this.timer) {
+      clearTimeout(this.timer);
+      this.timer = null;
+    }
+    if (this.queue.length === 0 || !this.stripeClient || this.isFlushing) {
+      return;
+    }
+    this.isFlushing = true;
+    const eventsToSend = [...this.queue];
+    this.queue = [];
+    try {
+      await this.sendEventsToStripe(eventsToSend);
+    } catch (error) {
+      const err = error instanceof Error ? error : new Error(String(error));
+      if (this.debug) {
+        console.error("[vibezcheck] Failed to send meter events to Stripe:", err);
+      }
+      if (this.onErrorCallback) {
+        this.onErrorCallback(err, eventsToSend);
+      }
+    } finally {
+      this.isFlushing = false;
+      if (this.queue.length > 0) {
+        this.flush().catch(() => {
+        });
+      }
+    }
+  }
+  /**
+   * Sends events to Stripe Billing Meter Events API
+   */
+  async sendEventsToStripe(events) {
+    if (!this.stripeClient) return;
+    for (const event of events) {
+      const customerId = event.customerId;
+      if (!customerId) {
+        continue;
+      }
+      const timestamp = event.timestamp || (/* @__PURE__ */ new Date()).toISOString();
+      const model = `${event.provider}/${event.model}`;
+      if (event.usage.inputTokens > 0) {
+        try {
+          await this.stripeClient.v2.billing.meterEvents.create({
+            event_name: this.eventName,
+            timestamp,
+            payload: {
+              stripe_customer_id: customerId,
+              value: event.usage.inputTokens.toString(),
+              model,
+              token_type: "input",
+              cached_tokens: (event.usage.cachedTokens ?? 0).toString(),
+              ...event.metadata ? event.metadata : {}
+            }
+          });
+        } catch (e) {
+          if (this.debug) console.warn("[vibezcheck] Input meter event error:", e);
+        }
+      }
+      if (event.usage.outputTokens > 0) {
+        try {
+          await this.stripeClient.v2.billing.meterEvents.create({
+            event_name: this.eventName,
+            timestamp,
+            payload: {
+              stripe_customer_id: customerId,
+              value: event.usage.outputTokens.toString(),
+              model,
+              token_type: "output",
+              reasoning_tokens: (event.usage.reasoningTokens ?? 0).toString(),
+              visible_tokens: (event.usage.visibleOutputTokens ?? event.usage.outputTokens).toString(),
+              ...event.metadata ? event.metadata : {}
+            }
+          });
+        } catch (e) {
+          if (this.debug) console.warn("[vibezcheck] Output meter event error:", e);
+        }
+      }
+    }
+  }
+  /**
+   * Updates internal in-memory ledger
+   */
+  recordInLedger(event) {
+    this.totalRequests += 1;
+    this.totalTokens += event.usage.totalTokens;
+    this.totalInputTokens += event.usage.inputTokens;
+    this.totalOutputTokens += event.usage.outputTokens;
+    this.totalReasoningTokens += event.usage.reasoningTokens ?? 0;
+    this.totalCostUSD += event.cost.totalUSD;
+    const modelKey = event.model;
+    if (!this.byModel[modelKey]) {
+      this.byModel[modelKey] = { requests: 0, tokens: 0, costUSD: 0 };
+    }
+    this.byModel[modelKey].requests += 1;
+    this.byModel[modelKey].tokens += event.usage.totalTokens;
+    this.byModel[modelKey].costUSD += event.cost.totalUSD;
+  }
+  /**
+   * Get in-memory usage summary
+   */
+  getSummary() {
+    return {
+      totalRequests: this.totalRequests,
+      totalTokens: this.totalTokens,
+      totalInputTokens: this.totalInputTokens,
+      totalOutputTokens: this.totalOutputTokens,
+      totalReasoningTokens: this.totalReasoningTokens,
+      totalCostUSD: Number(this.totalCostUSD.toFixed(6)),
+      byModel: { ...this.byModel }
+    };
+  }
+  /**
+   * Reset in-memory ledger
+   */
+  resetLedger() {
+    this.totalRequests = 0;
+    this.totalTokens = 0;
+    this.totalInputTokens = 0;
+    this.totalOutputTokens = 0;
+    this.totalReasoningTokens = 0;
+    this.totalCostUSD = 0;
+    this.byModel = {};
+  }
+};
+
+// src/meter/extractors/openai.ts
+function extractOpenAIResponseUsage(response) {
