@lingo.dev/compiler 0.1.0

package/build/translators/pluralization/service.cjs

@@ -0,0 +1,247 @@
+const require_rolldown_runtime = require('../../_virtual/rolldown_runtime.cjs');
+const require_parse_xml = require('../parse-xml.cjs');
+const require_model_factory = require('../lingo/model-factory.cjs');
+const require_timeout = require('../../utils/timeout.cjs');
+const require_pattern_detector = require('./pattern-detector.cjs');
+const require_prompt = require('./prompt.cjs');
+const require_shots = require('./shots.cjs');
+const require_icu_validator = require('./icu-validator.cjs');
+let ai = require("ai");
+
+//#region src/translators/pluralization/service.ts
+/**
+* Pluralization service with batching and model reuse
+*/
+var PluralizationService = class {
+	languageModel;
+	modelName;
+	cache = /* @__PURE__ */ new Map();
+	prompt;
+	sourceLocale;
+	constructor(config, logger) {
+		this.logger = logger;
+		const localeModel = require_model_factory.parseModelString(config.model);
+		if (!localeModel) throw new Error(`Invalid model format: "${config.model}"`);
+		const modelsConfig = { "*:*": config.model };
+		this.logger.info("Validating API keys for pluralization...");
+		const validatedKeys = require_model_factory.validateAndGetApiKeys(modelsConfig);
+		this.logger.info("✅ API keys validated for pluralization");
+		this.languageModel = require_model_factory.createAiModel(localeModel, validatedKeys);
+		this.modelName = `${localeModel.provider}:${localeModel.name}`;
+		this.sourceLocale = config.sourceLocale;
+		this.prompt = require_prompt.getSystemPrompt({ sourceLocale: config.sourceLocale });
+		this.logger.info(`Initialized pluralization service with ${this.modelName}`);
+	}
+	/**
+	* Generate ICU formats for multiple candidates in a single batch
+	*
+	* @param candidates Array of plural candidates
+	* @param batchSize Maximum candidates per batch (default: 10)
+	* @returns Map of hash -> ICU generation result
+	*/
+	async generateBatch(candidates, batchSize = 10) {
+		const results = /* @__PURE__ */ new Map();
+		const uncachedCandidates = candidates.filter((c) => {
+			const cached = this.cache.get(c.hash);
+			if (cached) {
+				results.set(c.hash, cached);
+				return false;
+			}
+			return true;
+		});
+		if (uncachedCandidates.length === 0) {
+			this.logger.debug(`All ${candidates.length} candidates found in cache, skipping LLM call`);
+			return results;
+		}
+		this.logger.info(`Processing ${uncachedCandidates.length} candidates (${candidates.length - uncachedCandidates.length} cached)`);
+		for (let i = 0; i < uncachedCandidates.length; i += batchSize) {
+			const batch = uncachedCandidates.slice(i, i + batchSize);
+			this.logger.info(`Processing batch ${Math.floor(i / batchSize) + 1}/${Math.ceil(uncachedCandidates.length / batchSize)} (${batch.length} candidates)`);
+			const batchResults = await this.processBatch(batch);
+			for (const [hash, result] of batchResults) {
+				results.set(hash, result);
+				this.cache.set(hash, result);
+			}
+		}
+		return results;
+	}
+	/**
+	* Process a single batch of candidates
+	*/
+	async processBatch(candidates) {
+		const results = /* @__PURE__ */ new Map();
+		try {
+			const batchRequest = {
+				version: .1,
+				sourceLocale: this.sourceLocale,
+				candidates: { candidate: candidates.map((c) => ({
+					hash: c.hash,
+					text: c.sourceText
+				})) }
+			};
+			const responseText = (await require_timeout.withTimeout((0, ai.generateText)({
+				model: this.languageModel,
+				messages: [
+					{
+						role: "system",
+						content: this.prompt
+					},
+					...require_shots.shots.flatMap((shotsTuple) => [{
+						role: "user",
+						content: require_parse_xml.obj2xml(shotsTuple[0])
+					}, {
+						role: "assistant",
+						content: require_parse_xml.obj2xml(shotsTuple[1])
+					}]),
+					{
+						role: "user",
+						content: require_parse_xml.obj2xml(batchRequest)
+					}
+				]
+			}), require_timeout.DEFAULT_TIMEOUTS.AI_API * 2, `Pluralization with ${this.modelName}`)).text.trim();
+			this.logger.debug(`LLM XML response: ${responseText.substring(0, 200)}...`);
+			const parsed = require_parse_xml.parseXmlFromResponseText(responseText);
+			const resultArray = Array.isArray(parsed.results.result) ? parsed.results.result : [parsed.results.result];
+			for (const result of resultArray) {
+				const candidate = candidates.find((c) => c.hash === result.hash);
+				if (!candidate) {
+					this.logger.warn(`No candidate found for hash: ${result.hash}`);
+					continue;
+				}
+				if (result.shouldPluralize && result.icuText) {
+					this.logger.debug(`✓ ICU format generated for "${candidate.sourceText}": "${result.icuText}"`);
+					results.set(result.hash, {
+						success: true,
+						icuText: result.icuText,
+						reasoning: result.reasoning
+					});
+				} else {
+					this.logger.debug(`✗ Pluralization not appropriate for "${candidate.sourceText}": ${result.reasoning}`);
+					results.set(result.hash, {
+						success: false,
+						reasoning: result.reasoning
+					});
+				}
+			}
+			for (const candidate of candidates) if (!results.has(candidate.hash)) {
+				this.logger.warn(`No result returned for candidate: ${candidate.sourceText}`);
+				results.set(candidate.hash, {
+					success: false,
+					error: "No result returned by LLM"
+				});
+			}
+		} catch (error) {
+			const errorMsg = error instanceof Error ? error.message : "Unknown error";
+			this.logger.error(`Failed to process batch: ${errorMsg}`);
+			for (const candidate of candidates) results.set(candidate.hash, {
+				success: false,
+				error: errorMsg
+			});
+		}
+		return results;
+	}
+	/**
+	* Process metadata entries for pluralization
+	*
+	* This is the main entry point that:
+	* 1. Detects plural candidates using pattern matching
+	* 2. Generates ICU format using LLM (batched)
+	* 3. Validates the ICU format
+	* 4. Updates metadata entries in-place (modifies sourceText)
+	* 5. Returns statistics
+	* @param metadata Metadata schema with translation entries
+	
+	* @returns Statistics about the pluralization process
+	*/
+	async process(metadata) {
+		const startTime = performance.now();
+		const totalEntries = Object.keys(metadata.entries).length;
+		if (totalEntries === 0) return {
+			total: 0,
+			candidates: 0,
+			pluralized: 0,
+			rejected: 0,
+			failed: 0,
+			durationMs: 0
+		};
+		this.logger.info(`Starting pluralization processing for ${totalEntries} entries`);
+		const candidates = require_pattern_detector.detectPluralCandidates(Object.fromEntries(Object.entries(metadata.entries).map(([hash, entry]) => [hash, entry.sourceText])), this.logger);
+		this.logger.info(`Found ${candidates.length} plural candidates (${(candidates.length / totalEntries * 100).toFixed(1)}%)`);
+		if (candidates.length === 0) return {
+			total: totalEntries,
+			candidates: 0,
+			pluralized: 0,
+			rejected: 0,
+			failed: 0,
+			durationMs: performance.now() - startTime
+		};
+		this.logger.debug("Generating ICU formats with batching...");
+		const icuResults = await this.generateBatch(candidates, 10);
+		this.logger.debug("Validating and updating entries...");
+		let pluralized = 0;
+		let rejected = 0;
+		let failed = 0;
+		for (const candidate of candidates) {
+			const result = icuResults.get(candidate.hash);
+			const entry = metadata.entries[candidate.hash];
+			this.logger.debug(`Processing candidate: ${candidate.sourceText}`);
+			if (!entry) {
+				this.logger.warn(`Entry not found for hash: ${candidate.hash}`);
+				failed++;
+				continue;
+			}
+			if (!result) {
+				this.logger.warn(`No result for hash: ${candidate.hash}`);
+				failed++;
+				continue;
+			}
+			if (result.error) {
+				this.logger.warn(`Error generating ICU for "${candidate.sourceText}": ${result.error}`);
+				failed++;
+				continue;
+			}
+			if (!result.success || !result.icuText) {
+				this.logger.debug(`Rejected pluralization for "${candidate.sourceText}": ${result.reasoning}`);
+				rejected++;
+				continue;
+			}
+			if (!require_icu_validator.validateICU(result.icuText, candidate.sourceText, this.logger)) {
+				this.logger.warn(`Invalid ICU format generated for "${candidate.sourceText}", falling back to original`);
+				failed++;
+				continue;
+			}
+			this.logger.info(`Pluralizing: "${entry.sourceText}" -> "${result.icuText}"`);
+			entry.sourceText = result.icuText;
+			pluralized++;
+		}
+		const duration = performance.now() - startTime;
+		this.logger.info(`Pluralization completed: ${pluralized} pluralized, ${rejected} rejected, ${failed} failed in ${duration.toFixed(0)}ms`);
+		return {
+			total: totalEntries,
+			candidates: candidates.length,
+			pluralized,
+			rejected,
+			failed,
+			durationMs: duration
+		};
+	}
+	/**
+	* Clear the cache
+	*/
+	clearCache() {
+		this.cache.clear();
+		this.logger.debug("Pluralization cache cleared");
+	}
+	/**
+	* Get cache statistics
+	*/
+	getCacheStats() {
+		return {
+			size: this.cache.size,
+			hits: 0
+		};
+	}
+};
+
+//#endregion
+exports.PluralizationService = PluralizationService;

