@llmops/core 0.1.5-beta.1 → 0.1.5

package/dist/index.mjs

@@ -1,5 +1,6 @@
-import { A as record, C as _enum, D as literal, E as boolean, M as union, N as unknown, O as number, S as zod_default, T as array, _ as schemas, a as matchType, b as variantsSchema, c as parsePartialTableData, d as validateTableData, f as SCHEMA_METADATA, g as environmentsSchema, h as environmentSecretsSchema, i as getMigrations, j as string, k as object, l as parseTableData, m as configsSchema, n as createDatabaseFromConnection, o as runAutoMigrations, p as configVariantsSchema, r as detectDatabaseType, s as logger, t as createDatabase, u as validatePartialTableData, v as targetingRulesSchema, w as any, x as workspaceSettingsSchema, y as variantVersionsSchema } from "./db-CGY-vZ3u.mjs";
+import { A as object, C as zod_default, D as boolean, E as array, M as string, N as union, O as literal, P as unknown, S as workspaceSettingsSchema, T as any, _ as llmRequestsSchema, a as matchType, b as variantVersionsSchema, c as parsePartialTableData, d as validateTableData, f as SCHEMA_METADATA, g as environmentsSchema, h as environmentSecretsSchema, i as getMigrations, j as record, k as number, l as parseTableData, m as configsSchema, n as createDatabaseFromConnection, o as runAutoMigrations, p as configVariantsSchema, r as detectDatabaseType, s as logger, t as createDatabase, u as validatePartialTableData, v as schemas, w as _enum, x as variantsSchema, y as targetingRulesSchema } from "./db-DSzwrW4p.mjs";
 import gateway from "@llmops/gateway";
+import { sql } from "kysely";
 import * as fs from "node:fs/promises";
 import * as path from "node:path";
 import { createRandomStringGenerator } from "@better-auth/utils/random";
@@ -1279,7 +1280,11 @@ const createConfigVariantDataLayer = (db) => {
 				"version"
 			]).where("variantId", "=", configVariant.variantId).orderBy("version", "desc").limit(1).executeTakeFirst();
 			if (!versionData) throw new LLMOpsError(`No variant version found for variant ${configVariant.variantId}`);
-			return versionData;
+			return {
+				...versionData,
+				configId: resolvedConfigId,
+				variantId: configVariant.variantId
+			};
 		}
 	};
 };
@@ -1439,6 +1444,287 @@ const createEnvironmentSecretDataLayer = (db) => {
 	};
 };
 
+//#endregion
+//#region src/datalayer/llmRequests.ts
+/**
+* Schema for inserting a new LLM request log
+*/
+const insertLLMRequestSchema = zod_default.object({
+	requestId: zod_default.string().uuid(),
+	configId: zod_default.string().uuid().nullable().optional(),
+	variantId: zod_default.string().uuid().nullable().optional(),
+	provider: zod_default.string(),
+	model: zod_default.string(),
+	promptTokens: zod_default.number().int().default(0),
+	completionTokens: zod_default.number().int().default(0),
+	totalTokens: zod_default.number().int().default(0),
+	cachedTokens: zod_default.number().int().default(0),
+	cost: zod_default.number().int().default(0),
+	inputCost: zod_default.number().int().default(0),
+	outputCost: zod_default.number().int().default(0),
+	endpoint: zod_default.string(),
+	statusCode: zod_default.number().int(),
+	latencyMs: zod_default.number().int().default(0),
+	isStreaming: zod_default.boolean().default(false),
+	userId: zod_default.string().nullable().optional(),
+	tags: zod_default.record(zod_default.string(), zod_default.string()).default({})
+});
+/**
+* Schema for listing LLM requests
+*/
+const listRequestsSchema = zod_default.object({
+	limit: zod_default.number().int().positive().max(1e3).default(100),