+  if (!response || typeof response !== "object") return null;
+  if ("choices" in response && "usage" in response && response.usage) {
+    const rawUsage = response.usage;
+    const model = response.model || "gpt-4o";
+    const inputTokens = rawUsage.prompt_tokens ?? 0;
+    const outputTokens = rawUsage.completion_tokens ?? 0;
+    const reasoningTokens = rawUsage.completion_tokens_details?.reasoning_tokens ?? 0;
+    const cachedTokens = rawUsage.prompt_tokens_details?.cached_tokens ?? 0;
+    return {
+      model,
+      provider: "openai",
+      usage: {
+        inputTokens,
+        outputTokens,
+        totalTokens: inputTokens + outputTokens,
+        reasoningTokens: reasoningTokens > 0 ? reasoningTokens : void 0,
+        visibleOutputTokens: Math.max(0, outputTokens - reasoningTokens),
+        cachedTokens: cachedTokens > 0 ? cachedTokens : void 0
+      }
+    };
+  }
+  if ("data" in response && "usage" in response && response.usage && "model" in response) {
+    const rawUsage = response.usage;
+    const inputTokens = rawUsage.prompt_tokens ?? 0;
+    return {
+      model: response.model || "text-embedding-3-small",
+      provider: "openai",
+      usage: {
+        inputTokens,
+        outputTokens: 0,
+        totalTokens: inputTokens
+      }
+    };
+  }
+  if ("status" in response && "usage" in response && response.usage) {
+    const rawUsage = response.usage;
+    const model = response.model || "gpt-5.6-sol";
+    const inputTokens = rawUsage.input_tokens ?? rawUsage.prompt_tokens ?? 0;
+    const outputTokens = rawUsage.output_tokens ?? rawUsage.completion_tokens ?? 0;
+    const reasoningTokens = rawUsage.output_token_details?.reasoning_tokens ?? rawUsage.completion_tokens_details?.reasoning_tokens ?? 0;
+    const cachedTokens = rawUsage.input_token_details?.cached_tokens ?? rawUsage.prompt_tokens_details?.cached_tokens ?? 0;
+    return {
+      model,
+      provider: "openai",
+      usage: {
+        inputTokens,
+        outputTokens,
+        totalTokens: inputTokens + outputTokens,
+        reasoningTokens: reasoningTokens > 0 ? reasoningTokens : void 0,
+        visibleOutputTokens: Math.max(0, outputTokens - reasoningTokens),
+        cachedTokens: cachedTokens > 0 ? cachedTokens : void 0
+      }
+    };
+  }
+  return null;
+}
+function inspectOpenAIStreamChunk(chunk) {
+  if (!chunk || typeof chunk !== "object") return {};
+  const model = chunk.model;
+  if (chunk.usage) {
+    const rawUsage = chunk.usage;
+    const inputTokens = rawUsage.prompt_tokens ?? rawUsage.input_tokens ?? 0;
+    const outputTokens = rawUsage.completion_tokens ?? rawUsage.output_tokens ?? 0;
+    const reasoningTokens = rawUsage.completion_tokens_details?.reasoning_tokens ?? rawUsage.output_token_details?.reasoning_tokens ?? 0;
+    const cachedTokens = rawUsage.prompt_tokens_details?.cached_tokens ?? rawUsage.input_token_details?.cached_tokens ?? 0;
+    return {
+      model,
+      usage: {
+        inputTokens,
+        outputTokens,
+        totalTokens: inputTokens + outputTokens,
+        reasoningTokens: reasoningTokens > 0 ? reasoningTokens : void 0,
+        visibleOutputTokens: Math.max(0, outputTokens - reasoningTokens),
+        cachedTokens: cachedTokens > 0 ? cachedTokens : void 0
+      }
+    };
+  }
+  if (chunk.type === "response.completed" || chunk.type === "response.done") {
+    if (chunk.response?.usage) {
+      const rawUsage = chunk.response.usage;
+      const inputTokens = rawUsage.input_tokens ?? 0;
+      const outputTokens = rawUsage.output_tokens ?? 0;
+      const reasoningTokens = rawUsage.output_token_details?.reasoning_tokens ?? 0;
+      const cachedTokens = rawUsage.input_token_details?.cached_tokens ?? 0;
+      return {
+        model: chunk.response.model || model,
+        usage: {
+          inputTokens,
+          outputTokens,
+          totalTokens: inputTokens + outputTokens,
+          reasoningTokens: reasoningTokens > 0 ? reasoningTokens : void 0,
+          visibleOutputTokens: Math.max(0, outputTokens - reasoningTokens),
+          cachedTokens: cachedTokens > 0 ? cachedTokens : void 0
+        }
+      };
+    }
+  }
