scrapex 0.5.3 → 1.0.0-alpha.1

package/dist/enhancer-Q6CSc1gA.mjs

@@ -0,0 +1,220 @@
+import { z } from "zod";
+
+//#region src/core/errors.ts
+/**
+* Custom error class for scraping failures with structured error codes
+*/
+var ScrapeError = class ScrapeError extends Error {
+	code;
+	statusCode;
+	constructor(message, code, statusCode, cause) {
+		super(message, { cause });
+		this.name = "ScrapeError";
+		this.code = code;
+		this.statusCode = statusCode;
+		if (Error.captureStackTrace) Error.captureStackTrace(this, ScrapeError);
+	}
+	/**
+	* Create a ScrapeError from an unknown error
+	*/
+	static from(error, code = "FETCH_FAILED") {
+		if (error instanceof ScrapeError) return error;
+		if (error instanceof Error) return new ScrapeError(error.message, code, void 0, error);
+		return new ScrapeError(String(error), code);
+	}
+	/**
+	* Check if error is retryable (network issues, timeouts)
+	*/
+	isRetryable() {
+		return this.code === "FETCH_FAILED" || this.code === "TIMEOUT";
+	}
+	/**
+	* Convert to a plain object for serialization
+	*/
+	toJSON() {
+		return {
+			name: this.name,
+			message: this.message,
+			code: this.code,
+			statusCode: this.statusCode,
+			stack: this.stack
+		};
+	}
+};
+
+//#endregion
+//#region src/llm/types.ts
+/**
+* Zod schemas for LLM outputs
+*/
+const SummarySchema = z.object({ summary: z.string().describe("A concise 2-3 sentence summary of the content") });
+const TagsSchema = z.object({ tags: z.array(z.string()).describe("5-10 relevant tags/keywords") });
+const EntitiesSchema = z.object({
+	people: z.array(z.string()).describe("People mentioned"),
+	organizations: z.array(z.string()).describe("Organizations/companies"),
+	technologies: z.array(z.string()).describe("Technologies/tools/frameworks"),
+	locations: z.array(z.string()).describe("Locations/places"),
+	concepts: z.array(z.string()).describe("Key concepts/topics")
+});
+const ClassifySchema = z.object({
+	contentType: z.enum([
+		"article",
+		"repo",
+		"docs",
+		"package",
+		"video",
+		"tool",
+		"product",
+		"unknown"
+	]).describe("The type of content"),
+	confidence: z.number().min(0).max(1).describe("Confidence score 0-1")
+});
+
+//#endregion
+//#region src/llm/enhancer.ts
+/**
+* Enhance scraped data with LLM-powered features
+*/
+async function enhance(data, provider, types) {
+	const results = {};
+	const content = data.excerpt || data.textContent.slice(0, 1e4);
+	const context = `Title: ${data.title}\nURL: ${data.url}\n\nContent:\n${content}`;
+	const promises = [];
+	if (types.includes("summarize")) promises.push(summarize(context, provider).then((summary) => {
+		results.summary = summary;
+	}));
+	if (types.includes("tags")) promises.push(extractTags(context, provider).then((tags) => {
+		results.suggestedTags = tags;
+	}));
+	if (types.includes("entities")) promises.push(extractEntities(context, provider).then((entities) => {
+		results.entities = entities;
+	}));
+	if (types.includes("classify")) promises.push(classify(context, provider).then((classification) => {
+		if (classification.confidence > .7) results.contentType = classification.contentType;
+	}));
+	await Promise.all(promises);
+	return results;
+}
+/**
+* Ask a custom question about the scraped content
+* Results are stored in the `custom` field of ScrapedData
+*/
+async function ask(data, provider, prompt, options) {
+	const key = options?.key || "response";
+	const content = data.excerpt || data.textContent.slice(0, 1e4);
+	const processedPrompt = applyPlaceholders(prompt, data, content);
+	if (options?.schema) {
+		const result = await extract(data, provider, options.schema, processedPrompt);
+		return { custom: { [key]: result } };
+	}
+	const fullPrompt = prompt.includes("{{content}}") ? processedPrompt : `${processedPrompt}\n\nTitle: ${data.title}\nURL: ${data.url}\n\nContent:\n${content}`;
+	const response = await provider.complete(fullPrompt);
+	return { custom: { [key]: response } };
+}
+/**
+* Apply placeholder replacements to a prompt template
+*/
+function applyPlaceholders(prompt, data, content) {
+	const domain = (() => {
+		try {
+			return new URL(data.url).hostname;
+		} catch {
+			return "";
+		}
+	})();
+	return prompt.replace(/\{\{title\}\}/g, data.title).replace(/\{\{url\}\}/g, data.url).replace(/\{\{content\}\}/g, content).replace(/\{\{description\}\}/g, data.description || "").replace(/\{\{excerpt\}\}/g, data.excerpt || "").replace(/\{\{domain\}\}/g, domain);
+}
+/**
+* Extract structured data using LLM and a custom schema
+*/
+async function extract(data, provider, schema, promptTemplate) {
+	const zodShape = {};
+	for (const [key, type] of Object.entries(schema)) {
+		const isOptional = type.endsWith("?");
+		const baseType = isOptional ? type.slice(0, -1) : type;
+		let zodType;
+		switch (baseType) {
+			case "string":
+				zodType = z.string();
+				break;
+			case "number":
+				zodType = z.number();
+				break;
+			case "boolean":
+				zodType = z.boolean();
+				break;
+			case "string[]":
+				zodType = z.array(z.string());
+				break;
+			case "number[]":
+				zodType = z.array(z.number());
+				break;
+			default: zodType = z.string();
+		}
+		zodShape[key] = isOptional ? zodType.optional() : zodType;
+	}
+	const zodSchema = z.object(zodShape);
+	const content = data.textContent.slice(0, 4e3);
+	let prompt;
+	if (promptTemplate) {
+		prompt = applyPlaceholders(promptTemplate, data, content);
+		if (!promptTemplate.includes("{{content}}")) prompt += `\n\nContext:\n${content}`;
+	} else prompt = `Extract the following information from this content:
+
+Title: ${data.title}
+URL: ${data.url}
+
+Content:
+${content}
+
+Extract these fields:
+${Object.entries(schema).map(([key, type]) => `- ${key} (${type})`).join("\n")}`;
+	return provider.completeJSON(prompt, zodSchema);
+}
+/**
+* Generate a summary of the content
+*/
+async function summarize(context, provider) {