package/build/translators/pluralization/service.mjs

@@ -0,0 +1,247 @@
+import { obj2xml, parseXmlFromResponseText } from "../parse-xml.mjs";
+import { createAiModel, parseModelString, validateAndGetApiKeys } from "../lingo/model-factory.mjs";
+import { DEFAULT_TIMEOUTS, withTimeout } from "../../utils/timeout.mjs";
+import { detectPluralCandidates } from "./pattern-detector.mjs";
+import { getSystemPrompt } from "./prompt.mjs";
+import { shots } from "./shots.mjs";
+import { validateICU } from "./icu-validator.mjs";
+import { generateText } from "ai";
+
+//#region src/translators/pluralization/service.ts
+/**
+* Pluralization service with batching and model reuse
+*/
+var PluralizationService = class {
+	languageModel;
+	modelName;
+	cache = /* @__PURE__ */ new Map();
+	prompt;
+	sourceLocale;
+	constructor(config, logger) {
+		this.logger = logger;
+		const localeModel = parseModelString(config.model);
+		if (!localeModel) throw new Error(`Invalid model format: "${config.model}"`);
+		const modelsConfig = { "*:*": config.model };
+		this.logger.info("Validating API keys for pluralization...");
+		const validatedKeys = validateAndGetApiKeys(modelsConfig);
+		this.logger.info("✅ API keys validated for pluralization");
+		this.languageModel = createAiModel(localeModel, validatedKeys);
+		this.modelName = `${localeModel.provider}:${localeModel.name}`;
+		this.sourceLocale = config.sourceLocale;
+		this.prompt = getSystemPrompt({ sourceLocale: config.sourceLocale });
+		this.logger.info(`Initialized pluralization service with ${this.modelName}`);
+	}
+	/**
+	* Generate ICU formats for multiple candidates in a single batch
+	*
+	* @param candidates Array of plural candidates
+	* @param batchSize Maximum candidates per batch (default: 10)
+	* @returns Map of hash -> ICU generation result
+	*/
+	async generateBatch(candidates, batchSize = 10) {
+		const results = /* @__PURE__ */ new Map();
+		const uncachedCandidates = candidates.filter((c) => {
+			const cached = this.cache.get(c.hash);
+			if (cached) {
+				results.set(c.hash, cached);
+				return false;
+			}
+			return true;
+		});
+		if (uncachedCandidates.length === 0) {
+			this.logger.debug(`All ${candidates.length} candidates found in cache, skipping LLM call`);
+			return results;
+		}
+		this.logger.info(`Processing ${uncachedCandidates.length} candidates (${candidates.length - uncachedCandidates.length} cached)`);
+		for (let i = 0; i < uncachedCandidates.length; i += batchSize) {
+			const batch = uncachedCandidates.slice(i, i + batchSize);
+			this.logger.info(`Processing batch ${Math.floor(i / batchSize) + 1}/${Math.ceil(uncachedCandidates.length / batchSize)} (${batch.length} candidates)`);
+			const batchResults = await this.processBatch(batch);
+			for (const [hash, result] of batchResults) {
+				results.set(hash, result);
+				this.cache.set(hash, result);
+			}
+		}
+		return results;
+	}
+	/**
+	* Process a single batch of candidates
+	*/
+	async processBatch(candidates) {
+		const results = /* @__PURE__ */ new Map();
+		try {
+			const batchRequest = {
+				version: .1,
+				sourceLocale: this.sourceLocale,
+				candidates: { candidate: candidates.map((c) => ({
+					hash: c.hash,
+					text: c.sourceText
+				})) }
+			};
+			const responseText = (await withTimeout(generateText({
+				model: this.languageModel,
+				messages: [
+					{
+						role: "system",
+						content: this.prompt
+					},
+					...shots.flatMap((shotsTuple) => [{
+						role: "user",
+						content: obj2xml(shotsTuple[0])
+					}, {
+						role: "assistant",
+						content: obj2xml(shotsTuple[1])
+					}]),
+					{
+						role: "user",
+						content: obj2xml(batchRequest)
+					}
+				]
+			}), DEFAULT_TIMEOUTS.AI_API * 2, `Pluralization with ${this.modelName}`)).text.trim();
+			this.logger.debug(`LLM XML response: ${responseText.substring(0, 200)}...`);
+			const parsed = parseXmlFromResponseText(responseText);
+			const resultArray = Array.isArray(parsed.results.result) ? parsed.results.result : [parsed.results.result];
+			for (const result of resultArray) {
+				const candidate = candidates.find((c) => c.hash === result.hash);
+				if (!candidate) {
+					this.logger.warn(`No candidate found for hash: ${result.hash}`);
+					continue;
+				}
+				if (result.shouldPluralize && result.icuText) {
+					this.logger.debug(`✓ ICU format generated for "${candidate.sourceText}": "${result.icuText}"`);
+					results.set(result.hash, {
+						success: true,
+						icuText: result.icuText,
+						reasoning: result.reasoning
+					});
+				} else {
+					this.logger.debug(`✗ Pluralization not appropriate for "${candidate.sourceText}": ${result.reasoning}`);
+					results.set(result.hash, {
+						success: false,
+						reasoning: result.reasoning
+					});
+				}
+			}
+			for (const candidate of candidates) if (!results.has(candidate.hash)) {
+				this.logger.warn(`No result returned for candidate: ${candidate.sourceText}`);
+				results.set(candidate.hash, {
+					success: false,
+					error: "No result returned by LLM"
+				});
+			}
+		} catch (error) {
+			const errorMsg = error instanceof Error ? error.message : "Unknown error";
+			this.logger.error(`Failed to process batch: ${errorMsg}`);
+			for (const candidate of candidates) results.set(candidate.hash, {
+				success: false,
+				error: errorMsg
+			});
+		}
+		return results;
+	}
+	/**
+	* Process metadata entries for pluralization
+	*
+	* This is the main entry point that:
+	* 1. Detects plural candidates using pattern matching
+	* 2. Generates ICU format using LLM (batched)
+	* 3. Validates the ICU format
+	* 4. Updates metadata entries in-place (modifies sourceText)
+	* 5. Returns statistics
+	* @param metadata Metadata schema with translation entries
+	
+	* @returns Statistics about the pluralization process
+	*/
+	async process(metadata) {
+		const startTime = performance.now();
+		const totalEntries = Object.keys(metadata.entries).length;
+		if (totalEntries === 0) return {
+			total: 0,
+			candidates: 0,
+			pluralized: 0,
+			rejected: 0,
+			failed: 0,
+			durationMs: 0
+		};
+		this.logger.info(`Starting pluralization processing for ${totalEntries} entries`);
+		const candidates = detectPluralCandidates(Object.fromEntries(Object.entries(metadata.entries).map(([hash, entry]) => [hash, entry.sourceText])), this.logger);
+		this.logger.info(`Found ${candidates.length} plural candidates (${(candidates.length / totalEntries * 100).toFixed(1)}%)`);
+		if (candidates.length === 0) return {
+			total: totalEntries,
+			candidates: 0,
+			pluralized: 0,
+			rejected: 0,
+			failed: 0,
+			durationMs: performance.now() - startTime
+		};
+		this.logger.debug("Generating ICU formats with batching...");
+		const icuResults = await this.generateBatch(candidates, 10);
+		this.logger.debug("Validating and updating entries...");
+		let pluralized = 0;
+		let rejected = 0;
+		let failed = 0;
+		for (const candidate of candidates) {
+			const result = icuResults.get(candidate.hash);
+			const entry = metadata.entries[candidate.hash];
+			this.logger.debug(`Processing candidate: ${candidate.sourceText}`);
+			if (!entry) {
+				this.logger.warn(`Entry not found for hash: ${candidate.hash}`);
+				failed++;
+				continue;
+			}
+			if (!result) {
+				this.logger.warn(`No result for hash: ${candidate.hash}`);
+				failed++;
+				continue;
+			}
+			if (result.error) {
+				this.logger.warn(`Error generating ICU for "${candidate.sourceText}": ${result.error}`);
+				failed++;
+				continue;
+			}
+			if (!result.success || !result.icuText) {
+				this.logger.debug(`Rejected pluralization for "${candidate.sourceText}": ${result.reasoning}`);
+				rejected++;
+				continue;
+			}
+			if (!validateICU(result.icuText, candidate.sourceText, this.logger)) {
+				this.logger.warn(`Invalid ICU format generated for "${candidate.sourceText}", falling back to original`);
+				failed++;
+				continue;
+			}
+			this.logger.info(`Pluralizing: "${entry.sourceText}" -> "${result.icuText}"`);
+			entry.sourceText = result.icuText;
+			pluralized++;
+		}
+		const duration = performance.now() - startTime;
+		this.logger.info(`Pluralization completed: ${pluralized} pluralized, ${rejected} rejected, ${failed} failed in ${duration.toFixed(0)}ms`);
+		return {
+			total: totalEntries,
+			candidates: candidates.length,
+			pluralized,
+			rejected,
+			failed,
+			durationMs: duration
+		};
+	}
+	/**
+	* Clear the cache
+	*/
+	clearCache() {
+		this.cache.clear();
+		this.logger.debug("Pluralization cache cleared");
+	}
+	/**
+	* Get cache statistics
+	*/
+	getCacheStats() {
+		return {
+			size: this.cache.size,
+			hits: 0
+		};
+	}
+};
+
+//#endregion
+export { PluralizationService };
+//# sourceMappingURL=service.mjs.map