+  return { model };
+}
+
+// src/meter/extractors/anthropic.ts
+function extractAnthropicResponseUsage(response) {
+  if (!response || typeof response !== "object") return null;
+  if (response.type === "message" || "content" in response && "usage" in response) {
+    const rawUsage = response.usage || {};
+    const model = response.model || "claude-3-7-sonnet";
+    const inputTokens = rawUsage.input_tokens ?? 0;
+    const outputTokens = rawUsage.output_tokens ?? 0;
+    const cachedTokens = rawUsage.cache_read_input_tokens ?? 0;
+    const cacheWriteTokens = rawUsage.cache_creation_input_tokens ?? 0;
+    let reasoningTokens = void 0;
+    if (Array.isArray(response.content)) {
+      const thinkingBlocks = response.content.filter((b) => b.type === "thinking");
+      if (thinkingBlocks.length > 0) {
+        reasoningTokens = rawUsage.thinking_tokens ?? void 0;
+      }
+    }
+    return {
+      model,
+      provider: "anthropic",
+      usage: {
+        inputTokens,
+        outputTokens,
+        totalTokens: inputTokens + outputTokens,
+        reasoningTokens,
+        visibleOutputTokens: reasoningTokens !== void 0 ? Math.max(0, outputTokens - reasoningTokens) : outputTokens,
+        cachedTokens: cachedTokens > 0 ? cachedTokens : void 0,
+        cacheWriteTokens: cacheWriteTokens > 0 ? cacheWriteTokens : void 0
+      }
+    };
+  }
+  return null;
+}
+var AnthropicStreamAccumulator = class {
+  model = "claude-3-7-sonnet";
+  inputTokens = 0;
+  outputTokens = 0;
+  cachedTokens = 0;
+  cacheWriteTokens = 0;
+  reasoningTokens = 0;
+  processEvent(event) {
+    if (!event || typeof event !== "object") return;
+    if (event.type === "message_start" && event.message) {
+      if (event.message.model) {
+        this.model = event.message.model;
+      }
+      if (event.message.usage) {
+        this.inputTokens = event.message.usage.input_tokens ?? 0;
+        this.cachedTokens = event.message.usage.cache_read_input_tokens ?? 0;
+        this.cacheWriteTokens = event.message.usage.cache_creation_input_tokens ?? 0;
+      }
+    }
+    if (event.type === "message_delta" && event.usage) {
+      this.outputTokens = event.usage.output_tokens ?? 0;
+      if (event.usage.thinking_tokens) {
+        this.reasoningTokens = event.usage.thinking_tokens;
+      }
+    }
+    if (event.type === "content_block_start" && event.content_block?.type === "thinking") {
+    }
+  }
+  getUsage() {
+    return {
+      model: this.model,
+      provider: "anthropic",
+      usage: {
+        inputTokens: this.inputTokens,
+        outputTokens: this.outputTokens,
+        totalTokens: this.inputTokens + this.outputTokens,
+        reasoningTokens: this.reasoningTokens > 0 ? this.reasoningTokens : void 0,
+        visibleOutputTokens: this.reasoningTokens > 0 ? Math.max(0, this.outputTokens - this.reasoningTokens) : this.outputTokens,
+        cachedTokens: this.cachedTokens > 0 ? this.cachedTokens : void 0,
+        cacheWriteTokens: this.cacheWriteTokens > 0 ? this.cacheWriteTokens : void 0
+      }
+    };
+  }
+};
+
+// src/meter/extractors/gemini.ts
+function extractGeminiResponseUsage(response, fallbackModel = "gemini-3.7-flash") {
+  if (!response || typeof response !== "object") return null;
+  const usageMetadata = response.usageMetadata || response.response?.usageMetadata;
+  if (usageMetadata) {
+    const inputTokens = usageMetadata.promptTokenCount ?? 0;
+    const baseOutputTokens = usageMetadata.candidatesTokenCount ?? 0;
+    const thoughtsTokenCount = usageMetadata.thoughtsTokenCount ?? usageMetadata.reasoningTokenCount ?? 0;
+    const cachedTokens = usageMetadata.cachedContentTokenCount ?? 0;
+    const totalOutput = baseOutputTokens + thoughtsTokenCount;
+    const model = response.model || response.response?.model || fallbackModel;
+    return {
+      model,
+      provider: "google",
+      usage: {
+        inputTokens,
+        outputTokens: totalOutput,
+        totalTokens: inputTokens + totalOutput,
+        reasoningTokens: thoughtsTokenCount > 0 ? thoughtsTokenCount : void 0,
+        visibleOutputTokens: baseOutputTokens,
+        cachedTokens: cachedTokens > 0 ? cachedTokens : void 0
