scrapex 0.5.2 → 1.0.0-alpha.1

package/dist/llm/index.mjs

@@ -0,0 +1,310 @@
+import { a as EntitiesSchema, c as ScrapeError, i as ClassifySchema, n as enhance, o as SummarySchema, r as extract, s as TagsSchema, t as ask } from "../enhancer-Q6CSc1gA.mjs";
+import { createRequire } from "node:module";
+
+//#region rolldown:runtime
+var __require = /* @__PURE__ */ createRequire(import.meta.url);
+
+//#endregion
+//#region src/llm/anthropic.ts
+const DEFAULT_MODEL$1 = "claude-3-5-haiku-20241022";
+const DEFAULT_MAX_TOKENS$1 = 1024;
+/**
+* Anthropic Claude provider
+*
+* Requires @anthropic-ai/sdk as a peer dependency.
+*
+* @example
+* ```ts
+* const provider = new AnthropicProvider({ apiKey: 'sk-...' });
+* const result = await scrape(url, { llm: provider, enhance: ['summarize'] });
+* ```
+*/
+var AnthropicProvider = class {
+	name = "anthropic";
+	client;
+	model;
+	constructor(config = {}) {
+		const apiKey = config.apiKey ?? process.env.ANTHROPIC_API_KEY;
+		if (!apiKey) throw new ScrapeError("Anthropic API key required. Set ANTHROPIC_API_KEY env var or pass apiKey in config.", "LLM_ERROR");
+		this.model = config.model ?? DEFAULT_MODEL$1;
+		try {
+			const { Anthropic } = __require("@anthropic-ai/sdk");
+			this.client = new Anthropic({
+				apiKey,
+				baseURL: config.baseUrl
+			});
+		} catch {
+			throw new ScrapeError("@anthropic-ai/sdk is required for Anthropic provider. Install with: npm install @anthropic-ai/sdk", "LLM_ERROR");
+		}
+	}
+	async complete(prompt, options = {}) {
+		try {
+			const content = (await this.client.messages.create({
+				model: this.model,
+				max_tokens: options.maxTokens ?? DEFAULT_MAX_TOKENS$1,
+				messages: [{
+					role: "user",
+					content: prompt
+				}],
+				system: options.systemPrompt,
+				temperature: options.temperature
+			})).content[0];
+			if (content?.type === "text" && content.text) return content.text;
+			throw new ScrapeError("Unexpected or empty response from Anthropic", "LLM_ERROR");
+		} catch (error) {
+			if (error instanceof ScrapeError) throw error;
+			throw new ScrapeError(`Anthropic API error: ${error instanceof Error ? error.message : String(error)}`, "LLM_ERROR", void 0, error instanceof Error ? error : void 0);
+		}
+	}
+	async completeJSON(prompt, schema, options = {}) {
+		const jsonPrompt = `${prompt}
+
+Respond ONLY with valid JSON matching this schema:
+${JSON.stringify(zodToJsonSchema$1(schema), null, 2)}
+
+Do not include any explanation or markdown formatting. Just the JSON object.`;
+		const response = await this.complete(jsonPrompt, {
+			...options,
+			systemPrompt: options.systemPrompt ?? "You are a helpful assistant that responds only with valid JSON."
+		});
+		try {
+			const jsonMatch = response.match(/\{[\s\S]*\}/);
+			if (!jsonMatch) throw new Error("No JSON object found in response");
+			const parsed = JSON.parse(jsonMatch[0]);
+			return schema.parse(parsed);
+		} catch (error) {
+			throw new ScrapeError(`Failed to parse LLM response as JSON: ${error instanceof Error ? error.message : String(error)}`, "VALIDATION_ERROR", void 0, error instanceof Error ? error : void 0);
+		}
+	}
+};
+/**
+* Convert a Zod schema to a simple JSON Schema representation
+* (simplified version for prompt engineering)
+*/
+function zodToJsonSchema$1(schema) {
+	const def = schema._def;
+	switch (def.typeName) {
+		case "ZodObject": {
+			const shape = schema.shape;
+			const properties = {};
+			for (const [key, value] of Object.entries(shape)) properties[key] = zodToJsonSchema$1(value);
+			return {
+				type: "object",
+				properties
+			};
+		}
+		case "ZodArray": return {
+			type: "array",
+			items: zodToJsonSchema$1(def.type)
+		};
+		case "ZodString": return { type: "string" };
+		case "ZodNumber": return { type: "number" };
+		case "ZodBoolean": return { type: "boolean" };
+		case "ZodEnum": return {
+			type: "string",
+			enum: def.values
+		};
+		default: return { type: "string" };
+	}
+}
+
+//#endregion
+//#region src/llm/openai.ts
+const DEFAULT_MODEL = "gpt-4o-mini";
+const DEFAULT_MAX_TOKENS = 1024;
+const DEFAULT_BASE_URL = "https://api.openai.com/v1";
+/**
+* OpenAI-compatible provider
+*
+* Works with:
+* - OpenAI API
+* - Ollama (http://localhost:11434/v1)
+* - LM Studio (http://localhost:1234/v1)
+* - LocalAI
+* - vLLM
+* - Any OpenAI-compatible API
+*
+* Requires `openai` as a peer dependency.
+*
+* @example
+* ```ts
+* // OpenAI
+* const provider = new OpenAIProvider({ apiKey: 'sk-...' });
+*
+* // Ollama
+* const provider = new OpenAIProvider({
+*   baseUrl: 'http://localhost:11434/v1',
+*   model: 'llama3.2',
+*   apiKey: 'ollama' // Ollama doesn't require a real key
+* });
+*
+* // LM Studio
+* const provider = new OpenAIProvider({
+*   baseUrl: 'http://localhost:1234/v1',
+*   model: 'local-model',
+*   apiKey: 'lm-studio'
+* });
+* ```
+*/
+var OpenAIProvider = class {
+	name = "openai";
+	client;
+	model;
+	constructor(config = {}) {
+		const apiKey = config.apiKey ?? process.env.OPENAI_API_KEY;
+		const baseUrl = config.baseUrl ?? DEFAULT_BASE_URL;
+		if (!apiKey && baseUrl === DEFAULT_BASE_URL) throw new ScrapeError("OpenAI API key required. Set OPENAI_API_KEY env var or pass apiKey in config.", "LLM_ERROR");
+		this.model = config.model ?? DEFAULT_MODEL;
+		try {
+			const { OpenAI } = __require("openai");
+			this.client = new OpenAI({
+				apiKey: apiKey ?? "local",
+				baseURL: baseUrl
+			});
+		} catch {
+			throw new ScrapeError("openai package is required for OpenAI provider. Install with: npm install openai", "LLM_ERROR");
+		}
+	}
+	async complete(prompt, options = {}) {
+		try {
+			const client = this.client;
+			const messages = [];
+			if (options.systemPrompt) messages.push({
+				role: "system",
+				content: options.systemPrompt
+			});
+			messages.push({
+				role: "user",
+				content: prompt
+			});
+			const content = (await client.chat.completions.create({
+				model: this.model,
+				max_tokens: options.maxTokens ?? DEFAULT_MAX_TOKENS,
+				messages,
+				temperature: options.temperature
+			})).choices[0]?.message?.content;
+			if (!content) throw new ScrapeError("Empty response from OpenAI", "LLM_ERROR");
+			return content;
+		} catch (error) {
+			if (error instanceof ScrapeError) throw error;
+			throw new ScrapeError(`OpenAI API error: ${error instanceof Error ? error.message : String(error)}`, "LLM_ERROR", void 0, error instanceof Error ? error : void 0);
+		}
+	}
+	async completeJSON(prompt, schema, options = {}) {
+		const client = this.client;
+		try {
+			const messages = [{
+				role: "system",
+				content: options.systemPrompt ?? "You are a helpful assistant that extracts information from content."