+	const prompt = `Summarize the following content in 2-3 concise sentences:
+
+${context}`;
+	return (await provider.completeJSON(prompt, SummarySchema)).summary;
+}
+/**
+* Extract relevant tags/keywords
+*/
+async function extractTags(context, provider) {
+	const prompt = `Extract 5-10 relevant tags or keywords from the following content. Focus on technologies, concepts, and topics mentioned:
+
+${context}`;
+	return (await provider.completeJSON(prompt, TagsSchema)).tags;
+}
+/**
+* Extract named entities from content
+*/
+async function extractEntities(context, provider) {
+	const prompt = `Extract named entities from the following content. Identify people, organizations, technologies, locations, and key concepts:
+
+${context}`;
+	return provider.completeJSON(prompt, EntitiesSchema);
+}
+/**
+* Classify content type using LLM
+*/
+async function classify(context, provider) {
+	const prompt = `Classify the following content into one of these categories:
+- article: Blog post, news article, essay
+- repo: Code repository, open source project
+- docs: Documentation, API reference, guides
+- package: npm/pip package page
+- video: Video content, YouTube
+- tool: Software tool, web application
+- product: Commercial product, e-commerce
+
+${context}`;
+	return provider.completeJSON(prompt, ClassifySchema);
+}
+
+//#endregion
+export { EntitiesSchema as a, ScrapeError as c, ClassifySchema as i, enhance as n, SummarySchema as o, extract as r, TagsSchema as s, ask as t };
+//# sourceMappingURL=enhancer-Q6CSc1gA.mjs.map

package/dist/enhancer-Q6CSc1gA.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"enhancer-Q6CSc1gA.mjs","names":["results: Partial<ScrapedData>","promises: Promise<void>[]","zodShape: Record<string, z.ZodTypeAny>","zodType: z.ZodTypeAny","prompt: string"],"sources":["../src/core/errors.ts","../src/llm/types.ts","../src/llm/enhancer.ts"],"sourcesContent":["/**\n * Error codes for scraping failures\n */\nexport type ScrapeErrorCode =\n  | 'FETCH_FAILED'\n  | 'TIMEOUT'\n  | 'INVALID_URL'\n  | 'BLOCKED'\n  | 'NOT_FOUND'\n  | 'ROBOTS_BLOCKED'\n  | 'PARSE_ERROR'\n  | 'LLM_ERROR'\n  | 'VALIDATION_ERROR';\n\n/**\n * Custom error class for scraping failures with structured error codes\n */\nexport class ScrapeError extends Error {\n  public readonly code: ScrapeErrorCode;\n  public readonly statusCode?: number;\n\n  constructor(message: string, code: ScrapeErrorCode, statusCode?: number, cause?: Error) {\n    super(message, { cause });\n    this.name = 'ScrapeError';\n    this.code = code;\n    this.statusCode = statusCode;\n\n    // Maintains proper stack trace in V8 environments\n    if (Error.captureStackTrace) {\n      Error.captureStackTrace(this, ScrapeError);\n    }\n  }\n\n  /**\n   * Create a ScrapeError from an unknown error\n   */\n  static from(error: unknown, code: ScrapeErrorCode = 'FETCH_FAILED'): ScrapeError {\n    if (error instanceof ScrapeError) {\n      return error;\n    }\n\n    if (error instanceof Error) {\n      return new ScrapeError(error.message, code, undefined, error);\n    }\n\n    return new ScrapeError(String(error), code);\n  }\n\n  /**\n   * Check if error is retryable (network issues, timeouts)\n   */\n  isRetryable(): boolean {\n    return this.code === 'FETCH_FAILED' || this.code === 'TIMEOUT';\n  }\n\n  /**\n   * Convert to a plain object for serialization\n   */\n  toJSON(): Record<string, unknown> {\n    return {\n      name: this.name,\n      message: this.message,\n      code: this.code,\n      statusCode: this.statusCode,\n      stack: this.stack,\n    };\n  }\n}\n","import { z } from 'zod';\n\n/**\n * LLM completion options\n */\nexport interface CompletionOptions {\n  maxTokens?: number;\n  temperature?: number;\n  systemPrompt?: string;\n}\n\n/**\n * LLM Provider interface - implemented by all providers\n */\nexport interface LLMProvider {\n  readonly name: string;\n\n  /**\n   * Generate a text completion\n   */\n  complete(prompt: string, options?: CompletionOptions): Promise<string>;\n\n  /**\n   * Generate a structured JSON completion with Zod validation\n   */\n  completeJSON<T>(prompt: string, schema: z.ZodType<T>, options?: CompletionOptions): Promise<T>;\n}\n\n/**\n * Provider configuration for Anthropic\n */\nexport interface AnthropicConfig {\n  apiKey?: string; // Falls back to ANTHROPIC_API_KEY env var\n  model?: string; // Default: claude-3-haiku-20240307\n  baseUrl?: string;\n}\n\n/**\n * Provider configuration for OpenAI-compatible APIs\n * Works with: OpenAI, Ollama, LM Studio, LocalAI, vLLM, etc.\n */\nexport interface OpenAICompatibleConfig {\n  apiKey?: string; // Falls back to OPENAI_API_KEY env var\n  model?: string; // Default: gpt-4o-mini\n  baseUrl?: string; // Default: https://api.openai.com/v1\n}\n\n/**\n * Enhancement result types\n */\nexport interface SummaryResult {\n  summary: string;\n}\n\nexport interface TagsResult {\n  tags: string[];\n}\n\nexport interface EntitiesResult {\n  people: string[];\n  organizations: string[];\n  technologies: string[];\n  locations: string[];\n  concepts: string[];\n}\n\nexport interface ClassifyResult {\n  contentType: string;\n  confidence: number;\n}\n\n/**\n * Zod schemas for LLM outputs\n */\nexport const SummarySchema = z.object({\n  summary: z.string().describe('A concise 2-3 sentence summary of the content'),\n});\n\nexport const TagsSchema = z.object({\n  tags: z.array(z.string()).describe('5-10 relevant tags/keywords'),\n});\n\nexport const EntitiesSchema = z.object({\n  people: z.array(z.string()).describe('People mentioned'),\n  organizations: z.array(z.string()).describe('Organizations/companies'),\n  technologies: z.array(z.string()).describe('Technologies/tools/frameworks'),\n  locations: z.array(z.string()).describe('Locations/places'),\n  concepts: z.array(z.string()).describe('Key concepts/topics'),\n});\n\nexport const ClassifySchema = z.object({\n  contentType: z\n    .enum(['article', 'repo', 'docs', 'package', 'video', 'tool', 'product', 'unknown'])\n    .describe('The type of content'),\n  confidence: z.number().min(0).max(1).describe('Confidence score 0-1'),\n});\n","import { z } from 'zod';\nimport type {\n  EnhancementType,\n  ExtractedEntities,\n  ExtractionSchema,\n  ScrapedData,\n} from '@/core/types.js';\nimport type { LLMProvider } from './types.js';\nimport { ClassifySchema, EntitiesSchema, SummarySchema, TagsSchema } from './types.js';\n\n/**\n * Enhance scraped data with LLM-powered features\n */\nexport async function enhance(\n  data: ScrapedData,\n  provider: LLMProvider,\n  types: EnhancementType[]\n): Promise<Partial<ScrapedData>> {\n  const results: Partial<ScrapedData> = {};\n\n  // Prepare content for LLM (use excerpt/textContent to save tokens)\n  const content = data.excerpt || data.textContent.slice(0, 10000);\n  const context = `Title: ${data.title}\\nURL: ${data.url}\\n\\nContent:\\n${content}`;\n\n  // Run enhancements in parallel\n  const promises: Promise<void>[] = [];\n\n  if (types.includes('summarize')) {\n    promises.push(\n      summarize(context, provider).then((summary) => {\n        results.summary = summary;\n      })\n    );\n  }\n\n  if (types.includes('tags')) {\n    promises.push(\n      extractTags(context, provider).then((tags) => {\n        results.suggestedTags = tags;\n      })\n    );\n  }\n\n  if (types.includes('entities')) {\n    promises.push(\n      extractEntities(context, provider).then((entities) => {\n        results.entities = entities;\n      })\n    );\n  }\n\n  if (types.includes('classify')) {\n    promises.push(\n      classify(context, provider).then((classification) => {\n        if (classification.confidence > 0.7) {\n          results.contentType = classification.contentType as ScrapedData['contentType'];\n        }\n      })\n    );\n  }\n\n  await Promise.all(promises);\n\n  return results;\n}\n\n/**\n * Options for the ask() function\n */\nexport interface AskOptions {\n  /** Key to store the result under in custom field */\n  key?: string;\n  /** Schema for structured response */\n  schema?: ExtractionSchema;\n}\n\n/**\n * Ask a custom question about the scraped content\n * Results are stored in the `custom` field of ScrapedData\n */\nexport async function ask(\n  data: ScrapedData,\n  provider: LLMProvider,\n  prompt: string,\n  options?: AskOptions\n): Promise<Partial<ScrapedData>> {\n  const key = options?.key || 'response';\n  const content = data.excerpt || data.textContent.slice(0, 10000);\n\n  // Apply placeholder replacements\n  const processedPrompt = applyPlaceholders(prompt, data, content);\n\n  if (options?.schema) {\n    // Use structured extraction\n    const result = await extract(data, provider, options.schema, processedPrompt);\n    return { custom: { [key]: result } };\n  }\n\n  // Simple string response\n  const fullPrompt = prompt.includes('{{content}}')\n    ? processedPrompt\n    : `${processedPrompt}\\n\\nTitle: ${data.title}\\nURL: ${data.url}\\n\\nContent:\\n${content}`;\n\n  const response = await provider.complete(fullPrompt);\n  return { custom: { [key]: response } };\n}\n\n/**\n * Apply placeholder replacements to a prompt template\n */\nfunction applyPlaceholders(prompt: string, data: ScrapedData, content: string): string {\n  const domain = (() => {\n    try {\n      return new URL(data.url).hostname;\n    } catch {\n      return '';\n    }\n  })();\n\n  return prompt\n    .replace(/\\{\\{title\\}\\}/g, data.title)\n    .replace(/\\{\\{url\\}\\}/g, data.url)\n    .replace(/\\{\\{content\\}\\}/g, content)\n    .replace(/\\{\\{description\\}\\}/g, data.description || '')\n    .replace(/\\{\\{excerpt\\}\\}/g, data.excerpt || '')\n    .replace(/\\{\\{domain\\}\\}/g, domain);\n}\n\n/**\n * Extract structured data using LLM and a custom schema\n */\nexport async function extract<T>(\n  data: ScrapedData,\n  provider: LLMProvider,\n  schema: ExtractionSchema,\n  promptTemplate?: string\n): Promise<T> {\n  // Convert simple schema to Zod schema\n  const zodShape: Record<string, z.ZodTypeAny> = {};\n\n  for (const [key, type] of Object.entries(schema)) {\n    const isOptional = type.endsWith('?');\n    const baseType = isOptional ? type.slice(0, -1) : type;\n\n    let zodType: z.ZodTypeAny;\n    switch (baseType) {\n      case 'string':\n        zodType = z.string();\n        break;\n      case 'number':\n        zodType = z.number();\n        break;\n      case 'boolean':\n        zodType = z.boolean();\n        break;\n      case 'string[]':\n        zodType = z.array(z.string());\n        break;\n      case 'number[]':\n        zodType = z.array(z.number());\n        break;\n      default:\n        zodType = z.string();\n    }\n\n    zodShape[key] = isOptional ? zodType.optional() : zodType;\n  }\n\n  const zodSchema = z.object(zodShape) as unknown as z.ZodType<T>;\n\n  const content = data.textContent.slice(0, 4000);\n\n  let prompt: string;\n\n  if (promptTemplate) {\n    // Apply all placeholder replacements\n    prompt = applyPlaceholders(promptTemplate, data, content);\n\n    // If content wasn't included via placeholder, append it\n    if (!promptTemplate.includes('{{content}}')) {\n      prompt += `\\n\\nContext:\\n${content}`;\n    }\n  } else {\n    prompt = `Extract the following information from this content:\n\nTitle: ${data.title}\nURL: ${data.url}\n\nContent:\n${content}\n\nExtract these fields:\n${Object.entries(schema)\n  .map(([key, type]) => `- ${key} (${type})`)\n  .join('\\n')}`;\n  }\n\n  return provider.completeJSON<T>(prompt, zodSchema as z.ZodType<T>);\n}\n\n/**\n * Generate a summary of the content\n */\nasync function summarize(context: string, provider: LLMProvider): Promise<string> {\n  const prompt = `Summarize the following content in 2-3 concise sentences:\n\n${context}`;\n\n  