+      }
+    };
+  }
+  return null;
+}
+
+// src/meter/extractors/generic.ts
+function extractGenericResponseUsage(response, fallbackModel = "generic-llm", fallbackProvider = "generic") {
+  if (!response || typeof response !== "object") return null;
+  const usage = response.usage || response.token_usage || response.usageMetadata;
+  if (usage) {
+    const inputTokens = usage.prompt_tokens ?? usage.input_tokens ?? usage.promptTokenCount ?? usage.prompt_eval_count ?? 0;
+    const outputTokens = usage.completion_tokens ?? usage.output_tokens ?? usage.candidatesTokenCount ?? usage.eval_count ?? 0;
+    const reasoningTokens = usage.reasoning_tokens ?? usage.thoughtsTokenCount ?? usage.completion_tokens_details?.reasoning_tokens ?? 0;
+    const cachedTokens = usage.prompt_tokens_details?.cached_tokens ?? usage.cached_tokens ?? usage.cachedContentTokenCount ?? 0;
+    const model = response.model || fallbackModel;
+    const provider = response.provider || fallbackProvider;
+    return {
+      model,
+      provider,
+      usage: {
+        inputTokens,
+        outputTokens,
+        totalTokens: inputTokens + outputTokens,
+        reasoningTokens: reasoningTokens > 0 ? reasoningTokens : void 0,
+        visibleOutputTokens: Math.max(0, outputTokens - reasoningTokens),
+        cachedTokens: cachedTokens > 0 ? cachedTokens : void 0
+      }
+    };
+  }
+  return null;
+}
+
+// src/meter/extractors/index.ts
+function detectAndExtractUsage(response, fallbackModel, fallbackProvider) {
+  if (!response || typeof response !== "object") return null;
+  const openaiResult = extractOpenAIResponseUsage(response);
+  if (openaiResult) return openaiResult;
+  const anthropicResult = extractAnthropicResponseUsage(response);
+  if (anthropicResult) return anthropicResult;
+  const geminiResult = extractGeminiResponseUsage(response, fallbackModel);
+  if (geminiResult) return geminiResult;
+  const genericResult = extractGenericResponseUsage(response, fallbackModel, fallbackProvider);
+  if (genericResult) return genericResult;
+  return null;
+}
+
+// src/pricing/table.ts
+var MODEL_PRICING_TABLE = {
+  // --- OpenAI ---
+  "gpt-5.6-sol": { inputPer1M: 4, outputPer1M: 20, cachedInputPer1M: 0.4 },
+  "gpt-5.6-terra": { inputPer1M: 2, outputPer1M: 12, cachedInputPer1M: 0.2 },
+  "gpt-5.6-luna": { inputPer1M: 0.2, outputPer1M: 1.2, cachedInputPer1M: 0.02 },
+  "gpt-5": { inputPer1M: 4, outputPer1M: 20, cachedInputPer1M: 0.4 },
+  "gpt-5-mini": { inputPer1M: 0.2, outputPer1M: 1.2, cachedInputPer1M: 0.02 },
+  "o1": { inputPer1M: 15, outputPer1M: 60, cachedInputPer1M: 7.5 },
+  "o1-mini": { inputPer1M: 1.1, outputPer1M: 4.4, cachedInputPer1M: 0.55 },
+  "o3": { inputPer1M: 15, outputPer1M: 60, cachedInputPer1M: 7.5 },
+  "o3-mini": { inputPer1M: 1.1, outputPer1M: 4.4, cachedInputPer1M: 0.55 },
+  "gpt-4o": { inputPer1M: 2.5, outputPer1M: 10, cachedInputPer1M: 1.25 },
+  "gpt-4o-mini": { inputPer1M: 0.15, outputPer1M: 0.6, cachedInputPer1M: 0.075 },
+  "gpt-4.1": { inputPer1M: 2, outputPer1M: 8, cachedInputPer1M: 1 },
+  "gpt-4.1-nano": { inputPer1M: 0.1, outputPer1M: 0.4, cachedInputPer1M: 0.05 },
+  "text-embedding-3-small": { inputPer1M: 0.02, outputPer1M: 0 },
+  "text-embedding-3-large": { inputPer1M: 0.13, outputPer1M: 0 },
+  // --- Anthropic ---
+  "claude-3-7-sonnet": { inputPer1M: 0.59, outputPer1M: 2.93, cachedInputPer1M: 0.3 },
+  "claude-sonnet-5": { inputPer1M: 2, outputPer1M: 10, cachedInputPer1M: 0.3 },
+  "claude-3-5-sonnet": { inputPer1M: 3, outputPer1M: 15, cachedInputPer1M: 0.3 },
+  "claude-3-5-haiku": { inputPer1M: 0.8, outputPer1M: 4, cachedInputPer1M: 0.08 },
+  "haiku-4.5": { inputPer1M: 1, outputPer1M: 5, cachedInputPer1M: 0.1 },
+  "claude-opus-5": { inputPer1M: 5, outputPer1M: 25, cachedInputPer1M: 1.5 },
+  "claude-3-opus": { inputPer1M: 15, outputPer1M: 75, cachedInputPer1M: 1.5 },
+  // --- Google Gemini ---
+  "gemini-3.7-flash": { inputPer1M: 0.75, outputPer1M: 3.75, cachedInputPer1M: 0.18 },
+  "gemini-3.1-pro": { inputPer1M: 2, outputPer1M: 12, cachedInputPer1M: 0.5 },
+  "gemini-3.5-flash": { inputPer1M: 1.5, outputPer1M: 9, cachedInputPer1M: 0.38 },