+			}, {
+				role: "user",
+				content: prompt
+			}];
+			const content = (await client.chat.completions.create({
+				model: this.model,
+				max_tokens: options.maxTokens ?? DEFAULT_MAX_TOKENS,
+				messages,
+				temperature: options.temperature,
+				response_format: { type: "json_object" }
+			})).choices[0]?.message?.content;
+			if (!content) throw new ScrapeError("Empty response from OpenAI", "LLM_ERROR");
+			const parsed = JSON.parse(content);
+			return schema.parse(parsed);
+		} catch (error) {
+			if (error instanceof ScrapeError) throw error;
+			const jsonPrompt = `${prompt}
+
+Respond ONLY with valid JSON matching this schema:
+${JSON.stringify(zodToJsonSchema(schema), null, 2)}
+
+Do not include any explanation or markdown formatting. Just the JSON object.`;
+			const response = await this.complete(jsonPrompt, {
+				...options,
+				systemPrompt: "You respond only with valid JSON."
+			});
+			try {
+				const jsonMatch = response.match(/\{[\s\S]*\}/);
+				if (!jsonMatch) throw new Error("No JSON object found in response");
+				const parsed = JSON.parse(jsonMatch[0]);
+				return schema.parse(parsed);
+			} catch (parseError) {
+				throw new ScrapeError(`Failed to parse LLM response as JSON: ${parseError instanceof Error ? parseError.message : String(parseError)}`, "VALIDATION_ERROR", void 0, parseError instanceof Error ? parseError : void 0);
+			}
+		}
+	}
+};
+/**
+* Convert a Zod schema to JSON Schema for structured outputs
+*/
+function zodToJsonSchema(schema) {
+	const def = schema._def;
+	switch (def.typeName) {
+		case "ZodObject": {
+			const shape = schema.shape;
+			const properties = {};
+			const required = [];
+			for (const [key, value] of Object.entries(shape)) {
+				properties[key] = zodToJsonSchema(value);
+				if (value._def.typeName !== "ZodOptional") required.push(key);
+			}
+			return {
+				type: "object",
+				properties,
+				required
+			};
+		}
+		case "ZodArray": return {
+			type: "array",
+			items: zodToJsonSchema(def.type)
+		};
+		case "ZodString": return { type: "string" };
+		case "ZodNumber": return { type: "number" };
+		case "ZodBoolean": return { type: "boolean" };
+		case "ZodEnum": return {
+			type: "string",
+			enum: def.values
+		};
+		case "ZodOptional": return zodToJsonSchema(def.innerType);
+		default: return { type: "string" };
+	}
+}
+/**
+* Create an OpenAI provider with default settings
+*/
+function createOpenAI(config) {
+	return new OpenAIProvider(config);
+}
+/**
+* Create an Ollama provider
+*
+* @example
+* ```ts
+* const provider = createOllama({ model: 'llama3.2' });
+* ```
+*/
+function createOllama(config = { model: "llama3.2" }) {
+	return new OpenAIProvider({
+		baseUrl: `http://localhost:${config.port ?? 11434}/v1`,
+		model: config.model,
+		apiKey: "ollama"
+	});
+}
+/**
+* Create an LM Studio provider
+*
+* @example
+* ```ts
+* const provider = createLMStudio({ model: 'local-model' });
+* ```
+*/
+function createLMStudio(config = { model: "local-model" }) {
+	return new OpenAIProvider({
+		baseUrl: `http://localhost:${config.port ?? 1234}/v1`,
+		model: config.model,
+		apiKey: "lm-studio"
+	});
+}
+
+//#endregion
+export { AnthropicProvider, ClassifySchema, EntitiesSchema, OpenAIProvider, SummarySchema, TagsSchema, ask, createLMStudio, createOllama, createOpenAI, enhance, extract };
+//# sourceMappingURL=index.mjs.map

package/dist/llm/index.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.mjs","names":["DEFAULT_MODEL","DEFAULT_MAX_TOKENS","zodToJsonSchema","properties: Record<string, object>","messages: Array<{ role: 'system' | 'user'; content: string }>","properties: Record<string, object>","required: string[]"],"sources":["../../src/llm/anthropic.ts","../../src/llm/openai.ts"],"sourcesContent":["import type { z } from 'zod';\nimport { ScrapeError } from '@/core/errors.js';\nimport type { AnthropicConfig, CompletionOptions, LLMProvider } from './types.js';\n\nconst DEFAULT_MODEL = 'claude-3-5-haiku-20241022';\nconst DEFAULT_MAX_TOKENS = 1024;\n\n/**\n * Anthropic Claude provider\n *\n * Requires @anthropic-ai/sdk as a peer dependency.\n *\n * @example\n * ```ts\n * const provider = new AnthropicProvider({ apiKey: 'sk-...' });\n * const result = await scrape(url, { llm: provider, enhance: ['summarize'] });\n * ```\n */\nexport class AnthropicProvider implements LLMProvider {\n  readonly name = 'anthropic';\n  private client: unknown;\n  private model: string;\n\n  constructor(config: AnthropicConfig = {}) {\n    const apiKey = config.apiKey ?? process.env.ANTHROPIC_API_KEY;\n    if (!apiKey) {\n      throw new ScrapeError(\n        'Anthropic API key required. Set ANTHROPIC_API_KEY env var or pass apiKey in config.',\n        'LLM_ERROR'\n      );\n    }\n\n    this.model = config.model ?? DEFAULT_MODEL;\n\n    // Dynamic import to avoid requiring the SDK if not used\n    try {\n      // eslint-disable-next-line @typescript-eslint/no-require-imports\n      const { Anthropic } = require('@anthropic-ai/sdk') as typeof import('@anthropic-ai/sdk');\n      this.client = new Anthropic({\n        apiKey,\n        baseURL: config.baseUrl,\n      });\n    } catch {\n      throw new ScrapeError(\n        '@anthropic-ai/sdk is required for Anthropic provider. Install with: npm install @anthropic-ai/sdk',\n        'LLM_ERROR'\n      );\n    }\n  }\n\n  async complete(prompt: string, options: CompletionOptions = {}): Promise<string> {\n    try {\n      const client = this.client as import('@anthropic-ai/sdk').Anthropic;\n      const