@open330/kiwimu 0.4.1 → 0.7.1

package/src/ingest/docx.ts

@@ -5,11 +5,3 @@ export async function extractTextFromDocx(filePath: string): Promise<{ title: st
   const title = filePath.split("/").pop()?.replace(/\.docx?$/i, "") || "Untitled";
   return { title, text };
 }
-
-export async function extractHtmlFromDocx(filePath: string): Promise<{ title: string; html: string }> {
-  const mammoth = require("mammoth");
-  const result = await mammoth.convertToHtml({ path: filePath });
-  const html: string = result.value;
-  const title = filePath.split("/").pop()?.replace(/\.docx?$/i, "") || "Untitled";
-  return { title, html };
-}

package/src/ingest/legacy.ts

@@ -9,7 +9,7 @@ export async function extractWithTextutil(filePath: string): Promise<{ title: st
   const textutilFormats = new Set(["doc", "rtf", "odt"]);
 
   if (textutilFormats.has(ext)) {
-    const proc = Bun.spawn(["textutil", "-convert", "txt", "-stdout", filePath], {
+    const proc = Bun.spawn(["textutil", "-convert", "txt", "-stdout", "--", filePath], {
       stdout: "pipe",
       stderr: "pipe",
     });
@@ -26,12 +26,12 @@ export async function extractWithTextutil(filePath: string): Promise<{ title: st
   if (ext === "key") {
     // Try to extract text using mdimport/spotlight metadata
     try {
-      const proc = Bun.spawn(["mdimport", "-d2", filePath], { stdout: "pipe", stderr: "pipe" });
+      const proc = Bun.spawn(["mdimport", "-d2", "--", filePath], { stdout: "pipe", stderr: "pipe" });
       await proc.exited;
     } catch {}
 
     // Keynote files are directories or zip-like packages. Try strings extraction.
-    const proc = Bun.spawn(["strings", filePath], { stdout: "pipe", stderr: "pipe" });
+    const proc = Bun.spawn(["strings", "--", filePath], { stdout: "pipe", stderr: "pipe" });
     const raw = await new Response(proc.stdout).text();
     await proc.exited;
 
@@ -50,7 +50,7 @@ export async function extractWithTextutil(filePath: string): Promise<{ title: st
 
   // For .ppt (legacy PowerPoint), try textutil or strings
   if (ext === "ppt") {
-    const proc = Bun.spawn(["strings", filePath], { stdout: "pipe", stderr: "pipe" });
+    const proc = Bun.spawn(["strings", "--", filePath], { stdout: "pipe", stderr: "pipe" });
     const raw = await new Response(proc.stdout).text();
     await proc.exited;
 

package/src/ingest/pdf.ts

@@ -1,5 +1,5 @@
 export async function extractTextFromPdf(pdfPath: string): Promise<{ title: string; text: string }> {
-  let pdfParse: any;
+  let pdfParse: (buffer: Buffer) => Promise<{ info?: { Title?: string }; text: string }>;
   try {
     pdfParse = require("pdf-parse");
   } catch {

package/src/ingest/pptx.ts

@@ -2,7 +2,6 @@ import { readFileSync } from "fs";
 
 export async function extractTextFromPptx(filePath: string): Promise<{ title: string; text: string }> {
   // PPTX is a ZIP containing XML files
-  const { Decompress } = await import("bun");
   const JSZip = (await import("jszip")).default;
 
   const buffer = readFileSync(filePath);