+	offset: zod_default.number().int().nonnegative().default(0),
+	configId: zod_default.string().uuid().optional(),
+	provider: zod_default.string().optional(),
+	model: zod_default.string().optional(),
+	startDate: zod_default.date().optional(),
+	endDate: zod_default.date().optional()
+});
+/**
+* Schema for date range queries
+*/
+const dateRangeSchema = zod_default.object({
+	startDate: zod_default.date(),
+	endDate: zod_default.date()
+});
+/**
+* Schema for cost summary with grouping
+*/
+const costSummarySchema = zod_default.object({
+	startDate: zod_default.date(),
+	endDate: zod_default.date(),
+	groupBy: zod_default.enum([
+		"day",
+		"hour",
+		"model",
+		"provider",
+		"config"
+	]).optional()
+});
+/**
+* Helper to create column reference for SQL
+* Uses sql.ref() to properly quote column names for the database
+*/
+const col = (name) => sql.ref(name);
+const tableCol = (table, name) => sql.ref(`${table}.${name}`);
+const createLLMRequestsDataLayer = (db) => {
+	return {
+		batchInsertRequests: async (requests) => {
+			if (requests.length === 0) return { count: 0 };
+			const validatedRequests = await Promise.all(requests.map(async (req) => {
+				const result = await insertLLMRequestSchema.safeParseAsync(req);
+				if (!result.success) throw new LLMOpsError(`Invalid request data: ${result.error.message}`);
+				return result.data;
+			}));
+			const now = (/* @__PURE__ */ new Date()).toISOString();
+			const values = validatedRequests.map((req) => ({
+				id: randomUUID(),
+				requestId: req.requestId,
+				configId: req.configId ?? null,
+				variantId: req.variantId ?? null,
+				provider: req.provider,
+				model: req.model,
+				promptTokens: req.promptTokens,
+				completionTokens: req.completionTokens,
+				totalTokens: req.totalTokens,
+				cachedTokens: req.cachedTokens,
+				cost: req.cost,
+				inputCost: req.inputCost,
+				outputCost: req.outputCost,
+				endpoint: req.endpoint,
+				statusCode: req.statusCode,
+				latencyMs: req.latencyMs,
+				isStreaming: req.isStreaming,
+				userId: req.userId ?? null,
+				tags: JSON.stringify(req.tags),
+				createdAt: now,
+				updatedAt: now
+			}));
+			await db.insertInto("llm_requests").values(values).execute();
+			return { count: values.length };
+		},
+		insertRequest: async (request) => {
+			const result = await insertLLMRequestSchema.safeParseAsync(request);
+			if (!result.success) throw new LLMOpsError(`Invalid request data: ${result.error.message}`);
+			const req = result.data;
+			const now = (/* @__PURE__ */ new Date()).toISOString();
+			return db.insertInto("llm_requests").values({
+				id: randomUUID(),
+				requestId: req.requestId,
+				configId: req.configId ?? null,
+				variantId: req.variantId ?? null,
+				provider: req.provider,
+				model: req.model,
+				promptTokens: req.promptTokens,
+				completionTokens: req.completionTokens,
+				totalTokens: req.totalTokens,
+				cachedTokens: req.cachedTokens,
+				cost: req.cost,
+				inputCost: req.inputCost,
+				outputCost: req.outputCost,
+				endpoint: req.endpoint,
+				statusCode: req.statusCode,
+				latencyMs: req.latencyMs,
+				isStreaming: req.isStreaming,
+				userId: req.userId ?? null,
+				tags: JSON.stringify(req.tags),
+				createdAt: now,
+				updatedAt: now
+			}).returningAll().executeTakeFirst();
+		},
+		listRequests: async (params) => {
+			const result = await listRequestsSchema.safeParseAsync(params || {});
+			if (!result.success) throw new LLMOpsError(`Invalid parameters: ${result.error.message}`);