response = await client.messages.create({\n        model: this.model,\n        max_tokens: options.maxTokens ?? DEFAULT_MAX_TOKENS,\n        messages: [{ role: 'user', content: prompt }],\n        system: options.systemPrompt,\n        temperature: options.temperature,\n      });\n\n      const content = response.content[0];\n      if (content?.type === 'text' && content.text) {\n        return content.text;\n      }\n\n      throw new ScrapeError('Unexpected or empty response from Anthropic', 'LLM_ERROR');\n    } catch (error) {\n      if (error instanceof ScrapeError) throw error;\n      throw new ScrapeError(\n        `Anthropic API error: ${error instanceof Error ? error.message : String(error)}`,\n        'LLM_ERROR',\n        undefined,\n        error instanceof Error ? error : undefined\n      );\n    }\n  }\n\n  async completeJSON<T>(\n    prompt: string,\n    schema: z.ZodType<T>,\n    options: CompletionOptions = {}\n  ): Promise<T> {\n    const jsonPrompt = `${prompt}\n\nRespond ONLY with valid JSON matching this schema:\n${JSON.stringify(zodToJsonSchema(schema), null, 2)}\n\nDo not include any explanation or markdown formatting. Just the JSON object.`;\n\n    const response = await this.complete(jsonPrompt, {\n      ...options,\n      systemPrompt:\n        options.systemPrompt ?? 'You are a helpful assistant that responds only with valid JSON.',\n    });\n\n    try {\n      // Try to extract JSON from the response\n      const jsonMatch = response.match(/\\{[\\s\\S]*\\}/);\n      if (!jsonMatch) {\n        throw new Error('No JSON object found in response');\n      }\n\n      const parsed = JSON.parse(jsonMatch[0]);\n      return schema.parse(parsed);\n    } catch (error) {\n      throw new ScrapeError(\n        `Failed to parse LLM response as JSON: ${error instanceof Error ? error.message : String(error)}`,\n        'VALIDATION_ERROR',\n        undefined,\n        error instanceof Error ? error : undefined\n      );\n    }\n  }\n}\n\n/**\n * Convert a Zod schema to a simple JSON Schema representation\n * (simplified version for prompt engineering)\n */\nfunction zodToJsonSchema(schema: z.ZodType<unknown>): object {\n  const def = (schema as z.ZodType<unknown> & { _def: { typeName: string } })._def;\n\n  switch (def.typeName) {\n    case 'ZodObject': {\n      const shape = (schema as z.ZodObject<z.ZodRawShape>).shape;\n      const properties: Record<string, object> = {};\n      for (const [key, value] of Object.entries(shape)) {\n        properties[key] = zodToJsonSchema(value as z.ZodType<unknown>);\n      }\n      return { type: 'object', properties };\n    }\n    case 'ZodArray': {\n      const arrayDef = def as unknown as { type: z.ZodType<unknown> };\n      return { type: 'array', items: zodToJsonSchema(arrayDef.type) };\n    }\n    case 'ZodString':\n      return { type: 'string' };\n    case 'ZodNumber':\n      return { type: 'number' };\n    case 'ZodBoolean':\n      return { type: 'boolean' };\n    case 'ZodEnum': {\n      const enumDef = def as unknown as { values: string[] };\n      return { type: 'string', enum: enumDef.values };\n    }\n    default:\n      return { type: 'string' };\n  }\n}\n","import type { z } from 'zod';\nimport { ScrapeError } from '@/core/errors.js';\nimport type { CompletionOptions, LLMProvider, OpenAICompatibleConfig } from './types.js';\n\nconst DEFAULT_MODEL = 'gpt-4o-mini';\nconst DEFAULT_MAX_TOKENS = 1024;\nconst DEFAULT_BASE_URL = 'https://api.openai.com/v1';\n\n/**\n * OpenAI-compatible provider\n *\n * Works with:\n * - OpenAI API\n * - Ollama (http://localhost:11434/v1)\n * - LM Studio (http://localhost:1234/v1)\n * - LocalAI\n * - vLLM\n * - Any OpenAI-compatible API\n *\n * Requires `openai` as a peer dependency.\n *\n * @example\n * ```ts\n * // OpenAI\n * const provider = new OpenAIProvider({ apiKey: 'sk-...' });\n *\n * // Ollama\n * const provider = new OpenAIProvider({\n *   baseUrl: 'http://localhost:11434/v1',\n *   model: 'llama3.2',\n *   apiKey: 'ollama' // Ollama doesn't require a real key\n * });\n *\n * // LM Studio\n * const provider = new OpenAIProvider({\n *   baseUrl: 'http://localhost:1234/v1',\n *   model: 'local-model',\n *   apiKey: 'lm-studio'\n * });\n * ```\n */\nexport class OpenAIProvider implements LLMProvider {\n  readonly name = 'openai';\n  private client: unknown;\n  private model: string;\n\n  constructor(config: OpenAICompatibleConfig = {}) {\n    const apiKey = config.apiKey ?? process.env.OPENAI_API_KEY;\n    const baseUrl = config.baseUrl ?? DEFAULT_BASE_URL;\n\n    // Only require API key for OpenAI (not for local providers)\n    if (!apiKey && baseUrl === DEFAULT_BASE_URL) {\n      throw new ScrapeError(\n        'OpenAI API key required. Set OPENAI_API_KEY env var or pass apiKey in config.',\n        'LLM_ERROR'\n      );\n    }\n\n    this.model = config.model ?? DEFAULT_MODEL;\n\n    // Dynamic import to avoid requiring the SDK if not used\n    try {\n      // eslint-disable-next-line @typescript-eslint/no-require-imports\n      const { OpenAI } = require('openai') as typeof import('openai');\n      this.client = new OpenAI({\n        apiKey: apiKey ?? 