const result = await provider.completeJSON(prompt, SummarySchema);\n  return result.summary;\n}\n\n/**\n * Extract relevant tags/keywords\n */\nasync function extractTags(context: string, provider: LLMProvider): Promise<string[]> {\n  const prompt = `Extract 5-10 relevant tags or keywords from the following content. Focus on technologies, concepts, and topics mentioned:\n\n${context}`;\n\n  const result = await provider.completeJSON(prompt, TagsSchema);\n  return result.tags;\n}\n\n/**\n * Extract named entities from content\n */\nasync function extractEntities(context: string, provider: LLMProvider): Promise<ExtractedEntities> {\n  const prompt = `Extract named entities from the following content. Identify people, organizations, technologies, locations, and key concepts:\n\n${context}`;\n\n  return provider.completeJSON(prompt, EntitiesSchema);\n}\n\n/**\n * Classify content type using LLM\n */\nasync function classify(\n  context: string,\n  provider: LLMProvider\n): Promise<{ contentType: string; confidence: number }> {\n  const prompt = `Classify the following content into one of these categories:\n- article: Blog post, news article, essay\n- repo: Code repository, open source project\n- docs: Documentation, API reference, guides\n- package: npm/pip package page\n- video: Video content, YouTube\n- tool: Software tool, web application\n- product: Commercial product, e-commerce\n\n${context}`;\n\n  return provider.completeJSON(prompt, ClassifySchema);\n}\n"],"mappings":";;;;;;AAiBA,IAAa,cAAb,MAAa,oBAAoB,MAAM;CACrC,AAAgB;CAChB,AAAgB;CAEhB,YAAY,SAAiB,MAAuB,YAAqB,OAAe;AACtF,QAAM,SAAS,EAAE,OAAO,CAAC;AACzB,OAAK,OAAO;AACZ,OAAK,OAAO;AACZ,OAAK,aAAa;AAGlB,MAAI,MAAM,kBACR,OAAM,kBAAkB,MAAM,YAAY;;;;;CAO9C,OAAO,KAAK,OAAgB,OAAwB,gBAA6B;AAC/E,MAAI,iBAAiB,YACnB,QAAO;AAGT,MAAI,iBAAiB,MACnB,QAAO,IAAI,YAAY,MAAM,SAAS,MAAM,QAAW,MAAM;AAG/D,SAAO,IAAI,YAAY,OAAO,MAAM,EAAE,KAAK;;;;;CAM7C,cAAuB;AACrB,SAAO,KAAK,SAAS,kBAAkB,KAAK,SAAS;;;;;CAMvD,SAAkC;AAChC,SAAO;GACL,MAAM,KAAK;GACX,SAAS,KAAK;GACd,MAAM,KAAK;GACX,YAAY,KAAK;GACjB,OAAO,KAAK;GACb;;;;;;;;;ACSL,MAAa,gBAAgB,EAAE,OAAO,EACpC,SAAS,EAAE,QAAQ,CAAC,SAAS,gDAAgD,EAC9E,CAAC;AAEF,MAAa,aAAa,EAAE,OAAO,EACjC,MAAM,EAAE,MAAM,EAAE,QAAQ,CAAC,CAAC,SAAS,8BAA8B,EAClE,CAAC;AAEF,MAAa,iBAAiB,EAAE,OAAO;CACrC,QAAQ,EAAE,MAAM,EAAE,QAAQ,CAAC,CAAC,SAAS,mBAAmB;CACxD,eAAe,EAAE,MAAM,EAAE,QAAQ,CAAC,CAAC,SAAS,0BAA0B;CACtE,cAAc,EAAE,MAAM,EAAE,QAAQ,CAAC,CAAC,SAAS,gCAAgC;CAC3E,WAAW,EAAE,MAAM,EAAE,QAAQ,CAAC,CAAC,SAAS,mBAAmB;CAC3D,UAAU,EAAE,MAAM,EAAE,QAAQ,CAAC,CAAC,SAAS,sBAAsB;CAC9D,CAAC;AAEF,MAAa,iBAAiB,EAAE,OAAO;CACrC,aAAa,EACV,KAAK;EAAC;EAAW;EAAQ;EAAQ;EAAW;EAAS;EAAQ;EAAW;EAAU,CAAC,CACnF,SAAS,sBAAsB;CAClC,YAAY,EAAE,QAAQ,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,CAAC,SAAS,uBAAuB;CACtE,CAAC;;;;;;;AClFF,eAAsB,QACpB,MACA,UACA,OAC+B;CAC/B,MAAMA,UAAgC,EAAE;CAGxC,MAAM,UAAU,KAAK,WAAW,KAAK,YAAY,MAAM,GAAG,IAAM;CAChE,MAAM,UAAU,UAAU,KAAK,MAAM,SAAS,KAAK,IAAI,gBAAgB;CAGvE,MAAMC,WAA4B,EAAE;AAEpC,KAAI,MAAM,SAAS,YAAY,CAC7B,UAAS,KACP,UAAU,SAAS,SAAS,CAAC,MAAM,YAAY;AAC7C,UAAQ,UAAU;GAClB,CACH;AAGH,KAAI,MAAM,SAAS,OAAO,CACxB,UAAS,KACP,YAAY,SAAS,SAAS,CAAC,MAAM,SAAS;AAC5C,UAAQ,gBAAgB;GACxB,CACH;AAGH,KAAI,MAAM,SAAS,WAAW,CAC5B,UAAS,KACP,gBAAgB,SAAS,SAAS,CAAC,MAAM,aAAa;AACpD,UAAQ,WAAW;GACnB,CACH;AAGH,KAAI,MAAM,SAAS,WAAW,CAC5B,UAAS,KACP,SAAS,SAAS,SAAS,CAAC,MAAM,mBAAmB;AACnD,MAAI,eAAe,aAAa,GAC9B,SAAQ,cAAc,eAAe;GAEvC,CACH;AAGH,OAAM,QAAQ,IAAI,SAAS;AAE3B,QAAO;;;;;;AAiBT,eAAsB,IACpB,MACA,UACA,QACA,SAC+B;CAC/B,MAAM,MAAM,SAAS,OAAO;CAC5B,MAAM,UAAU,KAAK,WAAW,KAAK,YAAY,MAAM,GAAG,IAAM;CAGhE,MAAM,kBAAkB,kBAAkB,QAAQ,MAAM,QAAQ;AAEhE,KAAI,SAAS,QAAQ;EAEnB,MAAM,SAAS,MAAM,QAAQ,MAAM,UAAU,QAAQ,QAAQ,gBAAgB;AAC7E,SAAO,EAAE,QAAQ,GAAG,MAAM,QAAQ,EAAE;;CAItC,MAAM,aAAa,OAAO,SAAS,cAAc,GAC7C,kBACA,GAAG,gBAAgB,aAAa,KAAK,MAAM,SAAS,KAAK,IAAI,gBAAgB;CAEjF,MAAM,WAAW,MAAM,SAAS,SAAS,WAAW;AACpD,QAAO,EAAE,QAAQ,GAAG,MAAM,UAAU,EAAE;;;;;AAMxC,SAAS,kBAAkB,QAAgB,MAAmB,SAAyB;CACrF,MAAM,gBAAgB;AACpB,MAAI;AACF,UAAO,IAAI,IAAI,KAAK,IAAI,CAAC;UACnB;AACN,UAAO;;KAEP;AAEJ,QAAO,OACJ,QAAQ,kBAAkB,KAAK,MAAM,CACrC,QAAQ,gBAAgB,KAAK,IAAI,CACjC,QAAQ,oBAAoB,QAAQ,CACpC,QAAQ,wBAAwB,KAAK,eAAe,GAAG,CACvD,QAAQ,oBAAoB,KAAK,WAAW,GAAG,CAC/C,QAAQ,mBAAmB,OAAO;;;;;AAMvC,eAAsB,QACpB,MACA,UACA,QACA,gBACY;CAEZ,MAAMC,WAAyC,EAAE;AAEjD,MAAK,MAAM,CAAC,KAAK,SAAS,OAAO,QAAQ,OAAO,EAAE;EAChD,MAAM,aAAa,KAAK,SAAS,IAAI;EACrC,MAAM,WAAW,aAAa,KAAK,MAAM,GAAG,GAAG,GAAG;EAElD,IAAIC;AACJ,UAAQ,UAAR;GACE,KAAK;AACH,cAAU,EAAE,QAAQ;AACpB;GACF,KAAK;AACH,cAAU,EAAE,QAAQ;AACpB;GACF,KAAK;AACH,cAAU,EAAE,SAAS;AACrB;GACF,KAAK;AACH,cAAU,EAAE,MAAM,EAAE,QAAQ,CAAC;AAC7B;GACF,KAAK;AACH,cAAU,EAAE,MAAM,EAAE,QAAQ,CAAC;AAC7B;GACF,QACE,WAAU,EAAE,QAAQ;;AAGxB,WAAS,OAAO,aAAa,QAAQ,UAAU,GAAG;;CAGpD,MAAM,YAAY,EAAE,OAAO,SAAS;CAEpC,MAAM,UAAU,KAAK,YAAY,MAAM,GAAG,IAAK;CAE/C,IAAIC;AAEJ,KAAI,gBAAgB;AAElB,WAAS,kBAAkB,gBAAgB,MAAM,QAAQ;AAGzD,MAAI,CAAC,eAAe,SAAS,cAAc,CACzC,WAAU,iBAAiB;OAG7B,UAAS;;SAEJ,KAAK,MAAM;OACb,KAAK,IAAI;;;EAGd,QAAQ;;;EAGR,OAAO,QAAQ,OAAO,CACrB,KAAK,CAAC,KAAK,UAAU,KAAK,IAAI,IAAI,KAAK,GAAG,CAC1C,KAAK,KAAK;AAGX,QAAO,SAAS,aAAgB,QAAQ,UAA0B;;;;;AAMpE,eAAe,UAAU,SAAiB,UAAwC;CAChF,MAAM,SAAS;;EAEf;AAGA,SADe,MAAM,SAAS,aAAa,QAAQ,cAAc,EACnD;;;;;AAMhB,eAAe,YAAY,SAAiB,UAA0C;CACpF,MAAM,SAAS;;EAEf;AAGA,SADe,MAAM,SAAS,aAAa,QAAQ,WAAW,EAChD;;;;;AAMhB,eAAe,gBAAgB,SAAiB,UAAmD;CACjG,MAAM,SAAS;;EAEf;AAEA,QAAO,SAAS,aAAa,QAAQ,eAAe;;;;;AAMtD,eAAe,SACb,SACA,UACsD;CACtD,MAAM,SAAS;;;;;;;;;EASf;AAEA,QAAO,SAAS,aAAa,QAAQ,eAAe"}

package/dist/enhancer-oM4BhYYS.cjs

@@ -0,0 +1,268 @@
+const require_index = require('./index.cjs');
+let zod = require("zod");
+
+//#region src/core/errors.ts
+/**
+* Custom error class for scraping failures with structured error codes
+*/
+var ScrapeError = class ScrapeError extends Error {
+	code;
+	statusCode;
+	constructor(message, code, statusCode, cause) {
+		super(message, { cause });
+		this.name = "ScrapeError";
+		this.code = code;
+		this.statusCode = statusCode;