+  "gemini-3.1-flash-lite": { inputPer1M: 0.25, outputPer1M: 1.5, cachedInputPer1M: 0.06 },
+  "gemini-2.0-flash": { inputPer1M: 0.1, outputPer1M: 0.4, cachedInputPer1M: 0.025 },
+  "gemini-1.5-pro": { inputPer1M: 1.25, outputPer1M: 5, cachedInputPer1M: 0.3125 },
+  "gemini-1.5-flash": { inputPer1M: 0.075, outputPer1M: 0.3, cachedInputPer1M: 0.01875 },
+  // --- xAI Grok ---
+  "grok-4.6": { inputPer1M: 3, outputPer1M: 15 },
+  "grok-2": { inputPer1M: 2, outputPer1M: 10 },
+  "grok-2-vision": { inputPer1M: 2, outputPer1M: 10 },
+  "grok-beta": { inputPer1M: 5, outputPer1M: 15 },
+  // --- Mistral ---
+  "mistral-large-3": { inputPer1M: 2, outputPer1M: 6 },
+  "mistral-large-latest": { inputPer1M: 2, outputPer1M: 6 },
+  "codestral-latest": { inputPer1M: 0.3, outputPer1M: 0.9 },
+  "mistral-small-latest": { inputPer1M: 0.2, outputPer1M: 0.6 },
+  "ministral-8b-latest": { inputPer1M: 0.1, outputPer1M: 0.1 },
+  // --- Groq LPUs ---
+  "llama-3.3-70b-versatile": { inputPer1M: 0.59, outputPer1M: 0.79 },
+  "llama-3.1-8b-instant": { inputPer1M: 0.05, outputPer1M: 0.08 },
+  "deepseek-r1-distill-llama-70b": { inputPer1M: 0.75, outputPer1M: 0.99 },
+  "qwen-2.5-32b": { inputPer1M: 0.29, outputPer1M: 0.39 },
+  // --- DeepSeek ---
+  "deepseek-v4-pro": { inputPer1M: 0.66, outputPer1M: 1.98, cachedInputPer1M: 0.15 },
+  "deepseek-v4-flash": { inputPer1M: 0.22, outputPer1M: 0.66, cachedInputPer1M: 0.05 },
+  "deepseek-chat": { inputPer1M: 0.22, outputPer1M: 0.66, cachedInputPer1M: 0.05 },
+  "deepseek-reasoner": { inputPer1M: 0.66, outputPer1M: 1.98, cachedInputPer1M: 0.15 },
+  // --- Cohere ---
+  "command-r-plus": { inputPer1M: 2.5, outputPer1M: 10 },
+  "command-r": { inputPer1M: 0.15, outputPer1M: 0.6 }
+};
+var customPricingRegistry = {};
+function normalizeModelKey(rawModel) {
+  if (!rawModel) return "unknown";
+  let model = rawModel.toLowerCase().trim();
+  if (model.includes("/")) {
+    model = model.split("/")[1] || model;
+  }
+  model = model.replace(/-\d{8}$/, "");
+  model = model.replace(/-\d{4}-\d{2}-\d{2}$/, "");
+  return model;
+}
+function getModelPricing(modelName) {
+  const normalized = normalizeModelKey(modelName);
+  if (customPricingRegistry[normalized]) {
+    return customPricingRegistry[normalized];
+  }
+  if (customPricingRegistry[modelName]) {
+    return customPricingRegistry[modelName];
+  }
+  if (MODEL_PRICING_TABLE[normalized]) {
+    return MODEL_PRICING_TABLE[normalized];
+  }
+  if (MODEL_PRICING_TABLE[modelName]) {
+    return MODEL_PRICING_TABLE[modelName];
+  }
+  return {
+    inputPer1M: 1,
+    outputPer1M: 3,
+    cachedInputPer1M: 0.5
+  };
+}
+
+// src/pricing/calculator.ts
+function calculateCost(params) {
+  const rates = getModelPricing(params.model);
+  const inputTokens = params.inputTokens ?? 0;
+  const outputTokens = params.outputTokens ?? 0;
+  const reasoningTokens = params.reasoningTokens ?? 0;
+  const cachedTokens = params.cachedTokens ?? 0;
+  const regularInputTokens = Math.max(0, inputTokens - cachedTokens);
+  const regularInputCost = regularInputTokens / 1e6 * rates.inputPer1M;
+  const cachedRate = rates.cachedInputPer1M ?? rates.inputPer1M * 0.5;
+  const cachedInputCost = cachedTokens / 1e6 * cachedRate;
+  const inputCostUSD = regularInputCost + cachedInputCost;
+  const outputCostUSD = outputTokens / 1e6 * rates.outputPer1M;
+  const reasoningRate = rates.reasoningPer1M ?? rates.outputPer1M;
+  const reasoningCostUSD = reasoningTokens / 1e6 * reasoningRate;
+  const standardCacheCost = cachedTokens / 1e6 * rates.inputPer1M;
+  const cachedDiscountUSD = Math.max(0, standardCacheCost - cachedInputCost);
+  const totalUSD = inputCostUSD + outputCostUSD;
+  const markup = params.markupMultiplier ?? 1;
+  const retailUSD = markup !== 1 ? totalUSD * markup : void 0;
+  return {
+    inputCostUSD: Number(inputCostUSD.toFixed(8)),
+    outputCostUSD: Number(outputCostUSD.toFixed(8)),
+    reasoningCostUSD: reasoningTokens > 0 ? Number(reasoningCostUSD.toFixed(8)) : void 0,
+    cachedDiscountUSD: cachedTokens > 0 ? Number(cachedDiscountUSD.toFixed(8)) : void 0,
+    totalUSD: Number(totalUSD.toFixed(8)),
+    retailUSD: retailUSD ? Number(retailUSD.toFixed(8)) : void 0,