+			const { limit, offset, configId, provider, model, startDate, endDate } = result.data;
+			let baseQuery = db.selectFrom("llm_requests");
+			if (configId) baseQuery = baseQuery.where("configId", "=", configId);
+			if (provider) baseQuery = baseQuery.where("provider", "=", provider);
+			if (model) baseQuery = baseQuery.where("model", "=", model);
+			if (startDate) baseQuery = baseQuery.where(sql`${col("createdAt")} >= ${startDate.toISOString()}`);
+			if (endDate) baseQuery = baseQuery.where(sql`${col("createdAt")} <= ${endDate.toISOString()}`);
+			const countResult = await baseQuery.select(sql`COUNT(*)`.as("total")).executeTakeFirst();
+			const total = Number(countResult?.total ?? 0);
+			return {
+				data: await baseQuery.selectAll().orderBy("createdAt", "desc").limit(limit).offset(offset).execute(),
+				total,
+				limit,
+				offset
+			};
+		},
+		getRequestByRequestId: async (requestId) => {
+			return db.selectFrom("llm_requests").selectAll().where("requestId", "=", requestId).executeTakeFirst();
+		},
+		getTotalCost: async (params) => {
+			const result = await dateRangeSchema.safeParseAsync(params);
+			if (!result.success) throw new LLMOpsError(`Invalid parameters: ${result.error.message}`);
+			const { startDate, endDate } = result.data;
+			return await db.selectFrom("llm_requests").select([
+				sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+				sql`COALESCE(SUM(${col("inputCost")}), 0)`.as("totalInputCost"),
+				sql`COALESCE(SUM(${col("outputCost")}), 0)`.as("totalOutputCost"),
+				sql`COALESCE(SUM(${col("promptTokens")}), 0)`.as("totalPromptTokens"),
+				sql`COALESCE(SUM(${col("completionTokens")}), 0)`.as("totalCompletionTokens"),
+				sql`COALESCE(SUM(${col("totalTokens")}), 0)`.as("totalTokens"),
+				sql`COUNT(*)`.as("requestCount")
+			]).where(sql`${col("createdAt")} >= ${startDate.toISOString()}`).where(sql`${col("createdAt")} <= ${endDate.toISOString()}`).executeTakeFirst();
+		},
+		getCostByModel: async (params) => {
+			const result = await dateRangeSchema.safeParseAsync(params);
+			if (!result.success) throw new LLMOpsError(`Invalid parameters: ${result.error.message}`);
+			const { startDate, endDate } = result.data;
+			return db.selectFrom("llm_requests").select([
+				"provider",
+				"model",
+				sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+				sql`COALESCE(SUM(${col("inputCost")}), 0)`.as("totalInputCost"),
+				sql`COALESCE(SUM(${col("outputCost")}), 0)`.as("totalOutputCost"),
+				sql`COALESCE(SUM(${col("totalTokens")}), 0)`.as("totalTokens"),
+				sql`COUNT(*)`.as("requestCount"),
+				sql`AVG(${col("latencyMs")})`.as("avgLatencyMs")
+			]).where(sql`${col("createdAt")} >= ${startDate.toISOString()}`).where(sql`${col("createdAt")} <= ${endDate.toISOString()}`).groupBy(["provider", "model"]).orderBy(sql`SUM(${col("cost")})`, "desc").execute();
+		},
+		getCostByProvider: async (params) => {
+			const result = await dateRangeSchema.safeParseAsync(params);
+			if (!result.success) throw new LLMOpsError(`Invalid parameters: ${result.error.message}`);
+			const { startDate, endDate } = result.data;
+			return db.selectFrom("llm_requests").select([
+				"provider",
+				sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+				sql`COALESCE(SUM(${col("inputCost")}), 0)`.as("totalInputCost"),
+				sql`COALESCE(SUM(${col("outputCost")}), 0)`.as("totalOutputCost"),
+				sql`COALESCE(SUM(${col("totalTokens")}), 0)`.as("totalTokens"),
+				sql`COUNT(*)`.as("requestCount"),
+				sql`AVG(${col("latencyMs")})`.as("avgLatencyMs")