+		if (Error.captureStackTrace) Error.captureStackTrace(this, ScrapeError);
+	}
+	/**
+	* Create a ScrapeError from an unknown error
+	*/
+	static from(error, code = "FETCH_FAILED") {
+		if (error instanceof ScrapeError) return error;
+		if (error instanceof Error) return new ScrapeError(error.message, code, void 0, error);
+		return new ScrapeError(String(error), code);
+	}
+	/**
+	* Check if error is retryable (network issues, timeouts)
+	*/
+	isRetryable() {
+		return this.code === "FETCH_FAILED" || this.code === "TIMEOUT";
+	}
+	/**
+	* Convert to a plain object for serialization
+	*/
+	toJSON() {
+		return {
+			name: this.name,
+			message: this.message,
+			code: this.code,
+			statusCode: this.statusCode,
+			stack: this.stack
+		};
+	}
+};
+
+//#endregion
+//#region src/llm/types.ts
+/**
+* Zod schemas for LLM outputs
+*/
+const SummarySchema = zod.z.object({ summary: zod.z.string().describe("A concise 2-3 sentence summary of the content") });
+const TagsSchema = zod.z.object({ tags: zod.z.array(zod.z.string()).describe("5-10 relevant tags/keywords") });
+const EntitiesSchema = zod.z.object({
+	people: zod.z.array(zod.z.string()).describe("People mentioned"),
+	organizations: zod.z.array(zod.z.string()).describe("Organizations/companies"),
+	technologies: zod.z.array(zod.z.string()).describe("Technologies/tools/frameworks"),
+	locations: zod.z.array(zod.z.string()).describe("Locations/places"),
+	concepts: zod.z.array(zod.z.string()).describe("Key concepts/topics")
+});
+const ClassifySchema = zod.z.object({
+	contentType: zod.z.enum([
+		"article",
+		"repo",
+		"docs",
+		"package",
+		"video",
+		"tool",
+		"product",
+		"unknown"
+	]).describe("The type of content"),
+	confidence: zod.z.number().min(0).max(1).describe("Confidence score 0-1")
+});
+
+//#endregion
+//#region src/llm/enhancer.ts
+/**
+* Enhance scraped data with LLM-powered features
+*/
+async function enhance(data, provider, types) {
+	const results = {};
+	const content = data.excerpt || data.textContent.slice(0, 1e4);
+	const context = `Title: ${data.title}\nURL: ${data.url}\n\nContent:\n${content}`;
+	const promises = [];
+	if (types.includes("summarize")) promises.push(summarize(context, provider).then((summary) => {
+		results.summary = summary;
+	}));
+	if (types.includes("tags")) promises.push(extractTags(context, provider).then((tags) => {
+		results.suggestedTags = tags;
+	}));
+	if (types.includes("entities")) promises.push(extractEntities(context, provider).then((entities) => {
+		results.entities = entities;
+	}));
+	if (types.includes("classify")) promises.push(classify(context, provider).then((classification) => {
+		if (classification.confidence > .7) results.contentType = classification.contentType;
+	}));
+	await Promise.all(promises);
+	return results;
+}
+/**
+* Ask a custom question about the scraped content
+* Results are stored in the `custom` field of ScrapedData
+*/
+async function ask(data, provider, prompt, options) {
+	const key = options?.key || "response";
+	const content = data.excerpt || data.textContent.slice(0, 1e4);
+	const processedPrompt = applyPlaceholders(prompt, data, content);
+	if (options?.schema) {
+		const result = await extract(data, provider, options.schema, processedPrompt);
+		return { custom: { [key]: result } };
+	}
+	const fullPrompt = prompt.includes("{{content}}") ? processedPrompt : `${processedPrompt}\n\nTitle: ${data.title}\nURL: ${data.url}\n\nContent:\n${content}`;
+	const response = await provider.complete(fullPrompt);
+	return { custom: { [key]: response } };
+}
+/**
+* Apply placeholder replacements to a prompt template
+*/
+function applyPlaceholders(prompt, data, content) {
+	const domain = (() => {
+		try {
+			return new URL(data.url).hostname;
+		} catch {
+			return "";
+		}
+	})();
+	return prompt.replace(/\{\{title\}\}/g, data.title).replace(/\{\{url\}\}/g, data.url).replace(/\{\{content\}\}/g, content).replace(/\{\{description\}\}/g, data.description || "").replace(/\{\{excerpt\}\}/g, data.excerpt || "").replace(/\{\{domain\}\}/g, domain);
+}
+/**
+* Extract structured data using LLM and a custom schema
+*/
+async function extract(data, provider, schema, promptTemplate) {
+	const zodShape = {};
+	for (const [key, type] of Object.entries(schema)) {
+		const isOptional = type.endsWith("?");
+		const baseType = isOptional ? type.slice(0, -1) : type;
+		let zodType;
+		switch (baseType) {
+			case "string":
+				zodType = zod.z.string();
+				break;
+			case "number":
+				zodType = zod.z.number();
+				break;
+			case "boolean":
+				zodType = zod.z.boolean();
+				break;
+			case "string[]":
+				zodType = zod.z.array(zod.z.string());
+				break;
+			case "number[]":
+				zodType = zod.z.array(zod.z.number());
+				break;
+			default: zodType = zod.z.string();
+		}
+		zodShape[key] = isOptional ? zodType.optional() : zodType;
+	}
+	const zodSchema = zod.z.object(zodShape);
+	const content = data.textContent.slice(0, 4e3);
+	let prompt;
+	if (promptTemplate) {
+		prompt = applyPlaceholders(promptTemplate, data, content);
+		if (!promptTemplate.includes("{{content}}")) prompt += `\n\nContext:\n${content}`;
+	} else prompt = `Extract the following information from this content:
+
+Title: ${data.title}
+URL: ${data.url}
+
+Content:
+${content}
+
+Extract these fields:
+${Object.entries(schema).map(([key, type]) => `- ${key} (${type})`).join("\n")}`;
+	return provider.completeJSON(prompt, zodSchema);
+}
+/**
+* Generate a summary of the content
+*/
+async function summarize(context, provider) {
+	const prompt = `Summarize the following content in 2-3 concise sentences:
+
+${context}`;
+	return (await provider.completeJSON(prompt, SummarySchema)).summary;
+}
+/**
+* Extract relevant tags/keywords
+*/
+async function extractTags(context, provider) {
+	const prompt = `Extract 5-10 relevant tags or keywords from the following content. Focus on technologies, concepts, and topics mentioned:
+
+${context}`;
+	return (await provider.completeJSON(prompt, TagsSchema)).tags;
+}
+/**
+* Extract named entities from content
+*/
+async function extractEntities(context, provider) {
+	const prompt = `Extract named entities from the following content. Identify people, organizations, technologies, locations, and key concepts:
+
+${context}`;
+	return provider.completeJSON(prompt, EntitiesSchema);
+}
+/**
+* Classify content type using LLM
+*/
+async function classify(context, provider) {
+	const prompt = `Classify the following content into one of these categories:
+- article: Blog post, news article, essay
+- repo: Code repository, open source project
+- docs: Documentation, API reference, guides
+- package: npm/pip package page
+- video: Video content, YouTube
+- tool: Software tool, web application
+- product: Commercial product, e-commerce
+
+${context}`;
+	return provider.completeJSON(prompt, ClassifySchema);
+}
+
+//#endregion
+Object.defineProperty(exports, 'ClassifySchema', {
+  enumerable: true,
+  get: function () {
+    return ClassifySchema;
+  }
+});
+Object.defineProperty(exports, 'EntitiesSchema', {
+  enumerable: true,
+  get: function () {
+    return EntitiesSchema;
+  }
+});
+Object.defineProperty(exports, 'ScrapeError', {
+  enumerable: true,
+  get: function () {
+    return ScrapeError;
+  }
+});
+Object.defineProperty(exports, 'SummarySchema', {
+  enumerable: true,
+  get: function () {
+    return SummarySchema;
+  }
+});
+Object.defineProperty(exports, 'TagsSchema', {
+  enumerable: true,
+  get: function () {
+    return TagsSchema;
+  }
+});
+Object.defineProperty(exports, 'ask', {
+  enumerable: true,
+  get: function () {
+    return ask;
+  }
+});
+Object.defineProperty(exports, 'enhance', {
+  enumerable: true,
+  get: function () {
+    return enhance;
+  }
+});
+Object.defineProperty(exports, 'extract', {
+  enumerable: true,
+  get: function () {
+    return extract;
+  }
+});
+//# sourceMappingURL=enhancer-oM4BhYYS.cjs.map

package/dist/enhancer-oM4BhYYS.cjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"enhancer-oM4BhYYS.cjs","names":["z","results: Partial<ScrapedData>","promises: Promise<void>[]","zodShape: Record<string, z.ZodTypeAny>","zodType: z.ZodTypeAny","z","prompt: string"],"sources":["../src/core/errors.ts","../src/llm/types.ts","../src/llm/enhancer.ts"],"sourcesContent":["/**\n * Error codes for scraping failures\n */\nexport type ScrapeErrorCode =\n  | 'FETCH_FAILED'\n  | 'TIMEOUT'\n  | 'INVALID_URL'\n  | 'BLOCKED'\n  | 'NOT_FOUND'\n  | 'ROBOTS_BLOCKED'\n  | 'PARSE_ERROR'\n  | 'LLM_ERROR'\n  | 'VALIDATION_ERROR';\n\n/**\n * Custom error class for scraping failures with structured error codes\n */\nexport class ScrapeError extends Error {\n  public readonly code: ScrapeErrorCode;\n  public readonly statusCode?: number;\n\n  constructor(message: string, code: ScrapeErrorCode, statusCode?: number, cause?: Error) {\n    super(message, { cause });\n    this.name = 'ScrapeError';\n    this.code = code;\n    this.statusCode = statusCode;\n\n    // Maintains proper stack trace in V8 environments\n    if (Error.captureStackTrace) {\n      Error.captureStackTrace(this, ScrapeError);\n    }\n  }\n\n  /**\n   * Create a ScrapeError from an unknown error\n   */\n  static from(error: unknown, code: ScrapeErrorCode = 'FETCH_FAILED'): ScrapeError {\n    if (error instanceof ScrapeError) {\n      return error;\n    }\n\n    if (error instanceof Error) {\n      return new ScrapeError(error.message, code, undefined, error);\n    }\n\n    return new ScrapeError(String(error), code);\n  }\n\n  /**\n   * Check if error is retryable (network issues, timeouts)\n   */\n  isRetryable(): boolean {\n    return this.code === 'FETCH_FAILED' || this.code === 'TIMEOUT';\n  }\n\n  /**\n   * Convert to a plain object for serialization\n   */\n  toJSON(): Record<string, unknown> {\n    return {\n      name: this.name,\n      message: this.message,\n      code: this.code,\n      statusCode: this.statusCode,\n      stack: this.stack,\n    };\n  }\n}\n","import { z } from 'zod';\n\n/**\n * LLM completion options\n */\nexport interface CompletionOptions {\n  maxTokens?: number;\n  temperature?: number;\n  systemPrompt?: string;\n}\n\n/**\n * LLM Provider interface - implemented by all providers\n */\nexport interface LLMProvider {\n  readonly name: string;\n\n  /**\n   * Generate a text completion\n   */\n  complete(prompt: string, options?: CompletionOptions): Promise<string>;\n\n  /**\n   * Generate a structured JSON completion with Zod validation\n   */\n  completeJSON<T>(prompt: string, schema: z.ZodType<T>, options?: CompletionOptions): Promise<T>;\n}\n\n/**\n * Provider configuration for Anthropic\n */\nexport interface AnthropicConfig {\n  apiKey?: string; // Falls back to ANTHROPIC_API_KEY env var\n  model?: string; // Default: claude-3-haiku-20240307\n  baseUrl?: string;\n}\n\n/**\n * Provider configuration for OpenAI-compatible APIs\n * Works with: OpenAI, Ollama, LM Studio, LocalAI, vLLM, etc.