'local', // Use 'local' as placeholder for local providers\n        baseURL: baseUrl,\n      });\n    } catch {\n      throw new ScrapeError(\n        'openai package is required for OpenAI provider. Install with: npm install openai',\n        'LLM_ERROR'\n      );\n    }\n  }\n\n  async complete(prompt: string, options: CompletionOptions = {}): Promise<string> {\n    try {\n      const client = this.client as import('openai').OpenAI;\n      const messages: Array<{ role: 'system' | 'user'; content: string }> = [];\n\n      if (options.systemPrompt) {\n        messages.push({ role: 'system', content: options.systemPrompt });\n      }\n      messages.push({ role: 'user', content: prompt });\n\n      const response = await client.chat.completions.create({\n        model: this.model,\n        max_tokens: options.maxTokens ?? DEFAULT_MAX_TOKENS,\n        messages,\n        temperature: options.temperature,\n      });\n\n      const content = response.choices[0]?.message?.content;\n      if (!content) {\n        throw new ScrapeError('Empty response from OpenAI', 'LLM_ERROR');\n      }\n\n      return content;\n    } catch (error) {\n      if (error instanceof ScrapeError) throw error;\n      throw new ScrapeError(\n        `OpenAI API error: ${error instanceof Error ? error.message : String(error)}`,\n        'LLM_ERROR',\n        undefined,\n        error instanceof Error ? error : undefined\n      );\n    }\n  }\n\n  async completeJSON<T>(\n    prompt: string,\n    schema: z.ZodType<T>,\n    options: CompletionOptions = {}\n  ): Promise<T> {\n    const client = this.client as import('openai').OpenAI;\n\n    try {\n      // Use JSON mode for structured outputs\n      const messages: Array<{ role: 'system' | 'user'; content: string }> = [\n        {\n          role: 'system',\n          content:\n            options.systemPrompt ??\n            'You are a helpful assistant that extracts information from content.',\n        },\n        { role: 'user', content: prompt },\n      ];\n\n      const response = await client.chat.completions.create({\n        model: this.model,\n        max_tokens: options.maxTokens ?? DEFAULT_MAX_TOKENS,\n        messages,\n        temperature: options.temperature,\n        response_format: { type: 'json_object' },\n      });\n\n      const content = response.choices[0]?.message?.content;\n      if (!content) {\n        throw new ScrapeError('Empty response from OpenAI', 'LLM_ERROR');\n      }\n\n      const parsed = JSON.parse(content);\n      return schema.parse(parsed);\n    } catch (error) {\n      // Fallback to regular completion with JSON instruction\n      if (error instanceof ScrapeError) throw error;\n\n      // If structured output failed, try regular completion\n      const jsonPrompt = `${prompt}\n\nRespond ONLY with valid JSON matching this schema:\n${JSON.stringify(zodToJsonSchema(schema), null, 2)}\n\nDo not include any explanation or markdown formatting. Just the JSON object.`;\n\n      const response = await this.complete(jsonPrompt, {\n        ...options,\n        systemPrompt: 'You respond only with valid JSON.',\n      });\n\n      try {\n        const jsonMatch = response.match(/\\{[\\s\\S]*\\}/);\n        if (!jsonMatch) {\n          throw new Error('No JSON object found in response');\n        }\n\n        const parsed = JSON.parse(jsonMatch[0]);\n        return schema.parse(parsed);\n      } catch (parseError) {\n        throw new ScrapeError(\n          `Failed to parse LLM response as JSON: ${parseError instanceof Error ? parseError.message : String(parseError)}`,\n          'VALIDATION_ERROR',\n          undefined,\n          parseError instanceof Error ? parseError : undefined\n        );\n      }\n    }\n  }\n}\n\n/**\n * Convert a Zod schema to JSON Schema for structured outputs\n */\nfunction zodToJsonSchema(schema: z.ZodType<unknown>): object {\n  const def = (schema as z.ZodType<unknown> & { _def: { typeName: string } })._def;\n\n  switch (def.typeName) {\n    case 'ZodObject': {\n      const shape = (schema as z.ZodObject<z.ZodRawShape>).shape;\n      const properties: Record<string, object> = {};\n      const required: string[] = [];\n\n      for (const [key, value] of Object.entries(shape)) {\n        properties[key] = zodToJsonSchema(value as z.ZodType<unknown>);\n        // Assume all fields are required unless wrapped in ZodOptional\n        const valueDef = (value as z.ZodType<unknown> & { _def: { typeName: string } })._def;\n        if (valueDef.typeName !== 'ZodOptional') {\n          required.push(key);\n        }\n      }\n      return { type: 'object', properties, required };\n    }\n    case 'ZodArray': {\n      const arrayDef = def as unknown as { type: z.ZodType<unknown> };\n      return { type: 'array', items: zodToJsonSchema(arrayDef.type) };\n    }\n    case 'ZodString':\n      return { type: 'string' };\n    case 'ZodNumber':\n      return { type: 'number' };\n    case 'ZodBoolean':\n      return { type: 'boolean' };\n    case 'ZodEnum': {\n      const enumDef = def as unknown as { values: string[] };\n      return { type: 'string', enum: enumDef.values };\n    }\n    case 'ZodOptional': {\n      const optionalDef = def as unknown as { innerType: z.ZodType<unknown> };\n      return zodToJsonSchema(optionalDef.innerType);\n    }\n    default:\n      return { type: 'string' };\n  }\n}\n\n// Convenience factory functions\n\n/**\n * Create an OpenAI provider with default settings\n */\nexport function createOpenAI(config?: OpenAICompatibleConfig): OpenAIProvider {\n  return new OpenAIProvider(config);\n}\n\n/**\n * Create an Ollama provider\n *\n * @example\n * ```ts\n * const provider = createOllama({ model: 'llama3.2' });\n * ```\n */\nexport function createOllama(\n  config: { model: string; port?: number } = { model: 'llama3.2' }\n): OpenAIProvider {\n  return new OpenAIProvider({\n    baseUrl: `http://localhost:${config.port ?? 