package/src/ingest/web.test.ts

@@ -0,0 +1,41 @@
+import { expect, test, describe } from "bun:test";
+import { validateUrl } from "./web";
+
+describe("validateUrl", () => {
+  test("정상 HTTP URL 허용", () => {
+    expect(() => validateUrl("http://example.com")).not.toThrow();
+  });
+  test("정상 HTTPS URL 허용", () => {
+    expect(() => validateUrl("https://example.com/page")).not.toThrow();
+  });
+  test("localhost 차단", () => {
+    expect(() => validateUrl("http://localhost:3000")).toThrow();
+  });
+  test("127.0.0.1 차단", () => {
+    expect(() => validateUrl("http://127.0.0.1")).toThrow();
+  });
+  test("10.x.x.x 차단", () => {
+    expect(() => validateUrl("http://10.0.0.1")).toThrow();
+  });
+  test("172.16.x.x 차단", () => {
+    expect(() => validateUrl("http://172.16.0.1")).toThrow();
+  });
+  test("192.168.x.x 차단", () => {
+    expect(() => validateUrl("http://192.168.1.1")).toThrow();
+  });
+  test("169.254.x.x 차단", () => {
+    expect(() => validateUrl("http://169.254.169.254")).toThrow();
+  });
+  test("file:// 프로토콜 차단", () => {
+    expect(() => validateUrl("file:///etc/passwd")).toThrow();
+  });
+  test("ftp:// 프로토콜 차단", () => {
+    expect(() => validateUrl("ftp://example.com")).toThrow();
+  });
+  test(".local 도메인 차단", () => {
+    expect(() => validateUrl("http://server.local")).toThrow();
+  });
+  test("0.0.0.0 차단", () => {
+    expect(() => validateUrl("http://0.0.0.0")).toThrow();
+  });
+});

package/src/ingest/web.ts

@@ -1,77 +1,76 @@
 import * as cheerio from "cheerio";
+import { URL } from "url";
 
-export interface Section {
-  level: number;
-  title: string;
-  htmlParts: string[];
-}
-
-const HEADING_TAGS = new Set(["h1", "h2", "h3", "h4"]);
-const SKIP_TAGS = new Set(["nav", "header", "footer", "script", "style", "noscript"]);
-const CONTAINER_TAGS = new Set([
-  "html", "head", "body", "div", "article", "main", "section", "aside", "details", "summary",
-]);
+/**
+ * Validate a URL to prevent SSRF attacks.
+ * Blocks private/internal IP ranges and non-http(s) schemes.
+ */
+export function validateUrl(urlStr: string): void {
+  let parsed: URL;
+  try {
+    parsed = new URL(urlStr);
+  } catch {
+    throw new Error("유효하지 않은 URL입니다");
+  }
 
-export async function fetchPage(url: string): Promise<{ title: string; html: string }> {
-  const resp = await fetch(url, {
-    headers: { "User-Agent": "kiwimu/0.2 (learning wiki builder)" },
-  });
-  if (!resp.ok) throw new Error(`Failed to fetch ${url}: ${resp.status}`);
-  const html = await resp.text();
-  const $ = cheerio.load(html);
-  const title = $("title").text().trim() || url;
-  const body = $("body").html() || html;
-  return { title, html: body };
-}
+  if (parsed.protocol !== "http:" && parsed.protocol !== "https:") {
+    throw new Error("http 또는 https URL만 허용됩니다");
+  }
 
-export function extractSections(html: string): Section[] {
-  const $ = cheerio.load(html, null, false);
-  const sections: Section[] = [];
-  let current: Section = { level: 1, title: "Introduction", htmlParts: [] };
+  const hostname = parsed.hostname;
 
-  function walk(el: cheerio.AnyNode): void {
-    if (el.type === "text") return;
-    if (el.type !== "tag") return;
+  // Block IP-based hostnames in private ranges
+  // IPv4 pattern
+  const ipv4Match = hostname.match(/^(\d{1,3})\.(\d{1,3})\.(\d{1,3})\.(\d{1,3})$/);
+  if (ipv4Match) {
+    const [, a, b, c, d] = ipv4Match.map(Number);
+    if (
+      a === 127 ||                              // 127.0.0.0/8
+      a === 10 ||                               // 10.0.0.0/8
+      (a === 172 && b >= 16 && b <= 31) ||      // 172.16.0.0/12
+      (a === 192 && b === 168) ||               // 192.168.0.0/16
+      (a === 169 && b === 254) ||               // 169.254.0.0/16
+      (a === 0 && b === 0 && c === 0 && d === 0) // 0.0.0.0
+    ) {
+      throw new Error("내부 네트워크 주소는 허용되지 않습니다");
+    }
+  }
 