package/build/translators/pluralization/service.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"service.mjs","names":["logger: Logger","modelsConfig: Record<string, string>","batchRequest: PluralizationBatch"],"sources":["../../../src/translators/pluralization/service.ts"],"sourcesContent":["/**\n * Pluralization service with batching and caching\n */\n\nimport type { LanguageModel } from \"ai\";\nimport { generateText } from \"ai\";\nimport type {\n  ICUGenerationResult,\n  PluralCandidate,\n  PluralizationBatch,\n  PluralizationConfig,\n  PluralizationResponse,\n  PluralizationStats,\n} from \"./types\";\nimport {\n  createAiModel,\n  parseModelString,\n  validateAndGetApiKeys,\n} from \"../lingo/model-factory\";\nimport { Logger } from \"../../utils/logger\";\nimport { DEFAULT_TIMEOUTS, withTimeout } from \"../../utils/timeout\";\nimport { getSystemPrompt } from \"./prompt\";\nimport { obj2xml, parseXmlFromResponseText } from \"../parse-xml\";\nimport { shots } from \"./shots\";\nimport type { MetadataSchema } from \"../../types\";\nimport { detectPluralCandidates } from \"./pattern-detector\";\nimport { validateICU } from \"./icu-validator\";\n\n/**\n * Pluralization service with batching and model reuse\n */\nexport class PluralizationService {\n  private readonly languageModel: LanguageModel;\n  private readonly modelName: string;\n  private cache = new Map<string, ICUGenerationResult>();\n  private readonly prompt: string;\n  private readonly sourceLocale: string;\n\n  constructor(\n    config: PluralizationConfig,\n    private logger: Logger,\n  ) {\n    const localeModel = parseModelString(config.model);\n    if (!localeModel) {\n      throw new Error(`Invalid model format: \"${config.model}\"`);\n    }\n\n    // Validate and fetch API keys for the pluralization provider\n    // We need to create a models config that validateAndFetchApiKeys can use\n    const modelsConfig: Record<string, string> = {\n      \"*:*\": config.model, // Single model for pluralization\n    };\n\n    this.logger.info(\"Validating API keys for pluralization...\");\n    const validatedKeys = validateAndGetApiKeys(modelsConfig);\n    this.logger.info(\"✅ API keys validated for pluralization\");\n\n    this.languageModel = createAiModel(localeModel, validatedKeys);\n    this.modelName = `${localeModel.provider}:${localeModel.name}`;\n    this.sourceLocale = config.sourceLocale;\n    this.prompt = getSystemPrompt({ sourceLocale: config.sourceLocale });\n\n    this.logger.info(\n      `Initialized pluralization service with ${this.modelName}`,\n    );\n  }\n\n  /**\n   * Generate ICU formats for multiple candidates in a single batch\n   *\n   * @param candidates Array of plural candidates\n   * @param batchSize Maximum candidates per batch (default: 10)\n   * @returns Map of hash -> ICU generation result\n   */\n  async generateBatch(\n    candidates: PluralCandidate[],\n    batchSize: number = 10,\n  ): Promise<Map<string, ICUGenerationResult>> {\n    const results = new Map<string, ICUGenerationResult>();\n\n    // Check cache first\n    const uncachedCandidates = candidates.filter((c) => {\n      const cached = this.cache.get(c.hash);\n      if (cached) {\n        results.set(c.hash, cached);\n        return false;\n      }\n      return true;\n    });\n\n    if (uncachedCandidates.length === 0) {\n      this.logger.debug(\n        `All ${candidates.length} candidates found in cache, skipping LLM call`,\n      );\n      return results;\n    }\n\n    this.logger.info(\n      `Processing ${uncachedCandidates.length} candidates (${candidates.length - uncachedCandidates.length} cached)`,\n    );\n\n    // Process in batches\n    for (let i = 0; i < uncachedCandidates.length; i += batchSize) {\n      const batch = uncachedCandidates.slice(i, i + batchSize);\n\n      this.logger.info(\n        `Processing batch ${Math.floor(i / batchSize) + 1}/${Math.ceil(uncachedCandidates.length / batchSize)} (${batch.length} candidates)`,\n      );\n\n      const batchResults = await this.processBatch(batch);\n\n      // Store results and cache them\n      for (const [hash, result] of batchResults) {\n        results.set(hash, result);\n        