+    currency: rates.currency || "USD"
+  };
+}
+function calculateUsageCost(model, usage, markupMultiplier) {
+  return calculateCost({
+    model,
+    inputTokens: usage.inputTokens,
+    outputTokens: usage.outputTokens,
+    reasoningTokens: usage.reasoningTokens,
+    cachedTokens: usage.cachedTokens,
+    cacheWriteTokens: usage.cacheWriteTokens,
+    markupMultiplier
+  });
+}
+
+// src/meter/stream.ts
+function wrapOpenAIStream(stream, options, onComplete) {
+  let detectedModel = options.model || "gpt-4o";
+  let finalUsage = null;
+  const wrappedAsyncIterable = {
+    async *[Symbol.asyncIterator]() {
+      try {
+        for await (const chunk of stream) {
+          const inspected = inspectOpenAIStreamChunk(chunk);
+          if (inspected.model) {
+            detectedModel = inspected.model;
+          }
+          if (inspected.usage) {
+            finalUsage = inspected.usage;
+          }
+          yield chunk;
+        }
+      } finally {
+        if (finalUsage) {
+          const cost = calculateUsageCost(detectedModel, finalUsage);
+          const customerId = typeof options.customer === "string" ? options.customer : options.customer?.id || options.customerId;
+          const event = {
+            timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+            model: detectedModel,
+            provider: "openai",
+            usage: finalUsage,
+            cost,
+            customerId,
+            metadata: options.metadata
+          };
+          onComplete(event);
+          if (options.onUsage) {
+            options.onUsage(event);
+          }
+        }
+      }
+    }
+  };
+  return wrappedAsyncIterable;
+}
+function wrapAnthropicStream(stream, options, onComplete) {
+  const accumulator = new AnthropicStreamAccumulator();
+  const wrappedAsyncIterable = {
+    async *[Symbol.asyncIterator]() {
+      try {
+        for await (const event of stream) {
+          accumulator.processEvent(event);
+          yield event;
+        }
+      } finally {
+        const extracted = accumulator.getUsage();
+        const model = options.model || extracted.model;
+        const cost = calculateUsageCost(model, extracted.usage);
+        const customerId = typeof options.customer === "string" ? options.customer : options.customer?.id || options.customerId;
+        const usageEvent = {
+          timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+          model,
+          provider: "anthropic",
+          usage: extracted.usage,
+          cost,
+          customerId,
+          metadata: options.metadata
+        };
+        onComplete(usageEvent);
+        if (options.onUsage) {
+          options.onUsage(usageEvent);
+        }
+      }
+    }
+  };
+  return wrappedAsyncIterable;
+}
+function wrapGeminiStream(result, options, onComplete) {
+  if (!result || !result.stream) return result;
+  const originalStream = result.stream;
+  const model = options.model || "gemini-3.7-flash";
+  let lastChunkWithUsage = null;
+  const wrappedStream = (async function* () {
+    try {
+      for await (const chunk of originalStream) {
+        if (chunk.usageMetadata) {
+          lastChunkWithUsage = chunk;
+        }
+        yield chunk;
+      }
+    } finally {
+      if (lastChunkWithUsage) {
+        const extracted = extractGeminiResponseUsage(lastChunkWithUsage, model);
+        if (extracted) {
+          const cost = calculateUsageCost(extracted.model, extracted.usage);
+          const customerId = typeof options.customer === "string" ? options.customer : options.customer?.id || options.customerId;
+          const event = {
+            timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+            model: extracted.model,
+            provider: "google",
+            usage: extracted.usage,
+            cost,
+            customerId,
+            metadata: options.metadata
+          };
+          onComplete(event);
+          if (options.onUsage) {
+            options.onUsage(event);
+          }
+        }
+      }
+    }
+  })();
+  return {
+    ...result,
+    stream: wrappedStream
+  };
+}
+function wrapUniversalStream(stream, options = {}, onComplete) {