+			]).where(sql`${col("createdAt")} >= ${startDate.toISOString()}`).where(sql`${col("createdAt")} <= ${endDate.toISOString()}`).groupBy("provider").orderBy(sql`SUM(${col("cost")})`, "desc").execute();
+		},
+		getCostByConfig: async (params) => {
+			const result = await dateRangeSchema.safeParseAsync(params);
+			if (!result.success) throw new LLMOpsError(`Invalid parameters: ${result.error.message}`);
+			const { startDate, endDate } = result.data;
+			return db.selectFrom("llm_requests").leftJoin("configs", "llm_requests.configId", "configs.id").select([
+				"llm_requests.configId",
+				"configs.name as configName",
+				"configs.slug as configSlug",
+				sql`COALESCE(SUM(${tableCol("llm_requests", "cost")}), 0)`.as("totalCost"),
+				sql`COALESCE(SUM(${tableCol("llm_requests", "inputCost")}), 0)`.as("totalInputCost"),
+				sql`COALESCE(SUM(${tableCol("llm_requests", "outputCost")}), 0)`.as("totalOutputCost"),
+				sql`COALESCE(SUM(${tableCol("llm_requests", "totalTokens")}), 0)`.as("totalTokens"),
+				sql`COUNT(*)`.as("requestCount")
+			]).where(sql`${tableCol("llm_requests", "createdAt")} >= ${startDate.toISOString()}`).where(sql`${tableCol("llm_requests", "createdAt")} <= ${endDate.toISOString()}`).groupBy([
+				"llm_requests.configId",
+				"configs.name",
+				"configs.slug"
+			]).orderBy(sql`SUM(${tableCol("llm_requests", "cost")})`, "desc").execute();
+		},
+		getDailyCosts: async (params) => {
+			const result = await dateRangeSchema.safeParseAsync(params);
+			if (!result.success) throw new LLMOpsError(`Invalid parameters: ${result.error.message}`);
+			const { startDate, endDate } = result.data;
+			return db.selectFrom("llm_requests").select([
+				sql`DATE(${col("createdAt")})`.as("date"),
+				sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+				sql`COALESCE(SUM(${col("inputCost")}), 0)`.as("totalInputCost"),
+				sql`COALESCE(SUM(${col("outputCost")}), 0)`.as("totalOutputCost"),
+				sql`COALESCE(SUM(${col("totalTokens")}), 0)`.as("totalTokens"),
+				sql`COUNT(*)`.as("requestCount")
+			]).where(sql`${col("createdAt")} >= ${startDate.toISOString()}`).where(sql`${col("createdAt")} <= ${endDate.toISOString()}`).groupBy(sql`DATE(${col("createdAt")})`).orderBy(sql`DATE(${col("createdAt")})`, "asc").execute();
+		},
+		getCostSummary: async (params) => {
+			const result = await costSummarySchema.safeParseAsync(params);
+			if (!result.success) throw new LLMOpsError(`Invalid parameters: ${result.error.message}`);
+			const { startDate, endDate, groupBy } = result.data;
+			const baseQuery = db.selectFrom("llm_requests").where(sql`${col("createdAt")} >= ${startDate.toISOString()}`).where(sql`${col("createdAt")} <= ${endDate.toISOString()}`);
+			switch (groupBy) {
+				case "day": return baseQuery.select([
+					sql`DATE(${col("createdAt")})`.as("groupKey"),
+					sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+					sql`COUNT(*)`.as("requestCount")
+				]).groupBy(sql`DATE(${col("createdAt")})`).orderBy(sql`DATE(${col("createdAt")})`, "asc").execute();
+				case "hour": return baseQuery.select([
+					sql`DATE_TRUNC('hour', ${col("createdAt")})`.as("groupKey"),
+					sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+					sql`COUNT(*)`.as("requestCount")
+				]).groupBy(sql`DATE_TRUNC('hour', ${col("createdAt")})`).orderBy(sql`DATE_TRUNC('hour', ${col("createdAt")})`, "asc").execute();
+				case "model": return baseQuery.select([
+					sql`${col("provider")} || '/' || ${col("model")}`.as("groupKey"),
+					sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+					sql`COUNT(*)`.as("requestCount")