this.cache.set(hash, result);\n      }\n    }\n\n    return results;\n  }\n\n  /**\n   * Process a single batch of candidates\n   */\n  private async processBatch(\n    candidates: PluralCandidate[],\n  ): Promise<Map<string, ICUGenerationResult>> {\n    const results = new Map<string, ICUGenerationResult>();\n\n    try {\n      // Prepare batch request in XML format\n      const batchRequest: PluralizationBatch = {\n        version: 0.1,\n        sourceLocale: this.sourceLocale,\n        candidates: {\n          candidate: candidates.map((c) => ({\n            hash: c.hash,\n            text: c.sourceText,\n          })),\n        },\n      };\n\n      // Call LLM with XML format and few-shot examples\n      const response = await withTimeout(\n        generateText({\n          model: this.languageModel,\n          messages: [\n            {\n              role: \"system\",\n              content: this.prompt,\n            },\n            // Add few-shot examples\n            ...shots.flatMap((shotsTuple) => [\n              {\n                role: \"user\" as const,\n                content: obj2xml(shotsTuple[0]),\n              },\n              {\n                role: \"assistant\" as const,\n                content: obj2xml(shotsTuple[1]),\n              },\n            ]),\n            {\n              role: \"user\",\n              content: obj2xml(batchRequest),\n            },\n          ],\n        }),\n        DEFAULT_TIMEOUTS.AI_API * 2, // Double timeout for batch\n        `Pluralization with ${this.modelName}`,\n      );\n\n      const responseText = response.text.trim();\n      this.logger.debug(\n        `LLM XML response: ${responseText.substring(0, 200)}...`,\n      );\n      // Parse XML response\n      const parsed =\n        parseXmlFromResponseText<PluralizationResponse>(responseText);\n\n      // Process results\n      const resultArray = Array.isArray(parsed.results.result)\n        ? parsed.results.result\n        : [parsed.results.result];\n\n      for (const result of resultArray) {\n        const candidate = candidates.find((c) => c.hash === result.hash);\n        if (!candidate) {\n          this.logger.warn(`No candidate found for hash: ${result.hash}`);\n          continue;\n        }\n\n        if (result.shouldPluralize && result.icuText) {\n          this.logger.debug(\n            `✓ ICU format generated for \"${candidate.sourceText}\": \"${result.icuText}\"`,\n          );\n          results.set(result.hash, {\n            success: true,\n            icuText: result.icuText,\n            reasoning: result.reasoning,\n          });\n        } else {\n          this.logger.debug(\n            `✗ Pluralization not appropriate for \"${candidate.sourceText}\": ${result.reasoning}`,\n          );\n          results.set(result.hash, {\n            success: false,\n            reasoning: result.reasoning,\n          });\n        }\n      }\n\n      // Handle missing results (LLM didn't return result for some candidates)\n      for (const candidate of candidates) {\n        if (!results.has(candidate.hash)) {\n          this.logger.warn(\n            `No result returned for candidate: ${candidate.sourceText}`,\n          );\n          results.set(candidate.hash, {\n            success: false,\n            error: \"No result returned by LLM\",\n          });\n        }\n      }\n    } catch (error) {\n      const errorMsg = error instanceof Error ? error.message : \"Unknown error\";\n      this.logger.error(`Failed to process batch: ${errorMsg}`);\n\n      // Mark all candidates as failed\n      for (const candidate of candidates) {\n        results.set(candidate.hash, {\n          success: false,\n          error: errorMsg,\n        });\n      }\n    }\n\n    return results;\n  }\n\n  /**\n   * Process metadata entries for pluralization\n   *\n   * This is the main entry point that:\n   * 1. Detects plural candidates using pattern matching\n   * 2. Generates ICU format using LLM (batched)\n   * 3. Validates the ICU format\n   * 4. Updates metadata entries in-place (modifies sourceText)\n   * 5. Returns statistics\n   * @param metadata Metadata schema with translation