11434}/v1`,\n    model: config.model,\n    apiKey: 'ollama',\n  });\n}\n\n/**\n * Create an LM Studio provider\n *\n * @example\n * ```ts\n * const provider = createLMStudio({ model: 'local-model' });\n * ```\n */\nexport function createLMStudio(\n  config: { model: string; port?: number } = { model: 'local-model' }\n): OpenAIProvider {\n  return new OpenAIProvider({\n    baseUrl: `http://localhost:${config.port ?? 1234}/v1`,\n    model: config.model,\n    apiKey: 'lm-studio',\n  });\n}\n"],"mappings":";;;;;;;;AAIA,MAAMA,kBAAgB;AACtB,MAAMC,uBAAqB;;;;;;;;;;;;AAa3B,IAAa,oBAAb,MAAsD;CACpD,AAAS,OAAO;CAChB,AAAQ;CACR,AAAQ;CAER,YAAY,SAA0B,EAAE,EAAE;EACxC,MAAM,SAAS,OAAO,UAAU,QAAQ,IAAI;AAC5C,MAAI,CAAC,OACH,OAAM,IAAI,YACR,uFACA,YACD;AAGH,OAAK,QAAQ,OAAO,SAASD;AAG7B,MAAI;GAEF,MAAM,EAAE,wBAAsB,oBAAoB;AAClD,QAAK,SAAS,IAAI,UAAU;IAC1B;IACA,SAAS,OAAO;IACjB,CAAC;UACI;AACN,SAAM,IAAI,YACR,qGACA,YACD;;;CAIL,MAAM,SAAS,QAAgB,UAA6B,EAAE,EAAmB;AAC/E,MAAI;GAUF,MAAM,WARW,MADF,KAAK,OACU,SAAS,OAAO;IAC5C,OAAO,KAAK;IACZ,YAAY,QAAQ,aAAaC;IACjC,UAAU,CAAC;KAAE,MAAM;KAAQ,SAAS;KAAQ,CAAC;IAC7C,QAAQ,QAAQ;IAChB,aAAa,QAAQ;IACtB,CAAC,EAEuB,QAAQ;AACjC,OAAI,SAAS,SAAS,UAAU,QAAQ,KACtC,QAAO,QAAQ;AAGjB,SAAM,IAAI,YAAY,+CAA+C,YAAY;WAC1E,OAAO;AACd,OAAI,iBAAiB,YAAa,OAAM;AACxC,SAAM,IAAI,YACR,wBAAwB,iBAAiB,QAAQ,MAAM,UAAU,OAAO,MAAM,IAC9E,aACA,QACA,iBAAiB,QAAQ,QAAQ,OAClC;;;CAIL,MAAM,aACJ,QACA,QACA,UAA6B,EAAE,EACnB;EACZ,MAAM,aAAa,GAAG,OAAO;;;EAG/B,KAAK,UAAUC,kBAAgB,OAAO,EAAE,MAAM,EAAE,CAAC;;;EAI/C,MAAM,WAAW,MAAM,KAAK,SAAS,YAAY;GAC/C,GAAG;GACH,cACE,QAAQ,gBAAgB;GAC3B,CAAC;AAEF,MAAI;GAEF,MAAM,YAAY,SAAS,MAAM,cAAc;AAC/C,OAAI,CAAC,UACH,OAAM,IAAI,MAAM,mCAAmC;GAGrD,MAAM,SAAS,KAAK,MAAM,UAAU,GAAG;AACvC,UAAO,OAAO,MAAM,OAAO;WACpB,OAAO;AACd,SAAM,IAAI,YACR,yCAAyC,iBAAiB,QAAQ,MAAM,UAAU,OAAO,MAAM,IAC/F,oBACA,QACA,iBAAiB,QAAQ,QAAQ,OAClC;;;;;;;;AASP,SAASA,kBAAgB,QAAoC;CAC3D,MAAM,MAAO,OAA+D;AAE5E,SAAQ,IAAI,UAAZ;EACE,KAAK,aAAa;GAChB,MAAM,QAAS,OAAsC;GACrD,MAAMC,aAAqC,EAAE;AAC7C,QAAK,MAAM,CAAC,KAAK,UAAU,OAAO,QAAQ,MAAM,CAC9C,YAAW,OAAOD,kBAAgB,MAA4B;AAEhE,UAAO;IAAE,MAAM;IAAU;IAAY;;EAEvC,KAAK,WAEH,QAAO;GAAE,MAAM;GAAS,OAAOA,kBADd,IACuC,KAAK;GAAE;EAEjE,KAAK,YACH,QAAO,EAAE,MAAM,UAAU;EAC3B,KAAK,YACH,QAAO,EAAE,MAAM,UAAU;EAC3B,KAAK,aACH,QAAO,EAAE,MAAM,WAAW;EAC5B,KAAK,UAEH,QAAO;GAAE,MAAM;GAAU,MADT,IACuB;GAAQ;EAEjD,QACE,QAAO,EAAE,MAAM,UAAU;;;;;;AC/I/B,MAAM,gBAAgB;AACtB,MAAM,qBAAqB;AAC3B,MAAM,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAmCzB,IAAa,iBAAb,MAAmD;CACjD,AAAS,OAAO;CAChB,AAAQ;CACR,AAAQ;CAER,YAAY,SAAiC,EAAE,EAAE;EAC/C,MAAM,SAAS,OAAO,UAAU,QAAQ,IAAI;EAC5C,MAAM,UAAU,OAAO,WAAW;AAGlC,MAAI,CAAC,UAAU,YAAY,iBACzB,OAAM,IAAI,YACR,iFACA,YACD;AAGH,OAAK,QAAQ,OAAO,SAAS;AAG7B,MAAI;GAEF,MAAM,EAAE,qBAAmB,SAAS;AACpC,QAAK,SAAS,IAAI,OAAO;IACvB,QAAQ,UAAU;IAClB,SAAS;IACV,CAAC;UACI;AACN,SAAM,IAAI,YACR,oFACA,YACD;;;CAIL,MAAM,SAAS,QAAgB,UAA6B,EAAE,EAAmB;AAC/E,MAAI;GACF,MAAM,SAAS,KAAK;GACpB,MAAME,WAAgE,EAAE;AAExE,OAAI,QAAQ,aACV,UAAS,KAAK;IAAE,MAAM;IAAU,SAAS,QAAQ;IAAc,CAAC;AAElE,YAAS,KAAK;IAAE,MAAM;IAAQ,SAAS;IAAQ,CAAC;GAShD,MAAM,WAPW,MAAM,OAAO,KAAK,YAAY,OAAO;IACpD,OAAO,KAAK;IACZ,YAAY,QAAQ,aAAa;IACjC;IACA,aAAa,QAAQ;IACtB,CAAC,EAEuB,QAAQ,IAAI,SAAS;AAC9C,OAAI,CAAC,QACH,OAAM,IAAI,YAAY,8BAA8B,YAAY;AAGlE,UAAO;WACA,OAAO;AACd,OAAI,iBAAiB,YAAa,OAAM;AACxC,SAAM,IAAI,YACR,qBAAqB,iBAAiB,QAAQ,MAAM,UAAU,OAAO,MAAM,IAC3E,aACA,QACA,iBAAiB,QAAQ,QAAQ,OAClC;;;CAIL,MAAM,aACJ,QACA,QACA,UAA6B,EAAE,EACnB;EACZ,MAAM,SAAS,KAAK;AAEpB,MAAI;GAEF,MAAMA,WAAgE,CACpE;IACE,MAAM;IACN,SACE,QAAQ,gBACR;IACH,EACD;IAAE,MAAM;IAAQ,SAAS;IAAQ,CAClC;GAUD,MAAM,WARW,MAAM,OAAO,KAAK,YAAY,OAAO;IACpD,OAAO,KAAK;IACZ,YAAY,QAAQ,aAAa;IACjC;IACA,aAAa,QAAQ;IACrB,iBAAiB,EAAE,MAAM,eAAe;IACzC,CAAC,EAEuB,QAAQ,IAAI,SAAS;AAC9C,OAAI,CAAC,QACH,OAAM,IAAI,YAAY,8BAA8B,YAAY;GAGlE,MAAM,SAAS,KAAK,MAAM,QAAQ;AAClC,UAAO,OAAO,MAAM,OAAO;WACpB,OAAO;AAEd,OAAI,iBAAiB,YAAa,OAAM;GAGxC,MAAM,aAAa,GAAG,OAAO;;;EAGjC,KAAK,UAAU,gBAAgB,OAAO,EAAE,MAAM,EAAE,CAAC;;;GAI7C,MAAM,WAAW,MAAM,KAAK,SAAS,YAAY;IAC/C,GAAG;IACH,cAAc;IACf,CAAC;AAEF,OAAI;IACF,MAAM,YAAY,SAAS,MAAM,cAAc;AAC/C,QAAI,CAAC,UACH,OAAM,IAAI,MAAM,mCAAmC;IAGrD,MAAM,SAAS,KAAK,MAAM,UAAU,GAAG;AACvC,WAAO,OAAO,MAAM,OAAO;YACpB,YAAY;AACnB,UAAM,IAAI,YACR,yCAAyC,sBAAsB,QAAQ,WAAW,UAAU,OAAO,WAAW,IAC9G,oBACA,QACA,sBAAsB,QAAQ,aAAa,OAC5C;;;;;;;;AAST,SAAS,gBAAgB,QAAoC;CAC3D,MAAM,MAAO,OAA+D;AAE5E,SAAQ,IAAI,UAAZ;EACE,KAAK,aAAa;GAChB,MAAM,QAAS,OAAsC;GACrD,MAAMC,aAAqC,EAAE;GAC7C,MAAMC,WAAqB,EAAE;AAE7B,QAAK,MAAM,CAAC,KAAK,UAAU,OAAO,QAAQ,MAAM,EAAE;AAChD,eAAW,OAAO,gBAAgB,MAA4B;AAG9D,QADkB,MAA8D,KACnE,aAAa,cACxB,UAAS,KAAK,IAAI;;AAGtB,UAAO;IAAE,MAAM;IAAU;IAAY;IAAU;;EAEjD,KAAK,WAEH,QAAO;GAAE,MAAM;GAAS,OAAO,gBADd,IACuC,KAAK;GAAE;EAEjE,KAAK,YACH,QAAO,EAAE,MAAM,UAAU;EAC3B,KAAK,YACH,QAAO,EAAE,MAAM,UAAU;EAC3B,KAAK,aACH,QAAO,EAAE,MAAM,WAAW;EAC5B,KAAK,UAEH,QAAO;GAAE,MAAM;GAAU,MADT,IACuB;GAAQ;EAEjD,KAAK,cAEH,QAAO,gBADa,IACe,UAAU;EAE/C,QACE,QAAO,EAAE,MAAM,UAAU;;;;;;AAS/B,SAAgB,aAAa,QAAiD;AAC5E,QAAO,IAAI,eAAe,OAAO;;;;;;;;;;AAWnC,SAAgB,aACd,SAA2C,EAAE,OAAO,YAAY,EAChD;AAChB,QAAO,IAAI,eAAe;EACxB,SAAS,oBAAoB,OAAO,QAAQ,MAAM;EAClD,OAAO,OAAO;EACd,QAAQ;EACT,CAAC;;;;;;;;;;AAWJ,SAAgB,eACd,SAA2C,EAAE,OAAO,eAAe,EACnD;AAChB,QAAO,IAAI,eAAe;EACxB,SAAS,oBAAoB,OAAO,QAAQ,KAAK;EACjD,OAAO,OAAO;EACd,QAAQ;EACT,CAAC"}

package/dist/parsers/index.cjs

@@ -0,0 +1,200 @@
+const require_index = require('../index.cjs');
+let mdast_util_from_markdown = require("mdast-util-from-markdown");
+let mdast_util_to_string = require("mdast-util-to-string");
+let unist_util_visit = require("unist-util-visit");
+
+//#region src/parsers/github.ts
+/**
+* GitHub-specific utilities for parsing repositories.