+				]).groupBy(["provider", "model"]).orderBy(sql`SUM(${col("cost")})`, "desc").execute();
+				case "provider": return baseQuery.select([
+					sql`${col("provider")}`.as("groupKey"),
+					sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+					sql`COUNT(*)`.as("requestCount")
+				]).groupBy("provider").orderBy(sql`SUM(${col("cost")})`, "desc").execute();
+				case "config": return baseQuery.select([
+					sql`COALESCE(${col("configId")}::text, 'no-config')`.as("groupKey"),
+					sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+					sql`COUNT(*)`.as("requestCount")
+				]).groupBy("configId").orderBy(sql`SUM(${col("cost")})`, "desc").execute();
+				default: return baseQuery.select([
+					sql`'total'`.as("groupKey"),
+					sql`COALESCE(SUM(${col("cost")}), 0)`.as("totalCost"),
+					sql`COUNT(*)`.as("requestCount")
+				]).execute();
+			}
+		},
+		getRequestStats: async (params) => {
+			const result = await dateRangeSchema.safeParseAsync(params);
+			if (!result.success) throw new LLMOpsError(`Invalid parameters: ${result.error.message}`);
+			const { startDate, endDate } = result.data;
+			return await db.selectFrom("llm_requests").select([
+				sql`COUNT(*)`.as("totalRequests"),
+				sql`COUNT(CASE WHEN ${col("statusCode")} >= 200 AND ${col("statusCode")} < 300 THEN 1 END)`.as("successfulRequests"),
+				sql`COUNT(CASE WHEN ${col("statusCode")} >= 400 THEN 1 END)`.as("failedRequests"),
+				sql`COUNT(CASE WHEN ${col("isStreaming")} = true THEN 1 END)`.as("streamingRequests"),
+				sql`AVG(${col("latencyMs")})`.as("avgLatencyMs"),
+				sql`MAX(${col("latencyMs")})`.as("maxLatencyMs"),
+				sql`MIN(${col("latencyMs")})`.as("minLatencyMs")
+			]).where(sql`${col("createdAt")} >= ${startDate.toISOString()}`).where(sql`${col("createdAt")} <= ${endDate.toISOString()}`).executeTakeFirst();
+		}
+	};
+};
+
 //#endregion
 //#region src/datalayer/targetingRules.ts
 const createTargetingRule = zod_default.object({
@@ -1865,6 +2151,7 @@ const createDataLayer = async (db) => {
 		...createConfigVariantDataLayer(db),
 		...createEnvironmentDataLayer(db),
 		...createEnvironmentSecretDataLayer(db),
+		...createLLMRequestsDataLayer(db),
 		...createTargetingRulesDataLayer(db),
 		...createVariantDataLayer(db),
 		...createVariantVersionsDataLayer(db),
@@ -1873,4 +2160,207 @@ const createDataLayer = async (db) => {
 };
 
 //#endregion
-export { CacheService, FileCacheBackend, MS, MemoryCacheBackend, SCHEMA_METADATA, SupportedProviders, chatCompletionCreateParamsBaseSchema, configVariantsSchema, configsSchema, createDataLayer, createDatabase, createDatabaseFromConnection, detectDatabaseType, environmentSecretsSchema, environmentsSchema, gateway, generateId, getMigrations, llmopsConfigSchema, logger, matchType, parsePartialTableData, parseTableData, runAutoMigrations, schemas, targetingRulesSchema, validateLLMOpsConfig, validatePartialTableData, validateTableData, variantJsonDataSchema, variantVersionsSchema, variantsSchema, workspaceSettingsSchema };
+//#region src/pricing/calculator.ts
+/**
+* Calculate the cost of an LLM request in micro-dollars
+*
+* Micro-dollars are used to avoid floating-point precision issues:
+* - 1 dollar = 1,000,000 micro-dollars
+* - $0.001 = 1,000 micro-dollars
+* - $0.000001 = 1 micro-dollar
+*
+* @param usage - Token usage data from the LLM response
+* @param pricing - Model pricing information
+* @returns Cost breakdown in micro-dollars
+*
+* @example
+* ```typescript
+* const usage = { promptTokens: 1000, completionTokens: 500 };