entries\n\n   * @returns Statistics about the pluralization process\n   */\n  async process(metadata: MetadataSchema): Promise<PluralizationStats> {\n    const startTime = performance.now();\n    const totalEntries = Object.keys(metadata.entries).length;\n\n    if (totalEntries === 0) {\n      return {\n        total: 0,\n        candidates: 0,\n        pluralized: 0,\n        rejected: 0,\n        failed: 0,\n        durationMs: 0,\n      };\n    }\n\n    this.logger.info(\n      `Starting pluralization processing for ${totalEntries} entries`,\n    );\n\n    // Step 1: Detect plural candidates using pattern matching\n    const entriesMap: Record<string, string> = Object.fromEntries(\n      Object.entries(metadata.entries).map(([hash, entry]) => [\n        hash,\n        entry.sourceText,\n      ]),\n    );\n\n    const candidates = detectPluralCandidates(entriesMap, this.logger);\n\n    this.logger.info(\n      `Found ${candidates.length} plural candidates (${((candidates.length / totalEntries) * 100).toFixed(1)}%)`,\n    );\n\n    if (candidates.length === 0) {\n      const endTime = performance.now();\n      return {\n        total: totalEntries,\n        candidates: 0,\n        pluralized: 0,\n        rejected: 0,\n        failed: 0,\n        durationMs: endTime - startTime,\n      };\n    }\n\n    // Step 2: Generate ICU formats with batching\n    this.logger.debug(\"Generating ICU formats with batching...\");\n    const icuResults = await this.generateBatch(candidates, 10);\n\n    // Step 3: Validate and update metadata entries\n    this.logger.debug(\"Validating and updating entries...\");\n    let pluralized = 0;\n    let rejected = 0;\n    let failed = 0;\n\n    for (const candidate of candidates) {\n      const result = icuResults.get(candidate.hash);\n      const entry = metadata.entries[candidate.hash];\n      this.logger.debug(`Processing candidate: ${candidate.sourceText}`);\n      if (!entry) {\n        this.logger.warn(`Entry not found for hash: ${candidate.hash}`);\n        failed++;\n        continue;\n      }\n\n      if (!result) {\n        this.logger.warn(`No result for hash: ${candidate.hash}`);\n        failed++;\n        continue;\n      }\n\n      if (result.error) {\n        this.logger.warn(\n          `Error generating ICU for \"${candidate.sourceText}\": ${result.error}`,\n        );\n        failed++;\n        continue;\n      }\n\n      if (!result.success || !result.icuText) {\n        this.logger.debug(\n          `Rejected pluralization for \"${candidate.sourceText}\": ${result.reasoning}`,\n        );\n        rejected++;\n        continue;\n      }\n\n      const isValid = validateICU(\n        result.icuText,\n        candidate.sourceText,\n        this.logger,\n      );\n\n      if (!isValid) {\n        this.logger.warn(\n          `Invalid ICU format generated for \"${candidate.sourceText}\", falling back to original`,\n        );\n        failed++;\n        continue;\n      }\n\n      // Update metadata entry in-place\n      this.logger.info(\n        `Pluralizing: \"${entry.sourceText}\" -> \"${result.icuText}\"`,\n      );\n      entry.sourceText = result.icuText;\n      pluralized++;\n    }\n\n    const endTime = performance.now();\n    const duration = endTime - startTime;\n\n    this.logger.info(\n      `Pluralization completed: ${pluralized} pluralized, ${rejected} rejected, ${failed} failed in ${duration.toFixed(0)}ms`,\n    );\n\n    return {\n      total: totalEntries,\n      candidates: candidates.length,\n      pluralized,\n      rejected,\n      failed,\n      durationMs: duration,\n    };\n  }\n\n  /**\n   * Clear the cache\n   */\n  clearCache(): void {\n    this.cache.clear();\n    this.logger.debug(\"Pluralization cache cleared\");\n  }\n\n  /**\n   * Get cache statistics\n   */\n  getCacheStats(): { size: number; hits: number } {\n    return {\n      size: this.cache.size,\n      hits: 0, // We don't track hits currently\n    };\n  }\n}\n"],"mappings":";;;;;;;;;;;;;AA+BA,IAAa,uBAAb,MAAkC;CAChC,AAAiB;CACjB,AAAiB;CACjB,AAAQ,wBAAQ,IAAI,KAAkC;CACtD,AAAiB;CACjB,AAAiB;CAEjB,YACE,QACA,AAAQA,QACR;EADQ;EAER,MAAM,cAAc,iBAAiB,OAAO,MAAM;AAClD,MAAI,CAAC,YACH,OAAM,IAAI,MAAM,0BAA0B,OAAO,MAAM,GAAG;EAK5D,MAAMC,eAAuC,EAC3C,OAAO,OAAO,OACf;AAED,OAAK,OAAO,KAAK,2CAA2C;EAC5D,MAAM,gBAAgB,sBAAsB,aAAa;AACzD,OAAK,OAAO,KAAK,yCAAyC;AAE1D,OAAK,gBAAgB,cAAc,aAAa,cAAc;AAC9D,OAAK,YAAY,GAAG,YAAY,SAAS,GAAG,YAAY;AACxD,OAAK,eAAe,OAAO;AAC3B,OAAK,SAAS,gBAAgB,EAAE,cAAc,OAAO,cAAc,CAAC;AAEpE,OAAK,OAAO,KACV,0CAA0C,KAAK,YAChD;;;;;;;;;CAUH,MAAM,cACJ,YACA,YAAoB,IACuB;EAC3C,MAAM,0BAAU,IAAI,KAAkC;EAGtD,MAAM,qBAAqB,WAAW,QAAQ,MAAM;GAClD,MAAM,SAAS,KAAK,MAAM,IAAI,EAAE,KAAK;AACrC,OAAI,QAAQ;AACV,YAAQ,IAAI,EAAE,MAAM,OAAO;AAC3B,WAAO;;AAET,UAAO;IACP;AAEF,MAAI,mBAAmB,WAAW,GAAG;AACnC,QAAK,OAAO,MACV,OAAO,WAAW,OAAO,+CAC1B;AACD,UAAO;;AAGT,OAAK,OAAO,KACV,cAAc,mBAAmB,OAAO,eAAe,WAAW,SAAS,mBAAmB,OAAO,UACtG;AAGD,OAAK,IAAI,IAAI,GAAG,IAAI,mBAAmB,QAAQ,KAAK,WAAW;GAC7D,MAAM,QAAQ,mBAAmB,MAAM,GAAG,IAAI,UAAU;AAExD,QAAK,OAAO,KACV,oBAAoB,KAAK,MAAM,IAAI,UAAU,GAAG,EAAE,GAAG,KAAK,KAAK,mBAAmB,SAAS,UAAU,CAAC,IAAI,MAAM,OAAO,cACxH;GAED,MAAM,eAAe,MAAM,KAAK,aAAa,MAAM;AAGnD,QAAK,MAAM,CAAC,MAAM,WAAW,cAAc;AACzC,YAAQ,IAAI,MAAM,OAAO;AACzB,SAAK,MAAM,IAAI,MAAM,OAAO;;;AAIhC,SAAO;;;;;CAMT,MAAc,aACZ,YAC2C;EAC3C,MAAM,0BAAU,IAAI,KAAkC;AAEtD,MAAI;GAEF,MAAMC,eAAmC;IACvC,SAAS;IACT,cAAc,KAAK;IACnB,YAAY,EACV,WAAW,WAAW,KAAK,OAAO;KAChC,MAAM,EAAE;KACR,MAAM,EAAE;KACT,EAAE,EACJ;IACF;GAgCD,MAAM,gBA7BW,MAAM,YACrB,aAAa;IACX,OAAO,KAAK;IACZ,UAAU;KACR;MACE,MAAM;MACN,SAAS,KAAK;MACf;KAED,GAAG,MAAM,SAAS,eAAe,CAC/B;MACE,MAAM;MACN,SAAS,QAAQ,WAAW,GAAG;MAChC,EACD;MACE,MAAM;MACN,SAAS,QAAQ,WAAW,GAAG;MAChC,CACF,CAAC;KACF;MACE,MAAM;MACN,SAAS,QAAQ,aAAa;MAC/B;KACF;IACF,CAAC,EACF,iBAAiB,SAAS,GAC1B,sBAAsB,KAAK,YAC5B,EAE6B,KAAK,MAAM;AACzC,QAAK,OAAO,MACV,qBAAqB,aAAa,UAAU,GAAG,IAAI,CAAC,KACrD;GAED,MAAM,SACJ,yBAAgD,aAAa;GAG/D,MAAM,cAAc,MAAM,QAAQ,OAAO,QAAQ,OAAO,GACpD,OAAO,QAAQ,SACf,CAAC,OAAO,QAAQ,OAAO;AAE3B,QAAK,MAAM,UAAU,aAAa;IAChC,MAAM,YAAY,WAAW,MAAM,MAAM,EAAE,SAAS,OAAO,KAAK;AAChE,QAAI,CAAC,WAAW;AACd,UAAK,OAAO,KAAK,gCAAgC,OAAO,OAAO;AAC/D;;AAGF,QAAI,OAAO,mBAAmB,OAAO,SAAS;AAC5C,UAAK,OAAO,MACV,+BAA+B,UAAU,WAAW,MAAM,OAAO,QAAQ,GAC1E;AACD,aAAQ,IAAI,OAAO,MAAM;MACvB,SAAS;MACT,SAAS,OAAO;MAChB,WAAW,OAAO;MACnB,CAAC;WACG;AACL,UAAK,OAAO,MACV,wCAAwC,UAAU,WAAW,KAAK,OAAO,YAC1E;AACD,aAAQ,IAAI,OAAO,MAAM;MACvB,SAAS;MACT,WAAW,OAAO;MACnB,CAAC;;;AAKN,QAAK,MAAM,aAAa,WACtB,KAAI,CAAC,QAAQ,IAAI,UAAU,KAAK,EAAE;AAChC,SAAK,OAAO,KACV,qCAAqC,UAAU,aAChD;AACD,YAAQ,IAAI,UAAU,MAAM;KAC1B,SAAS;KACT,OAAO;KACR,CAAC;;WAGC,OAAO;GACd,MAAM,WAAW,iBAAiB,QAAQ,MAAM,UAAU;AAC1D,QAAK,OAAO,MAAM,4BAA4B,WAAW;AAGzD,QAAK,MAAM,aAAa,WACtB,SAAQ,IAAI,UAAU,MAAM;IAC1B,SAAS;IACT,OAAO;IACR,CAAC;;AAIN,SAAO;;;;;;;;;;;;;;;CAgBT,MAAM,QAAQ,UAAuD;EACnE,MAAM,YAAY,YAAY,KAAK;EACnC,MAAM,eAAe,OAAO,KAAK,SAAS,QAAQ,CAAC;AAEnD,MAAI,iBAAiB,EACnB,QAAO;GACL,OAAO;GACP,YAAY;GACZ,YAAY;GACZ,UAAU;GACV,QAAQ;GACR,YAAY;GACb;AAGH,OAAK,OAAO,KACV,yCAAyC,aAAa,UACvD;EAUD,MAAM,aAAa,uBAPwB,OAAO,YAChD,OAAO,QAAQ,SAAS,QAAQ,CAAC,KAAK,CAAC,MAAM,WAAW,CACtD,MACA,MAAM,WACP,CAAC,CACH,EAEqD,KAAK,OAAO;AAElE,OAAK,OAAO,KACV,SAAS,WAAW,OAAO,uBAAwB,WAAW,SAAS,eAAgB,KAAK,QAAQ,EAAE,CAAC,IACxG;AAED,MAAI,WAAW,WAAW,EAExB,QAAO;GACL,OAAO;GACP,YAAY;GACZ,YAAY;GACZ,UAAU;GACV,QAAQ;GACR,YAPc,YAAY,KAAK,GAOT;GACvB;AAIH,OAAK,OAAO,MAAM,0CAA0C;EAC5D,MAAM,aAAa,MAAM,KAAK,cAAc,YAAY,GAAG;AAG3D,OAAK,OAAO,MAAM,qCAAqC;EACvD,IAAI,aAAa;EACjB,IAAI,WAAW;EACf,IAAI,SAAS;AAEb,OAAK,MAAM,aAAa,YAAY;GAClC,MAAM,SAAS,WAAW,IAAI,UAAU,KAAK;GAC7C,MAAM,QAAQ,SAAS,QAAQ,UAAU;AACzC,QAAK,OAAO,MAAM,yBAAyB,UAAU,aAAa;AAClE,OAAI,CAAC,OAAO;AACV,SAAK,OAAO,KAAK,6BAA6B,UAAU,OAAO;AAC/D;AACA;;AAGF,OAAI,CAAC,QAAQ;AACX,SAAK,OAAO,KAAK,uBAAuB,UAAU,OAAO;AACzD;AACA;;AAGF,OAAI,OAAO,OAAO;AAChB,SAAK,OAAO,KACV,6BAA6B,UAAU,WAAW,KAAK,OAAO,QAC/D;AACD;AACA;;AAGF,OAAI,CAAC,OAAO,WAAW,CAAC,OAAO,SAAS;AACtC,SAAK,OAAO,MACV,+BAA+B,UAAU,WAAW,KAAK,OAAO,YACjE;AACD;AACA;;AASF,OAAI,CANY,YACd,OAAO,SACP,UAAU,YACV,KAAK,OACN,EAEa;AACZ,SAAK,OAAO,KACV,qCAAqC,UAAU,WAAW,6BAC3D;AACD;AACA;;AAIF,QAAK,OAAO,KACV,iBAAiB,MAAM,WAAW,QAAQ,OAAO,QAAQ,GAC1D;AACD,SAAM,aAAa,OAAO;AAC1B;;EAIF,MAAM,WADU,YAAY,KAAK,GACN;AAE3B,OAAK,OAAO,KACV,4BAA4B,WAAW,eAAe,SAAS,aAAa,OAAO,aAAa,SAAS,QAAQ,EAAE,CAAC,IACrH;AAED,SAAO;GACL,OAAO;GACP,YAAY,WAAW;GACvB;GACA;GACA;GACA,YAAY;GACb;;;;;CAMH,aAAmB;AACjB,OAAK,MAAM,OAAO;AAClB,OAAK,OAAO,MAAM,8BAA8B;;;;;CAMlD,gBAAgD;AAC9C,SAAO;GACL,MAAM,KAAK,MAAM;GACjB,MAAM;GACP"}