+*/
+/**
+* Check if a URL is a GitHub repository
+*/
+function isGitHubRepo(url) {
+	return /^https?:\/\/(www\.)?github\.com\/[^/]+\/[^/]+\/?$/.test(url);
+}
+/**
+* Extract GitHub repo info from URL
+*/
+function parseGitHubUrl(url) {
+	const match = url.match(/github\.com\/([^/]+)\/([^/]+)/);
+	if (!match || !match[1] || !match[2]) return null;
+	return {
+		owner: match[1],
+		repo: match[2].replace(/\.git$/, "")
+	};
+}
+/**
+* Convert a GitHub repo URL to raw content URL
+*/
+function toRawUrl(url, branch = "main", file = "README.md") {
+	const info = parseGitHubUrl(url);
+	if (!info) return url;
+	return `https://raw.githubusercontent.com/${info.owner}/${info.repo}/${branch}/${file}`;
+}
+/**
+* Fetch GitHub API metadata for a repository
+* Note: This is a placeholder - actual implementation would need GitHub API access
+*/
+async function fetchRepoMeta(owner, repo, _token) {
+	return {
+		repoOwner: owner,
+		repoName: repo
+	};
+}
+/**
+* Group links by their category/section
+*/
+function groupByCategory(links) {
+	const groups = /* @__PURE__ */ new Map();
+	for (const link of links) {
+		const category = link.context || "Uncategorized";
+		const existing = groups.get(category) || [];
+		existing.push(link);
+		groups.set(category, existing);
+	}
+	return groups;
+}
+
+//#endregion
+//#region src/parsers/markdown.ts
+/**
+* Generic Markdown parser.
+* Extracts structure, links, and code blocks from markdown content.
+*
+* @example
+* ```ts
+* const parser = new MarkdownParser();
+* const result = parser.parse(markdownContent);
+* console.log(result.data.sections);
+* console.log(result.data.links);
+* ```
+*/
+var MarkdownParser = class {
+	name = "markdown";
+	canParse(content) {
+		return content.includes("# ") || content.includes("## ") || content.includes("- [") || content.includes("* [") || content.includes("```");
+	}
+	parse(content) {
+		const tree = (0, mdast_util_from_markdown.fromMarkdown)(content);
+		const sections = [];
+		const allLinks = [];
+		const codeBlocks = [];
+		let frontmatter;
+		if (content.startsWith("---")) {
+			const endIndex = content.indexOf("---", 3);
+			if (endIndex !== -1) {
+				const frontmatterContent = content.slice(3, endIndex).trim();
+				frontmatter = this.parseFrontmatter(frontmatterContent);
+			}
+		}
+		let currentSection = null;
+		(0, unist_util_visit.visit)(tree, (node) => {
+			if (node.type === "heading") {
+				const heading = node;
+				const title = (0, mdast_util_to_string.toString)(heading);
+				if (currentSection) sections.push(currentSection);
+				currentSection = {
+					level: heading.depth,
+					title,
+					content: "",
+					links: []
+				};
+			}
+			if (node.type === "link") {
+				const link = node;
+				const text = (0, mdast_util_to_string.toString)(link);
+				const linkData = {
+					url: link.url,
+					text,
+					title: link.title ?? void 0,
+					context: currentSection?.title
+				};
+				allLinks.push(linkData);
+				if (currentSection) currentSection.links.push(linkData);
+			}
+			if (node.type === "code") {
+				const code = node;
+				codeBlocks.push({
+					language: code.lang ?? void 0,
+					code: code.value,
+					meta: code.meta ?? void 0
+				});
+			}
+			if (currentSection && node.type === "paragraph") {
+				const text = (0, mdast_util_to_string.toString)(node);
+				currentSection.content += (currentSection.content ? "\n\n" : "") + text;
+			}
+		});
+		if (currentSection) sections.push(currentSection);
+		return { data: {
+			title: frontmatter?.title ?? sections.find((s) => s.level === 1)?.title,
+			description: frontmatter?.description ?? this.extractDescription(tree),
+			sections,
+			links: allLinks,
+			codeBlocks,
+			frontmatter
+		} };
+	}
+	parseFrontmatter(content) {
+		const result = {};
+		const lines = content.split("\n");
+		for (const line of lines) {
+			const colonIndex = line.indexOf(":");
+			if (colonIndex > 0) {
+				const key = line.slice(0, colonIndex).trim();
+				let value = line.slice(colonIndex + 1).trim();
+				if (value === "true") value = true;
+				else if (value === "false") value = false;
+				else if (/^-?\d+(\.\d+)?$/.test(value)) value = Number(value);
+				else if (value.startsWith("\"") && value.endsWith("\"")) value = value.slice(1, -1);
+				else if (value.startsWith("'") && value.endsWith("'")) value = value.slice(1, -1);
+				result[key] = value;
+			}
+		}
+		return result;
+	}
+	extractDescription(tree) {
+		for (const node of tree.children) {
+			if (node.type === "heading") break;
+			if (node.type === "paragraph") return (0, mdast_util_to_string.toString)(node);
+		}
+	}
+};
+/**
+* Extract links from a list-based markdown structure (like awesome lists)
+*/
+function extractListLinks(markdown) {
+	const tree = (0, mdast_util_from_markdown.fromMarkdown)(markdown);
+	const links = [];
+	let currentHeading = "";
+	(0, unist_util_visit.visit)(tree, (node) => {
+		if (node.type === "heading") currentHeading = (0, mdast_util_to_string.toString)(node);
+		if (node.type === "listItem") (0, unist_util_visit.visit)(node, "link", (linkNode) => {
+			links.push({
+				url: linkNode.url,
+				text: (0, mdast_util_to_string.toString)(linkNode),
+				title: linkNode.title ?? void 0,
+				context: currentHeading || void 0
+			});
+		});
+	});
+	return links;
+}
+/**
+* Parse markdown into sections by heading level
+*/
+function parseByHeadings(markdown, minLevel = 2) {
+	return new MarkdownParser().parse(markdown).data.sections.filter((s) => s.level >= minLevel);
+}
+
+//#endregion
+exports.MarkdownParser = MarkdownParser;
+exports.extractListLinks = extractListLinks;
+exports.fetchRepoMeta = fetchRepoMeta;
+exports.groupByCategory = groupByCategory;
+exports.isGitHubRepo = isGitHubRepo;
+exports.parseByHeadings = parseByHeadings;