+  if (!stream || typeof stream !== "object") return stream;
+  if ("stream" in stream && "response" in stream) {
+    return wrapGeminiStream(stream, options, onComplete);
+  }
+  if (Symbol.asyncIterator in stream) {
+    if (options.provider === "anthropic") {
+      return wrapAnthropicStream(stream, options, onComplete);
+    }
+    return wrapOpenAIStream(stream, options, onComplete);
+  }
+  return stream;
+}
+
+// src/meter/client.ts
+var VibezMeter = class {
+  batcher;
+  stripeClient;
+  markupMultiplier;
+  constructor(options = {}) {
+    this.markupMultiplier = options.markupMultiplier;
+    if (options.stripe) {
+      this.stripeClient = options.stripe;
+    } else if (options.apiKey || process.env.STRIPE_SECRET_KEY) {
+      const key = options.apiKey || process.env.STRIPE_SECRET_KEY;
+      this.stripeClient = new Stripe(key, {
+        appInfo: {
+          name: "vibezcheck",
+          version: "0.1.0",
+          url: "https://vibezcheck.xyz"
+        }
+      });
+    }
+    this.batcher = new MeterBatcher({
+      ...options,
+      stripe: this.stripeClient
+    });
+  }
+  /**
+   * Track token usage from a non-streaming response object (OpenAI, Anthropic, Gemini, etc.)
+   */
+  trackUsage(response, options = {}) {
+    const extracted = detectAndExtractUsage(response, options.model, options.provider);
+    if (!extracted) {
+      return null;
+    }
+    const model = options.model || extracted.model;
+    const cost = calculateUsageCost(model, extracted.usage, this.markupMultiplier);
+    const customerId = typeof options.customer === "string" ? options.customer : options.customer?.id || options.customerId;
+    const event = {
+      timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+      model,
+      provider: extracted.provider,
+      usage: extracted.usage,
+      cost,
+      customerId,
+      metadata: options.metadata
+    };
+    this.batcher.enqueue(event);
+    return event;
+  }
+  /**
+   * Wrap any LLM stream (OpenAI, Anthropic, Gemini) with zero added latency
+   */
+  wrapStream(stream, options = {}) {
+    return wrapUniversalStream(stream, options, (event) => {
+      this.batcher.enqueue(event);
+    });
+  }
+  /**
+   * Directly record token usage manually
+   */
+  recordUsage(options) {
+    const inputTokens = options.inputTokens ?? 0;
+    const outputTokens = options.outputTokens ?? 0;
+    const reasoningTokens = options.reasoningTokens;
+    const cachedTokens = options.cachedTokens;
+    const usage = {
+      inputTokens,
+      outputTokens,
+      totalTokens: inputTokens + outputTokens,
+      reasoningTokens,
+      visibleOutputTokens: reasoningTokens !== void 0 ? Math.max(0, outputTokens - reasoningTokens) : outputTokens,
+      cachedTokens
+    };
+    const cost = calculateUsageCost(options.model, usage, this.markupMultiplier);
+    const customerId = typeof options.customer === "string" ? options.customer : options.customer?.id || options.customerId;
+    const event = {
+      timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+      model: options.model,
+      provider: options.provider || "custom",
+      usage,
+      cost,
+      customerId,
+      metadata: options.metadata
+    };
+    this.batcher.enqueue(event);
+    return event;
+  }
+  /**
+   * Flush pending events to Stripe (vital for Serverless & Edge environments)
+   */
+  async flush() {
+    await this.batcher.flush();
+  }
+  /**
+   * Get in-memory aggregated usage statistics
+   */
+  getUsageSummary() {
+    return this.batcher.getSummary();
+  }
+  /**
+   * Reset in-memory ledger
+   */
+  resetSummary() {
+    this.batcher.resetLedger();
+  }
+};
+function createMeter(options = {}) {
+  return new VibezMeter(options);
+}
+
+// src/ai-sdk/with-billing.ts
+function withBilling(model, options = {}) {
+  if (!model || typeof model !== "object") {
+    return model;
+  }
+  const meter = options.meter || createMeter({
+    apiKey: options.stripeApiKey,
+    eventName: options.eventName
+  });
+  const customerId = typeof options.customer === "string" ? options.customer : options.customer?.id || options.customerId;