package/build/translators/pluralization/shots.cjs

@@ -0,0 +1,53 @@
+
+//#region src/translators/pluralization/shots.ts
+/**
+* Few-shot examples: [input, output]
+*/
+const shots = [[{
+	version: .1,
+	sourceLocale: "en",
+	candidates: { candidate: [{
+		hash: "h1",
+		text: "You have {count} items in your cart"
+	}, {
+		hash: "h2",
+		text: "{fileCount} files uploaded"
+	}] }
+}, {
+	version: .1,
+	results: { result: [{
+		hash: "h1",
+		shouldPluralize: true,
+		icuText: "You have {count, plural, =0 {no items} one {# item} other {# items}} in your cart",
+		reasoning: "Message varies based on item count"
+	}, {
+		hash: "h2",
+		shouldPluralize: true,
+		icuText: "{fileCount, plural, =0 {No files uploaded} one {# file uploaded} other {# files uploaded}}",
+		reasoning: "Message varies based on file count"
+	}] }
+}], [{
+	version: .1,
+	sourceLocale: "en",
+	candidates: { candidate: [{
+		hash: "h3",
+		text: "Welcome back, {name}"
+	}, {
+		hash: "h4",
+		text: "Your email is {email}"
+	}] }
+}, {
+	version: .1,
+	results: { result: [{
+		hash: "h3",
+		shouldPluralize: false,
+		reasoning: "Variable {name} is a person's name, not a count. No pluralization needed."
+	}, {
+		hash: "h4",
+		shouldPluralize: false,
+		reasoning: "Variable {email} is an email address, not a count. No pluralization needed."
+	}] }
+}]];
+
+//#endregion
+exports.shots = shots;