\n */\nexport interface OpenAICompatibleConfig {\n  apiKey?: string; // Falls back to OPENAI_API_KEY env var\n  model?: string; // Default: gpt-4o-mini\n  baseUrl?: string; // Default: https://api.openai.com/v1\n}\n\n/**\n * Enhancement result types\n */\nexport interface SummaryResult {\n  summary: string;\n}\n\nexport interface TagsResult {\n  tags: string[];\n}\n\nexport interface EntitiesResult {\n  people: string[];\n  organizations: string[];\n  technologies: string[];\n  locations: string[];\n  concepts: string[];\n}\n\nexport interface ClassifyResult {\n  contentType: string;\n  confidence: number;\n}\n\n/**\n * Zod schemas for LLM outputs\n */\nexport const SummarySchema = z.object({\n  summary: z.string().describe('A concise 2-3 sentence summary of the content'),\n});\n\nexport const TagsSchema = z.object({\n  tags: z.array(z.string()).describe('5-10 relevant tags/keywords'),\n});\n\nexport const EntitiesSchema = z.object({\n  people: z.array(z.string()).describe('People mentioned'),\n  organizations: z.array(z.string()).describe('Organizations/companies'),\n  technologies: z.array(z.string()).describe('Technologies/tools/frameworks'),\n  locations: z.array(z.string()).describe('Locations/places'),\n  concepts: z.array(z.string()).describe('Key concepts/topics'),\n});\n\nexport const ClassifySchema = z.object({\n  contentType: z\n    .enum(['article', 'repo', 'docs', 'package', 'video', 'tool', 'product', 'unknown'])\n    .describe('The type of content'),\n  confidence: z.number().min(0).max(1).describe('Confidence score 0-1'),\n});\n","import { z } from 'zod';\nimport type {\n  EnhancementType,\n  ExtractedEntities,\n  ExtractionSchema,\n  ScrapedData,\n} from '@/core/types.js';\nimport type { LLMProvider } from './types.js';\nimport { ClassifySchema, EntitiesSchema, SummarySchema, TagsSchema } from './types.js';\n\n/**\n * Enhance scraped data with LLM-powered features\n */\nexport async function enhance(\n  data: ScrapedData,\n  provider: LLMProvider,\n  types: EnhancementType[]\n): Promise<Partial<ScrapedData>> {\n  const results: Partial<ScrapedData> = {};\n\n  // Prepare content for LLM (use excerpt/textContent to save tokens)\n  const content = data.excerpt || data.textContent.slice(0, 10000);\n  const context = `Title: ${data.title}\\nURL: ${data.url}\\n\\nContent:\\n${content}`;\n\n  // Run enhancements in parallel\n  const promises: Promise<void>[] = [];\n\n  if (types.includes('summarize')) {\n    promises.push(\n      summarize(context, provider).then((summary) => {\n        results.summary = summary;\n      })\n    );\n  }\n\n  if (types.includes('tags')) {\n    promises.push(\n      extractTags(context, provider).then((tags) => {\n        results.suggestedTags = tags;\n      })\n    );\n  }\n\n  if (types.includes('entities')) {\n    promises.push(\n      extractEntities(context, provider).then((entities) => {\n        results.entities = entities;\n      })\n    );\n  }\n\n  if (types.includes('classify')) {\n    promises.push(\n      classify(context, provider).then((classification) => {\n        if (classification.confidence > 0.7) {\n          results.contentType = classification.contentType as ScrapedData['contentType'];\n        }\n      })\n    );\n  }\n\n  await Promise.all(promises);\n\n  return results;\n}\n\n/**\n * Options for the ask() function\n */\nexport interface AskOptions {\n  /** Key to store the result under in custom field */\n  key?: string;\n  /** Schema for structured response */\n  schema?: ExtractionSchema;\n}\n\n/**\n * Ask a custom question about the scraped content\n * Results are stored in the `custom` field of ScrapedData\n */\nexport async function ask(\n  data: ScrapedData,\n  provider: LLMProvider,\n  prompt: string,\n  options?: AskOptions\n): Promise<Partial<ScrapedData>> {\n  const key = options?.key || 'response';\n  const content = data.excerpt || data.textContent.slice(0, 10000);\n\n  // Apply placeholder replacements\n  const processedPrompt = applyPlaceholders(prompt, data, content);\n\n  if (options?.schema) {\n    // Use structured extraction\n    const result = await extract(data, provider, options.schema, processedPrompt);\n    return { custom: { [key]: result } };\n  }\n\n  // Simple string response\n  const fullPrompt = prompt.includes('{{content}}')\n    ? processedPrompt\n    : `${processedPrompt}\\n\\nTitle: ${data.title}\\nURL: ${data.url}\\n\\nContent:\\n${content}`;\n\n  const response = await provider.complete(fullPrompt);\n  return { custom: { [key]: response } };\n}\n\n/**\n * Apply placeholder replacements to a prompt template\n */\nfunction applyPlaceholders(prompt: string, data: ScrapedData, content: string): string {\n  const domain = (() => {\n    try {\n      return new URL(data.url).hostname;\n    } catch {\n      return '';\n    }\n  })();\n\n  return prompt\n    .replace(/\\{\\{title\\}\\}/g, data.title)\n    .replace(/\\{\\{url\\}\\}/g, data.url)\n    .replace(/\\{\\{content\\}\\}/g, content)\n    .replace(/\\{\\{description\\}\\}/g, data.description || '')\n    .replace(/\\{\\{excerpt\\}\\}/g, data.excerpt || '')\n    .replace(/\\{\\{domain\\}\\}/g, domain);\n}\n\n/**\n * Extract structured data using LLM and a custom schema\n */\nexport async function extract<T>(\n  data: ScrapedData,\n  provider: LLMProvider,\n  schema: ExtractionSchema,\n  promptTemplate?: string\n): Promise<T> {\n  // Convert simple schema to Zod schema\n  const zodShape: Record<string, z.ZodTypeAny> = {};\n\n  for (const [key, type] of Object.entries(schema)) {\n    const isOptional = type.endsWith('?');\n    const baseType = isOptional ? type.slice(0, -1) : type;\n\n    let zodType: z.ZodTypeAny;\n    switch (baseType) {\n      case 'string':\n        zodType = z.string();\n        break;\n      case 'number':\n        zodType = z.number();\n        break;\n      case 'boolean':\n        zodType = z.boolean();\n        break;\n      case 'string[]':\n        zodType = z.array(z.string());\n        break;\n      case 'number[]':\n        zodType = z.array(z.number());\n        break;\n      default:\n        zodType = z.string();\n    }\n\n    zodShape[key] = isOptional ? zodType.optional() : zodType;\n  }\n\n  const zodSchema = z.object(zodShape) as unknown as z.ZodType<T>;\n\n  const content = data.textContent.slice(0, 4000);\n\n  let prompt: string;\n\n  if (promptTemplate) {\n    // Apply all placeholder replacements\n    prompt = applyPlaceholders(promptTemplate, data, content);\n\n    // If content wasn't included via placeholder, append it\n    if (!promptTemplate.includes('{{content}}')) {\n      prompt += `\\n\\nContext:\\n${content}`;\n    }\n  } else {\n    prompt = `Extract the following information from this content:\n\nTitle: ${data.title}\nURL: ${data.url}\n\nContent:\n${content}\n\nExtract these fields:\n${Object.entries(schema)\n  .map(([key, type]) => `- ${key} (${type})`)\n  .join('\\n')}`;\n  }\n\n  return provider.completeJSON<T>(prompt, zodSchema as z.ZodType<T>);\n}\n\n/**\n * Generate a summary of the content\n */\nasync function summarize(context: string, provider: LLMProvider): Promise<string> {\n  const prompt = `Summarize the following content in 2-3 concise sentences:\n\n${context}`;\n\n  const result = await provider.completeJSON(prompt, SummarySchema);\n  return result.summary;\n}\n\n/**\n * Extract relevant tags/keywords\n */\nasync function extractTags(context: string, provider: LLMProvider): Promise<string[]> {\n  const prompt = `Extract 5-10 relevant tags or keywords from the following content. Focus on technologies, concepts, and topics mentioned:\n\n${context}`;\n\n  const result = await provider.completeJSON(prompt, TagsSchema);\n  return result.tags;\n}\n\n/**\n * Extract named entities from content\n */\nasync function extractEntities(context: string, provider: LLMProvider): Promise<ExtractedEntities> {\n  const prompt = `Extract named entities from the following content. Identify people, organizations, technologies, locations, and key concepts:\n\n${context}`;\n\n  return provider.completeJSON(prompt, EntitiesSchema);\n}\n\n/**\n * Classify content type using LLM\n */\nasync function classify(\n  context: string,\n  provider: LLMProvider\n): Promise<{ contentType: string; confidence: number }> {\n  const prompt = `Classify the following content into one of these categories:\n- article: Blog post, news article, essay\n- repo: Code repository, open source project\n- docs: Documentation, API reference, guides\n- package: npm/pip package page\n- video: Video content, YouTube\n- tool: Software tool, web application\n- product: Commercial product, e-commerce\n\n${context}`;\n\n  return provider.completeJSON(prompt, ClassifySchema);\n}\n"],"mappings":";;;;;;;AAiBA,IAAa,cAAb,MAAa,oBAAoB,MAAM;CACrC,AAAgB;CAChB,AAAgB;CAEhB,YAAY,SAAiB,MAAuB,YAAqB,OAAe;AACtF,QAAM,SAAS,EAAE,OAAO,CAAC;AACzB,OAAK,OAAO;AACZ,OAAK,OAAO;AACZ,OAAK,aAAa;AAGlB,MAAI,MAAM,kBACR,OAAM,kBAAkB,MAAM,YAAY;;;;;CAO9C,OAAO,KAAK,OAAgB,OAAwB,gBAA6B;AAC/E,MAAI,iBAAiB,YACnB,QAAO;AAGT,MAAI,iBAAiB,MACnB,QAAO,IAAI,YAAY,MAAM,SAAS,MAAM,QAAW,MAAM;AAG/D,SAAO,IAAI,YAAY,OAAO,MAAM,EAAE,KAAK;;;;;CAM7C,cAAuB;AACrB,SAAO,KAAK,SAAS,kBAAkB,KAAK,SAAS;;;;;CAMvD,SAAkC;AAChC,SAAO;GACL,MAAM,KAAK;GACX,SAAS,KAAK;GACd,MAAM,KAAK;GACX,YAAY,KAAK;GACjB,OAAO,KAAK;GACb;;;;;;;;;ACSL,MAAa,gBAAgBA,MAAE,OAAO,EACpC,SAASA,MAAE,QAAQ,CAAC,SAAS,gDAAgD,EAC9E,CAAC;AAEF,MAAa,aAAaA,MAAE,OAAO,EACjC,MAAMA,MAAE,MAAMA,MAAE,QAAQ,CAAC,CAAC,SAAS,8BAA8B,EAClE,CAAC;AAEF,MAAa,iBAAiBA,MAAE,OAAO;CACrC,QAAQA,MAAE,MAAMA,MAAE,QAAQ,CAAC,CAAC,SAAS,mBAAmB;CACxD,eAAeA,MAAE,MAAMA,MAAE,QAAQ,CAAC,CAAC,SAAS,0BAA0B;CACtE,cAAcA,MAAE,MAAMA,MAAE,QAAQ,CAAC,CAAC,SAAS,gCAAgC;CAC3E,WAAWA,MAAE,MAAMA,MAAE,QAAQ,CAAC,CAAC,SAAS,mBAAmB;CAC3D,UAAUA,MAAE,MAAMA,MAAE,QAAQ,CAAC,CAAC,SAAS,sBAAsB;CAC9D,CAAC;AAEF,MAAa,iBAAiBA,MAAE,OAAO;CACrC,aAAaA,MACV,KAAK;EAAC;EAAW;EAAQ;EAAQ;EAAW;EAAS;EAAQ;EAAW;EAAU,CAAC,CACnF,SAAS,sBAAsB;CAClC,YAAYA,MAAE,QAAQ,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,CAAC,SAAS,uBAAuB;CACtE,CAAC;;;;;;;AClFF,eAAsB,QACpB,MACA,UACA,OAC+B;CAC/B,MAAMC,UAAgC,EAAE;CAGxC,MAAM,UAAU,KAAK,WAAW,KAAK,YAAY,MAAM,GAAG,IAAM;CAChE,MAAM,UAAU,UAAU,KAAK,MAAM,SAAS,KAAK,IAAI,gBAAgB;CAGvE,MAAMC,WAA4B,EAAE;AAEpC,KAAI,MAAM,SAAS,YAAY,CAC7B,UAAS,KACP,UAAU,SAAS,SAAS,CAAC,MAAM,YAAY;AAC7C,UAAQ,UAAU;GAClB,CACH;AAGH,KAAI,MAAM,SAAS,OAAO,CACxB,UAAS,KACP,YAAY,SAAS,SAAS,CAAC,MAAM,SAAS;AAC5C,UAAQ,gBAAgB;GACxB,CACH;AAGH,KAAI,MAAM,SAAS,WAAW,CAC5B,UAAS,KACP,gBAAgB,SAAS,SAAS,CAAC,MAAM,aAAa;AACpD,UAAQ,WAAW;GACnB,CACH;AAGH,KAAI,MAAM,SAAS,WAAW,CAC5B,UAAS,KACP,SAAS,SAAS,SAAS,CAAC,MAAM,mBAAmB;AACnD,MAAI,eAAe,aAAa,GAC9B,SAAQ,cAAc,eAAe;GAEvC,CACH;AAGH,OAAM,QAAQ,IAAI,SAAS;AAE3B,QAAO;;;;;;AAiBT,eAAsB,IACpB,MACA,UACA,QACA,SAC+B;CAC/B,MAAM,MAAM,SAAS,OAAO;CAC5B,MAAM,UAAU,KAAK,WAAW,KAAK,YAAY,MAAM,GAAG,IAAM;CAGhE,MAAM,kBAAkB,kBAAkB,QAAQ,MAAM,QAAQ;AAEhE,KAAI,SAAS,QAAQ;EAEnB,MAAM,SAAS,MAAM,QAAQ,MAAM,UAAU,QAAQ,QAAQ,gBAAgB;AAC7E,SAAO,EAAE,QAAQ,GAAG,MAAM,QAAQ,EAAE;;CAItC,MAAM,aAAa,OAAO,SAAS,cAAc,GAC7C,kBACA,GAAG,gBAAgB,aAAa,KAAK,MAAM,SAAS,KAAK,IAAI,gBAAgB;CAEjF,MAAM,WAAW,MAAM,SAAS,SAAS,WAAW;AACpD,QAAO,EAAE,QAAQ,GAAG,MAAM,UAAU,EAAE;;;;;AAMxC,SAAS,kBAAkB,QAAgB,MAAmB,SAAyB;CACrF,MAAM,gBAAgB;AACpB,MAAI;AACF,UAAO,IAAI,IAAI,KAAK,IAAI,CAAC;UACnB;AACN,UAAO;;KAEP;AAEJ,QAAO,OACJ,QAAQ,kBAAkB,KAAK,MAAM,CACrC,QAAQ,gBAAgB,KAAK,IAAI,CACjC,QAAQ,oBAAoB,QAAQ,CACpC,QAAQ,wBAAwB,KAAK,eAAe,GAAG,CACvD,QAAQ,oBAAoB,KAAK,WAAW,GAAG,CAC/C,QAAQ,mBAAmB,OAAO;;;;;AAMvC,eAAsB,QACpB,MACA,UACA,QACA,gBACY;CAEZ,MAAMC,WAAyC,EAAE;AAEjD,MAAK,MAAM,CAAC,KAAK,SAAS,OAAO,QAAQ,OAAO,EAAE;EAChD,MAAM,aAAa,KAAK,SAAS,IAAI;EACrC,MAAM,WAAW,aAAa,KAAK,MAAM,GAAG,GAAG,GAAG;EAElD,IAAIC;AACJ,UAAQ,UAAR;GACE,KAAK;AACH,cAAUC,MAAE,QAAQ;AACpB;GACF,KAAK;AACH,cAAUA,MAAE,QAAQ;AACpB;GACF,KAAK;AACH,cAAUA,MAAE,SAAS;AACrB;GACF,KAAK;AACH,cAAUA,MAAE,MAAMA,MAAE,QAAQ,CAAC;AAC7B;GACF,KAAK;AACH,cAAUA,MAAE,MAAMA,MAAE,QAAQ,CAAC;AAC7B;GACF,QACE,WAAUA,MAAE,QAAQ;;AAGxB,WAAS,OAAO,aAAa,QAAQ,UAAU,GAAG;;CAGpD,MAAM,YAAYA,MAAE,OAAO,SAAS;CAEpC,MAAM,UAAU,KAAK,YAAY,MAAM,GAAG,IAAK;CAE/C,IAAIC;AAEJ,KAAI,gBAAgB;AAElB,WAAS,kBAAkB,gBAAgB,MAAM,QAAQ;AAGzD,MAAI,CAAC,eAAe,SAAS,cAAc,CACzC,WAAU,iBAAiB;OAG7B,UAAS;;SAEJ,KAAK,MAAM;OACb,KAAK,IAAI;;;EAGd,QAAQ;;;EAGR,OAAO,QAAQ,OAAO,CACrB,KAAK,CAAC,KAAK,UAAU,KAAK,IAAI,IAAI,KAAK,GAAG,CAC1C,KAAK,KAAK;AAGX,QAAO,SAAS,aAAgB,QAAQ,UAA0B;;;;;AAMpE,eAAe,UAAU,SAAiB,UAAwC;CAChF,MAAM,SAAS;;EAEf;AAGA,SADe,MAAM,SAAS,aAAa,QAAQ,cAAc,EACnD;;;;;AAMhB,eAAe,YAAY,SAAiB,UAA0C;CACpF,MAAM,SAAS;;EAEf;AAGA,SADe,MAAM,SAAS,aAAa,QAAQ,WAAW,EAChD;;;;;AAMhB,eAAe,gBAAgB,SAAiB,UAAmD;CACjG,MAAM,SAAS;;EAEf;AAEA,QAAO,SAAS,aAAa,QAAQ,eAAe;;;;;AAMtD,eAAe,SACb,SACA,UACsD;CACtD,MAAM,SAAS;;;;;;;;;EASf;AAEA,QAAO,SAAS,aAAa,QAAQ,eAAe"}