+exports.parseGitHubUrl = parseGitHubUrl;
+exports.toRawUrl = toRawUrl;
+//# sourceMappingURL=index.cjs.map

package/dist/parsers/index.cjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.cjs","names":["sections: MarkdownSection[]","allLinks: MarkdownLink[]","codeBlocks: CodeBlock[]","frontmatter: Record<string, unknown> | undefined","currentSection: MarkdownSection | null","linkData: MarkdownLink","result: Record<string, unknown>","value: string | boolean | number","links: MarkdownLink[]"],"sources":["../../src/parsers/github.ts","../../src/parsers/markdown.ts"],"sourcesContent":["import type { GitHubMeta, MarkdownLink } from './types.js';\n\n/**\n * GitHub-specific utilities for parsing repositories.\n */\n\n/**\n * Check if a URL is a GitHub repository\n */\nexport function isGitHubRepo(url: string): boolean {\n  return /^https?:\\/\\/(www\\.)?github\\.com\\/[^/]+\\/[^/]+\\/?$/.test(url);\n}\n\n/**\n * Extract GitHub repo info from URL\n */\nexport function parseGitHubUrl(url: string): { owner: string; repo: string } | null {\n  const match = url.match(/github\\.com\\/([^/]+)\\/([^/]+)/);\n  if (!match || !match[1] || !match[2]) return null;\n  return {\n    owner: match[1],\n    repo: match[2].replace(/\\.git$/, ''),\n  };\n}\n\n/**\n * Convert a GitHub repo URL to raw content URL\n */\nexport function toRawUrl(url: string, branch = 'main', file = 'README.md'): string {\n  const info = parseGitHubUrl(url);\n  if (!info) return url;\n  return `https://raw.githubusercontent.com/${info.owner}/${info.repo}/${branch}/${file}`;\n}\n\n/**\n * Fetch GitHub API metadata for a repository\n * Note: This is a placeholder - actual implementation would need GitHub API access\n */\nexport async function fetchRepoMeta(\n  owner: string,\n  repo: string,\n  _token?: string\n): Promise<GitHubMeta> {\n  // This would make actual API calls in a full implementation\n  // For now, return basic info\n  return {\n    repoOwner: owner,\n    repoName: repo,\n  };\n}\n\n/**\n * Group links by their category/section\n */\nexport function groupByCategory(links: MarkdownLink[]): Map<string, MarkdownLink[]> {\n  const groups = new Map<string, MarkdownLink[]>();\n\n  for (const link of links) {\n    const category = link.context || 'Uncategorized';\n    const existing = groups.get(category) || [];\n    existing.push(link);\n    groups.set(category, existing);\n  }\n\n  return groups;\n}\n","import type { Code, Heading, Link, ListItem, Root } from 'mdast';\nimport { fromMarkdown } from 'mdast-util-from-markdown';\nimport { toString as mdastToString } from 'mdast-util-to-string';\nimport { visit } from 'unist-util-visit';\nimport type {\n  CodeBlock,\n  MarkdownLink,\n  MarkdownSection,\n  ParsedMarkdown,\n  ParserResult,\n  SourceParser,\n} from './types.js';\n\n/**\n * Generic Markdown parser.\n * Extracts structure, links, and code blocks from markdown content.\n *\n * @example\n * ```ts\n * const parser = new MarkdownParser();\n * const result = parser.parse(markdownContent);\n * console.log(result.data.sections);\n * console.log(result.data.links);\n * ```\n */\nexport class MarkdownParser implements SourceParser<ParsedMarkdown> {\n  readonly name = 'markdown';\n\n  canParse(content: string): boolean {\n    // Check for common markdown patterns\n    return (\n      content.includes('# ') ||\n      content.includes('## ') ||\n      content.includes('- [') ||\n      content.includes('* [') ||\n      content.includes('```')\n    );\n  }\n\n  parse(content: string): ParserResult<ParsedMarkdown> {\n    const tree = fromMarkdown(content);\n    const sections: MarkdownSection[] = [];\n    const allLinks: MarkdownLink[] = [];\n    const codeBlocks: CodeBlock[] = [];\n    let frontmatter: Record<string, unknown> | undefined;\n\n    // Extract frontmatter if present\n    if (content.startsWith('---')) {\n      const endIndex = content.indexOf('---', 3);\n      if (endIndex !== -1) {\n        const frontmatterContent = content.slice(3, endIndex).trim();\n        frontmatter = this.parseFrontmatter(frontmatterContent);\n      }\n    }\n\n    // Track current section\n    let currentSection: MarkdownSection | null = null;\n\n    // Process the AST\n    visit(tree, (node) => {\n      // Handle headings\n      if (node.type === 'heading') {\n        const heading = node as Heading;\n        const title = mdastToString(heading);\n\n        // Finalize previous section\n        if (currentSection) {\n          sections.push(currentSection);\n        }\n\n        currentSection = {\n          level: heading.depth,\n          title,\n          content: '',\n          links: [],\n        };\n      }\n\n      // Handle links\n      if (node.type === 'link') {\n        const link = node as Link;\n        const text = mdastToString(link);\n        const linkData: MarkdownLink = {\n          url: link.url,\n          text,\n          title: link.title ?? undefined,\n          context: currentSection?.title,\n        };\n\n        allLinks.push(linkData);\n        if (currentSection) {\n          currentSection.links.push(linkData);\n        }\n      }\n\n      // Handle code blocks\n      if (node.type === 'code') {\n        const code = node as Code;\n        codeBlocks.push({\n          language: code.lang ?? undefined,\n          code: code.value,\n          meta: code.meta ?? undefined,\n        });\n      }\n\n      // Accumulate content for current section\n      if (currentSection && node.type === 'paragraph') {\n        const text = mdastToString(node);\n        currentSection.content += (currentSection.content ? '\\n\\n' : '') + text;\n      }\n    });\n\n    // Finalize last section\n    if (currentSection) {\n      sections.push(currentSection);\n    }\n\n    // Extract title from first h1 or frontmatter\n    const title = (frontmatter?.title