+  const modelId = model.modelId || "unknown-model";
+  const provider = model.provider?.replace(/^@ai-sdk\//, "") || "ai-sdk";
+  const handleUsage = (rawUsage) => {
+    if (!rawUsage) return;
+    const inputTokens = rawUsage.promptTokens ?? rawUsage.inputTokens ?? 0;
+    const outputTokens = rawUsage.completionTokens ?? rawUsage.outputTokens ?? 0;
+    const reasoningTokens = rawUsage.reasoningTokens ?? rawUsage.completionTokensDetails?.reasoningTokens ?? rawUsage.outputTokenDetails?.reasoningTokens ?? 0;
+    const cachedTokens = rawUsage.promptTokensDetails?.cachedTokens ?? rawUsage.inputTokenDetails?.cachedTokens ?? 0;
+    const usage = {
+      inputTokens,
+      outputTokens,
+      totalTokens: inputTokens + outputTokens,
+      reasoningTokens: reasoningTokens > 0 ? reasoningTokens : void 0,
+      visibleOutputTokens: Math.max(0, outputTokens - reasoningTokens),
+      cachedTokens: cachedTokens > 0 ? cachedTokens : void 0
+    };
+    const cost = calculateUsageCost(modelId, usage);
+    const event = {
+      timestamp: (/* @__PURE__ */ new Date()).toISOString(),
+      model: modelId,
+      provider,
+      usage,
+      cost,
+      customerId,
+      metadata: options.metadata
+    };
+    meter.recordUsage({
+      model: modelId,
+      provider,
+      inputTokens,
+      outputTokens,
+      reasoningTokens: usage.reasoningTokens,
+      cachedTokens: usage.cachedTokens,
+      customerId,
+      metadata: options.metadata
+    });
+    if (options.onUsage) {
+      options.onUsage(event);
+    }
+  };
+  return new Proxy(model, {
+    get(target, prop, receiver) {
+      const originalValue = Reflect.get(target, prop, receiver);
+      if (prop === "doGenerate" && typeof originalValue === "function") {
+        return async function(...args) {
+          const result = await originalValue.apply(target, args);
+          if (result && result.usage) {
+            handleUsage(result.usage);
+          }
+          return result;
+        };
+      }
+      if (prop === "doStream" && typeof originalValue === "function") {
+        return async function(...args) {
+          const result = await originalValue.apply(target, args);
+          if (!result || !result.stream) {
+            return result;
+          }
+          const originalStream = result.stream;
+          if (typeof originalStream.getReader === "function") {
+            const reader = originalStream.getReader();
+            const transformedStream = new ReadableStream({
+              async start(controller) {
+                try {
+                  while (true) {
+                    const { done, value } = await reader.read();
+                    if (done) {
+                      controller.close();
+                      break;
+                    }
+                    if (value && typeof value === "object") {
+                      if (value.type === "finish" && value.usage) {
+                        handleUsage(value.usage);
+                      }
+                    }
+                    controller.enqueue(value);
+                  }
+                } catch (err) {
+                  controller.error(err);
+                }
+              }
+            });
+            return {
+              ...result,
+              stream: transformedStream
+            };
+          }
+          if (Symbol.asyncIterator in originalStream) {
+            const wrappedAsyncIterable = {
+              async *[Symbol.asyncIterator]() {
+                for await (const chunk of originalStream) {
+                  if (chunk && typeof chunk === "object") {
+                    if (chunk.type === "finish" && chunk.usage) {
+                      handleUsage(chunk.usage);
+                    }
+                  }
+                  yield chunk;
+                }
+              }
+            };
+            return {
+              ...result,
+              stream: wrappedAsyncIterable
+            };
+          }
+          return result;
+        };
+      }
+      return originalValue;
+    }
+  });
+}
+var meteredModel = withBilling;
+export {
+  meteredModel,
+  withBilling
+};
+//# sourceMappingURL=index.mjs.map