package/build/translators/pluralization/shots.mjs

@@ -0,0 +1,53 @@
+//#region src/translators/pluralization/shots.ts
+/**
+* Few-shot examples: [input, output]
+*/
+const shots = [[{
+	version: .1,
+	sourceLocale: "en",
+	candidates: { candidate: [{
+		hash: "h1",
+		text: "You have {count} items in your cart"
+	}, {
+		hash: "h2",
+		text: "{fileCount} files uploaded"
+	}] }
+}, {
+	version: .1,
+	results: { result: [{
+		hash: "h1",
+		shouldPluralize: true,
+		icuText: "You have {count, plural, =0 {no items} one {# item} other {# items}} in your cart",
+		reasoning: "Message varies based on item count"
+	}, {
+		hash: "h2",
+		shouldPluralize: true,
+		icuText: "{fileCount, plural, =0 {No files uploaded} one {# file uploaded} other {# files uploaded}}",
+		reasoning: "Message varies based on file count"
+	}] }
+}], [{
+	version: .1,
+	sourceLocale: "en",
+	candidates: { candidate: [{
+		hash: "h3",
+		text: "Welcome back, {name}"
+	}, {
+		hash: "h4",
+		text: "Your email is {email}"
+	}] }
+}, {
+	version: .1,
+	results: { result: [{
+		hash: "h3",
+		shouldPluralize: false,
+		reasoning: "Variable {name} is a person's name, not a count. No pluralization needed."
+	}, {
+		hash: "h4",
+		shouldPluralize: false,
+		reasoning: "Variable {email} is an email address, not a count. No pluralization needed."
+	}] }
+}]];
+
+//#endregion
+export { shots };
+//# sourceMappingURL=shots.mjs.map

package/build/translators/pluralization/shots.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"shots.mjs","names":["shots: [PluralizationBatch, PluralizationResponse][]"],"sources":["../../../src/translators/pluralization/shots.ts"],"sourcesContent":["import type { PluralizationBatch, PluralizationResponse } from \"./types\";\n\n/**\n * Few-shot examples: [input, output]\n */\nexport const shots: [PluralizationBatch, PluralizationResponse][] = [\n  // Example 1: Should pluralize\n  [\n    {\n      version: 0.1,\n      sourceLocale: \"en\",\n      candidates: {\n        candidate: [\n          {\n            hash: \"h1\",\n            text: \"You have {count} items in your cart\",\n          },\n          {\n            hash: \"h2\",\n            text: \"{fileCount} files uploaded\",\n          },\n        ],\n      },\n    },\n    {\n      version: 0.1,\n      results: {\n        result: [\n          {\n            hash: \"h1\",\n            shouldPluralize: true,\n            icuText:\n              \"You have {count, plural, =0 {no items} one {# item} other {# items}} in your cart\",\n            reasoning: \"Message varies based on item count\",\n          },\n          {\n            hash: \"h2\",\n            shouldPluralize: true,\n            icuText:\n              \"{fileCount, plural, =0 {No files uploaded} one {# file uploaded} other {# files uploaded}}\",\n            reasoning: \"Message varies based on file count\",\n          },\n        ],\n      },\n    },\n  ],\n\n  // Example 2: Should NOT pluralize\n  [\n    {\n      version: 0.1,\n      sourceLocale: \"en\",\n      candidates: {\n        candidate: [\n          {\n            hash: \"h3\",\n            text: \"Welcome back, {name}\",\n          },\n          {\n            hash: \"h4\",\n            text: \"Your email is {email}\",\n          },\n        ],\n      },\n    },\n    {\n      version: 0.1,\n      results: {\n        result: [\n          {\n            hash: \"h3\",\n            shouldPluralize: false,\n            reasoning:\n              \"Variable {name} is a person's name, not a count. No pluralization needed.\",\n          },\n          {\n            hash: \"h4\",\n            shouldPluralize: false,\n            reasoning:\n              \"Variable {email} is an email address, not a count. No pluralization needed.\",\n          },\n        ],\n      },\n    },\n  ],\n];\n"],"mappings":";;;;AAKA,MAAaA,QAAuD,CAElE,CACE;CACE,SAAS;CACT,cAAc;CACd,YAAY,EACV,WAAW,CACT;EACE,MAAM;EACN,MAAM;EACP,EACD;EACE,MAAM;EACN,MAAM;EACP,CACF,EACF;CACF,EACD;CACE,SAAS;CACT,SAAS,EACP,QAAQ,CACN;EACE,MAAM;EACN,iBAAiB;EACjB,SACE;EACF,WAAW;EACZ,EACD;EACE,MAAM;EACN,iBAAiB;EACjB,SACE;EACF,WAAW;EACZ,CACF,EACF;CACF,CACF,EAGD,CACE;CACE,SAAS;CACT,cAAc;CACd,YAAY,EACV,WAAW,CACT;EACE,MAAM;EACN,MAAM;EACP,EACD;EACE,MAAM;EACN,MAAM;EACP,CACF,EACF;CACF,EACD;CACE,SAAS;CACT,SAAS,EACP,QAAQ,CACN;EACE,MAAM;EACN,iBAAiB;EACjB,WACE;EACH,EACD;EACE,MAAM;EACN,iBAAiB;EACjB,WACE;EACH,CACF,EACF;CACF,CACF,CACF"}

package/build/translators/pluralization/types.d.cts

@@ -0,0 +1,17 @@
+import { LocaleCode } from "lingo.dev/spec";
+
+//#region src/translators/pluralization/types.d.ts
+
+type PluralizationConfig = {
+  sourceLocale: LocaleCode;
+  enabled: boolean;
+  /**
+   * LLM provider for pluralization detection
+   * Format: "provider:model" (e.g., "groq:llama3-8b-8192")
+   * @default "groq:llama3-8b-8192"
+   */
+  model: string;
+};
+//#endregion
+export { PluralizationConfig };
+//# sourceMappingURL=types.d.cts.map

package/build/translators/pluralization/types.d.cts.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.d.cts","names":[],"sources":["../../../src/translators/pluralization/types.ts"],"sourcesContent":[],"mappings":";;;;KAkBY,mBAAA;gBACI"}

package/build/translators/pluralization/types.d.mts

@@ -0,0 +1,17 @@
+import { LocaleCode } from "lingo.dev/spec";
+
+//#region src/translators/pluralization/types.d.ts
+
+type PluralizationConfig = {
+  sourceLocale: LocaleCode;
+  enabled: boolean;
+  /**
+   * LLM provider for pluralization detection
+   * Format: "provider:model" (e.g., "groq:llama3-8b-8192")
+   * @default "groq:llama3-8b-8192"
+   */
+  model: string;
+};
+//#endregion
+export { PluralizationConfig };
+//# sourceMappingURL=types.d.mts.map