+* const pricing = { inputCostPer1M: 2.5, outputCostPer1M: 10.0 };
+* const cost = calculateCost(usage, pricing);
+* // cost = { inputCost: 2500, outputCost: 5000, totalCost: 7500 }
+* // In dollars: $0.0025 input + $0.005 output = $0.0075 total
+* ```
+*/
+function calculateCost(usage, pricing) {
+	const inputCost = Math.round(usage.promptTokens * pricing.inputCostPer1M);
+	const outputCost = Math.round(usage.completionTokens * pricing.outputCostPer1M);
+	return {
+		inputCost,
+		outputCost,
+		totalCost: inputCost + outputCost
+	};
+}
+/**
+* Convert micro-dollars to dollars
+*
+* @param microDollars - Amount in micro-dollars
+* @returns Amount in dollars
+*
+* @example
+* ```typescript
+* microDollarsToDollars(7500); // 0.0075
+* microDollarsToDollars(1000000); // 1.0
+* ```
+*/
+function microDollarsToDollars(microDollars) {
+	return microDollars / 1e6;
+}
+/**
+* Convert dollars to micro-dollars
+*
+* @param dollars - Amount in dollars
+* @returns Amount in micro-dollars (rounded to nearest integer)
+*
+* @example
+* ```typescript
+* dollarsToMicroDollars(0.0075); // 7500
+* dollarsToMicroDollars(1.0); // 1000000
+* ```
+*/
+function dollarsToMicroDollars(dollars) {
+	return Math.round(dollars * 1e6);
+}
+/**
+* Format micro-dollars as a human-readable dollar string
+*
+* @param microDollars - Amount in micro-dollars
+* @param decimals - Number of decimal places (default: 6)
+* @returns Formatted dollar string
+*
+* @example
+* ```typescript
+* formatCost(7500); // "$0.007500"
+* formatCost(1234567, 2); // "$1.23"
+* ```
+*/
+function formatCost(microDollars, decimals = 6) {
+	return `$${microDollarsToDollars(microDollars).toFixed(decimals)}`;
+}
+
+//#endregion
+//#region src/pricing/provider.ts
+const MODELS_DEV_API = "https://models.dev/api.json";
+/**
+* Pricing provider that fetches data from models.dev API
+*
+* Features:
+* - Caches pricing data with configurable TTL (default 5 minutes)
+* - Supports fallback to local cache on fetch failure
+* - Thread-safe cache refresh
+*/
+var ModelsDevPricingProvider = class {
+	cache = /* @__PURE__ */ new Map();
+	lastFetch = 0;
+	cacheTTL;
+	fetchPromise = null;
+	ready = false;
+	/**
+	* Create a new ModelsDevPricingProvider
+	*
+	* @param cacheTTL - Cache TTL in milliseconds (default: 5 minutes)
+	*/
+	constructor(cacheTTL = 300 * 1e3) {
+		this.cacheTTL = cacheTTL;
+	}
+	/**
+	* Generate a cache key for a provider/model combination
+	*/
+	getCacheKey(provider, model) {
+		return `${provider.toLowerCase()}:${model.toLowerCase()}`;
+	}
+	/**
+	* Fetch pricing data from models.dev API
+	*/
+	async fetchPricingData() {
+		try {
+			logger.debug("[Pricing] Fetching pricing data from models.dev");
+			const response = await fetch(MODELS_DEV_API);
+			if (!response.ok) throw new Error(`Failed to fetch models.dev API: ${response.status}`);
+			const data = await response.json();
+			this.cache.clear();
+			for (const [providerId, provider] of Object.entries(data)) {
+				if (!provider.models) continue;
+				for (const [_modelId, model] of Object.entries(provider.models)) {
+					if (!model.cost) continue;
+					const cacheKey = this.getCacheKey(providerId, model.id);
+					this.cache.set(cacheKey, {
+						inputCostPer1M: model.cost.input ?? 0,
+						outputCostPer1M: model.cost.output ?? 0,
+						cacheReadCostPer1M: model.cost.cache_read,
+						cacheWriteCostPer1M: model.cost.cache_write,
+						reasoningCostPer1M: model.cost.reasoning
+					});
+					const nameKey = this.getCacheKey(providerId, model.name);