as string) ?? sections.find((s) => s.level === 1)?.title;\n\n    // Extract description from frontmatter or first paragraph before any heading\n    const description = (frontmatter?.description as string) ?? this.extractDescription(tree);\n\n    return {\n      data: {\n        title,\n        description,\n        sections,\n        links: allLinks,\n        codeBlocks,\n        frontmatter,\n      },\n    };\n  }\n\n  private parseFrontmatter(content: string): Record<string, unknown> {\n    const result: Record<string, unknown> = {};\n    const lines = content.split('\\n');\n\n    for (const line of lines) {\n      const colonIndex = line.indexOf(':');\n      if (colonIndex > 0) {\n        const key = line.slice(0, colonIndex).trim();\n        let value: string | boolean | number = line.slice(colonIndex + 1).trim();\n\n        // Parse simple types\n        if (value === 'true') value = true;\n        else if (value === 'false') value = false;\n        else if (/^-?\\d+(\\.\\d+)?$/.test(value)) value = Number(value);\n        else if (value.startsWith('\"') && value.endsWith('\"')) value = value.slice(1, -1);\n        else if (value.startsWith(\"'\") && value.endsWith(\"'\")) value = value.slice(1, -1);\n\n        result[key] = value;\n      }\n    }\n\n    return result;\n  }\n\n  private extractDescription(tree: Root): string | undefined {\n    // Find first paragraph before any heading\n    for (const node of tree.children) {\n      if (node.type === 'heading') break;\n      if (node.type === 'paragraph') {\n        return mdastToString(node);\n      }\n    }\n    return undefined;\n  }\n}\n\n/**\n * Extract links from a list-based markdown structure (like awesome lists)\n */\nexport function extractListLinks(markdown: string): MarkdownLink[] {\n  const tree = fromMarkdown(markdown);\n  const links: MarkdownLink[] = [];\n  let currentHeading = '';\n\n  visit(tree, (node) => {\n    if (node.type === 'heading') {\n      currentHeading = mdastToString(node as Heading);\n    }\n\n    if (node.type === 'listItem') {\n      const listItem = node as ListItem;\n\n      // Find links in this list item\n      visit(listItem, 'link', (linkNode: Link) => {\n        links.push({\n          url: linkNode.url,\n          text: mdastToString(linkNode),\n          title: linkNode.title ?? undefined,\n          context: currentHeading || undefined,\n        });\n      });\n    }\n  });\n\n  return links;\n}\n\n/**\n * Parse markdown into sections by heading level\n */\nexport function parseByHeadings(markdown: string, minLevel = 2): MarkdownSection[] {\n  const parser = new MarkdownParser();\n  const result = parser.parse(markdown);\n  return result.data.sections.filter((s) => s.level >= minLevel);\n}\n"],"mappings":";;;;;;;;;;;;AASA,SAAgB,aAAa,KAAsB;AACjD,QAAO,oDAAoD,KAAK,IAAI;;;;;AAMtE,SAAgB,eAAe,KAAqD;CAClF,MAAM,QAAQ,IAAI,MAAM,gCAAgC;AACxD,KAAI,CAAC,SAAS,CAAC,MAAM,MAAM,CAAC,MAAM,GAAI,QAAO;AAC7C,QAAO;EACL,OAAO,MAAM;EACb,MAAM,MAAM,GAAG,QAAQ,UAAU,GAAG;EACrC;;;;;AAMH,SAAgB,SAAS,KAAa,SAAS,QAAQ,OAAO,aAAqB;CACjF,MAAM,OAAO,eAAe,IAAI;AAChC,KAAI,CAAC,KAAM,QAAO;AAClB,QAAO,qCAAqC,KAAK,MAAM,GAAG,KAAK,KAAK,GAAG,OAAO,GAAG;;;;;;AAOnF,eAAsB,cACpB,OACA,MACA,QACqB;AAGrB,QAAO;EACL,WAAW;EACX,UAAU;EACX;;;;;AAMH,SAAgB,gBAAgB,OAAoD;CAClF,MAAM,yBAAS,IAAI,KAA6B;AAEhD,MAAK,MAAM,QAAQ,OAAO;EACxB,MAAM,WAAW,KAAK,WAAW;EACjC,MAAM,WAAW,OAAO,IAAI,SAAS,IAAI,EAAE;AAC3C,WAAS,KAAK,KAAK;AACnB,SAAO,IAAI,UAAU,SAAS;;AAGhC,QAAO;;;;;;;;;;;;;;;;;ACvCT,IAAa,iBAAb,MAAoE;CAClE,AAAS,OAAO;CAEhB,SAAS,SAA0B;AAEjC,SACE,QAAQ,SAAS,KAAK,IACtB,QAAQ,SAAS,MAAM,IACvB,QAAQ,SAAS,MAAM,IACvB,QAAQ,SAAS,MAAM,IACvB,QAAQ,SAAS,MAAM;;CAI3B,MAAM,SAA+C;EACnD,MAAM,kDAAoB,QAAQ;EAClC,MAAMA,WAA8B,EAAE;EACtC,MAAMC,WAA2B,EAAE;EACnC,MAAMC,aAA0B,EAAE;EAClC,IAAIC;AAGJ,MAAI,QAAQ,WAAW,MAAM,EAAE;GAC7B,MAAM,WAAW,QAAQ,QAAQ,OAAO,EAAE;AAC1C,OAAI,aAAa,IAAI;IACnB,MAAM,qBAAqB,QAAQ,MAAM,GAAG,SAAS,CAAC,MAAM;AAC5D,kBAAc,KAAK,iBAAiB,mBAAmB;;;EAK3D,IAAIC,iBAAyC;AAG7C,8BAAM,OAAO,SAAS;AAEpB,OAAI,KAAK,SAAS,WAAW;IAC3B,MAAM,UAAU;IAChB,MAAM,2CAAsB,QAAQ;AAGpC,QAAI,eACF,UAAS,KAAK,eAAe;AAG/B,qBAAiB;KACf,OAAO,QAAQ;KACf;KACA,SAAS;KACT,OAAO,EAAE;KACV;;AAIH,OAAI,KAAK,SAAS,QAAQ;IACxB,MAAM,OAAO;IACb,MAAM,0CAAqB,KAAK;IAChC,MAAMC,WAAyB;KAC7B,KAAK,KAAK;KACV;KACA,OAAO,KAAK,SAAS;KACrB,SAAS,gBAAgB;KAC1B;AAED,aAAS,KAAK,SAAS;AACvB,QAAI,eACF,gBAAe,MAAM,KAAK,SAAS;;AAKvC,OAAI,KAAK,SAAS,QAAQ;IACxB,MAAM,OAAO;AACb,eAAW,KAAK;KACd,UAAU,KAAK,QAAQ;KACvB,MAAM,KAAK;KACX,MAAM,KAAK,QAAQ;KACpB,CAAC;;AAIJ,OAAI,kBAAkB,KAAK,SAAS,aAAa;IAC/C,MAAM,0CAAqB,KAAK;AAChC,mBAAe,YAAY,eAAe,UAAU,SAAS,MAAM;;IAErE;AAGF,MAAI,eACF,UAAS,KAAK,eAAe;AAS/B,SAAO,EACL,MAAM;GACJ,OAPW,aAAa,SAAoB,SAAS,MAAM,MAAM,EAAE,UAAU,EAAE,EAAE;GAQjF,aALiB,aAAa,eAA0B,KAAK,mBAAmB,KAAK;GAMrF;GACA,OAAO;GACP;GACA;GACD,EACF;;CAGH,AAAQ,iBAAiB,SAA0C;EACjE,MAAMC,SAAkC,EAAE;EAC1C,MAAM,QAAQ,QAAQ,MAAM,KAAK;AAEjC,OAAK,MAAM,QAAQ,OAAO;GACxB,MAAM,aAAa,KAAK,QAAQ,IAAI;AACpC,OAAI,aAAa,GAAG;IAClB,MAAM,MAAM,KAAK,MAAM,GAAG,WAAW,CAAC,MAAM;IAC5C,IAAIC,QAAmC,KAAK,MAAM,aAAa,EAAE,CAAC,MAAM;AAGxE,QAAI,UAAU,OAAQ,SAAQ;aACrB,UAAU,QAAS,SAAQ;aAC3B,kBAAkB,KAAK,MAAM,CAAE,SAAQ,OAAO,MAAM;aACpD,MAAM,WAAW,KAAI,IAAI,MAAM,SAAS,KAAI,CAAE,SAAQ,MAAM,MAAM,GAAG,GAAG;aACxE,MAAM,WAAW,IAAI,IAAI,MAAM,SAAS,IAAI,CAAE,SAAQ,MAAM,MAAM,GAAG,GAAG;AAEjF,WAAO,OAAO;;;AAIlB,SAAO;;CAGT,AAAQ,mBAAmB,MAAgC;AAEzD,OAAK,MAAM,QAAQ,KAAK,UAAU;AAChC,OAAI,KAAK,SAAS,UAAW;AAC7B,OAAI,KAAK,SAAS,YAChB,2CAAqB,KAAK;;;;;;;AAUlC,SAAgB,iBAAiB,UAAkC;CACjE,MAAM,kDAAoB,SAAS;CACnC,MAAMC,QAAwB,EAAE;CAChC,IAAI,iBAAiB;AAErB,6BAAM,OAAO,SAAS;AACpB,MAAI,KAAK,SAAS,UAChB,qDAA+B,KAAgB;AAGjD,MAAI,KAAK,SAAS,WAIhB,6BAHiB,MAGD,SAAS,aAAmB;AAC1C,SAAM,KAAK;IACT,KAAK,SAAS;IACd,yCAAoB,SAAS;IAC7B,OAAO,SAAS,SAAS;IACzB,SAAS,kBAAkB;IAC5B,CAAC;IACF;GAEJ;AAEF,QAAO;;;;;AAMT,SAAgB,gBAAgB,UAAkB,WAAW,GAAsB;AAGjF,QAFe,IAAI,gBAAgB,CACb,MAAM,SAAS,CACvB,KAAK,SAAS,QAAQ,MAAM,EAAE,SAAS,SAAS"}