+					if (nameKey !== cacheKey) this.cache.set(nameKey, this.cache.get(cacheKey));
+				}
+			}
+			this.lastFetch = Date.now();
+			this.ready = true;
+			logger.debug(`[Pricing] Cached pricing for ${this.cache.size} models from models.dev`);
+		} catch (error) {
+			logger.error(`[Pricing] Failed to fetch pricing data: ${error instanceof Error ? error.message : String(error)}`);
+			if (this.cache.size === 0) throw error;
+		}
+	}
+	/**
+	* Ensure cache is fresh, fetching if necessary
+	*/
+	async ensureFreshCache() {
+		if (!(Date.now() - this.lastFetch > this.cacheTTL) && this.cache.size > 0) return;
+		if (!this.fetchPromise) this.fetchPromise = this.fetchPricingData().finally(() => {
+			this.fetchPromise = null;
+		});
+		await this.fetchPromise;
+	}
+	/**
+	* Get pricing for a specific model
+	*/
+	async getModelPricing(provider, model) {
+		await this.ensureFreshCache();
+		const cacheKey = this.getCacheKey(provider, model);
+		const pricing = this.cache.get(cacheKey);
+		if (!pricing) {
+			logger.debug(`[Pricing] No pricing found for ${provider}/${model}, trying partial match`);
+			for (const [key, value] of this.cache.entries()) if (key.startsWith(`${provider.toLowerCase()}:`)) {
+				const modelPart = key.split(":")[1];
+				if (model.toLowerCase().includes(modelPart)) {
+					logger.debug(`[Pricing] Found partial match: ${key}`);
+					return value;
+				}
+			}
+			return null;
+		}
+		return pricing;
+	}
+	/**
+	* Force refresh the pricing cache
+	*/
+	async refreshCache() {
+		this.lastFetch = 0;
+		await this.ensureFreshCache();
+	}
+	/**
+	* Check if the provider is ready
+	*/
+	isReady() {
+		return this.ready;
+	}
+	/**
+	* Get the number of cached models (for debugging)
+	*/
+	getCacheSize() {
+		return this.cache.size;
+	}
+};
+let defaultProvider = null;
+/**
+* Get the default pricing provider instance
+*/
+function getDefaultPricingProvider() {
+	if (!defaultProvider) defaultProvider = new ModelsDevPricingProvider();
+	return defaultProvider;
+}
+
+//#endregion
+export { CacheService, FileCacheBackend, MS, MemoryCacheBackend, ModelsDevPricingProvider, SCHEMA_METADATA, SupportedProviders, calculateCost, chatCompletionCreateParamsBaseSchema, configVariantsSchema, configsSchema, createDataLayer, createDatabase, createDatabaseFromConnection, createLLMRequestsDataLayer, detectDatabaseType, dollarsToMicroDollars, environmentSecretsSchema, environmentsSchema, formatCost, gateway, generateId, getDefaultPricingProvider, getMigrations, llmRequestsSchema, llmopsConfigSchema, logger, matchType, microDollarsToDollars, parsePartialTableData, parseTableData, runAutoMigrations, schemas, targetingRulesSchema, validateLLMOpsConfig, validatePartialTableData, validateTableData, variantJsonDataSchema, variantVersionsSchema, variantsSchema, workspaceSettingsSchema };

package/dist/{node-sqlite-dialect-CQlHW438.cjs → node-sqlite-dialect-b2V910TJ.cjs}

@@ -1,4 +1,4 @@
-const require_db = require('./db-C9-M-kdS.cjs');
+const require_db = require('./db-eEfIe5dO.cjs');
 let kysely = require("kysely");
 
 //#region src/db/node-sqlite-dialect.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@llmops/core",
-  "version": "0.1.5-beta.1",
+  "version": "0.1.5",
   "description": "Core LLMOps functionality and utilities",
   "type": "module",
   "license": "Apache-2.0",
@@ -51,7 +51,7 @@
     "hono": "^4.10.7",
     "kysely": "^0.28.8",
     "pino": "^10.1.0",
-    "@llmops/gateway": "^0.1.5-beta.1"
+    "@llmops/gateway": "^0.1.5"
   },
   "scripts": {
     "build": "tsdown",