-    const tagName = (el as cheerio.Element).tagName.toLowerCase();
+  // Block common private hostnames
+  if (hostname === "localhost" || hostname === "[::1]" || hostname.endsWith(".local")) {
+    throw new Error("내부 네트워크 주소는 허용되지 않습니다");
+  }
+}
 
-    if (SKIP_TAGS.has(tagName)) return;
+export async function fetchPage(url: string): Promise<{ title: string; html: string }> {
+  validateUrl(url);
 
-    if (HEADING_TAGS.has(tagName)) {
-      if (current.htmlParts.length > 0) {
-        sections.push(current);
-      }
-      current = {
-        level: parseInt(tagName[1]),
-        title: $(el).text().trim(),
-        htmlParts: [],
-      };
-      return;
-    }
+  let currentUrl = url;
+  const maxRedirects = 5;
 
-    if (CONTAINER_TAGS.has(tagName)) {
-      for (const child of (el as cheerio.Element).children) {
-        walk(child);
-      }
-      return;
-    }
+  for (let i = 0; i <= maxRedirects; i++) {
+    const resp = await fetch(currentUrl, {
+      headers: { "User-Agent": "kiwimu/0.4 (learning wiki builder)" },
+      redirect: "manual",
+    });
 
-    // Content element
-    const html = $.html(el)?.trim();
-    if (html) {
-      current.htmlParts.push(html);
+    if (resp.status >= 300 && resp.status < 400) {
+      const location = resp.headers.get("location");
+      if (!location) throw new Error(`Redirect without location header from ${currentUrl}`);
+      // Resolve relative redirect URLs
+      const redirectUrl = new URL(location, currentUrl).href;
+      validateUrl(redirectUrl); // Re-validate redirect target to prevent SSRF bypass
+      currentUrl = redirectUrl;
+      continue;
     }
-  }
-
-  // Walk root children
-  const root = $.root();
-  for (const child of root.contents().toArray()) {
-    walk(child);
-  }
 
-  if (current.htmlParts.length > 0) {
-    sections.push(current);
+    if (!resp.ok) throw new Error(`Failed to fetch ${currentUrl}: ${resp.status}`);
+    const html = await resp.text();
+    const $ = cheerio.load(html);
+    const title = $("title").text().trim() || url;
+    const body = $("body").html() || html;
+    return { title, html: body };
   }
 
-  return sections.filter((s) => s.htmlParts.length > 0);
+  throw new Error(`Too many redirects fetching ${url}`);
 }

package/src/llm-client.ts

@@ -8,73 +8,14 @@ export interface UsageStats {
   totalTokens: number;
 }
 
-const _usage: UsageStats = {
-  totalCalls: 0,
-  promptTokens: 0,
-  completionTokens: 0,
-  totalTokens: 0,
-};
-
-let _llmConfig: LLMConfig | null = null;
-
-export function setLLMConfig(config: LLMConfig): void {
-  _llmConfig = config;
-}
-
-export function getLLMConfig(): LLMConfig {
-  if (!_llmConfig) throw new Error("LLM config not set. Call setLLMConfig() first.");
-  return _llmConfig;
-}
-
-export function getUsageStats(): UsageStats {
-  return { ..._usage };
-}
-
-export function resetUsageStats(): void {
-  _usage.totalCalls = 0;
-  _usage.promptTokens = 0;
-  _usage.completionTokens = 0;
-  _usage.totalTokens = 0;
-}
-
-export function getEstimatedCost(): number {
-  const config = _llmConfig;
-  if (!config) return 0;
-
-  // Pricing per 1M tokens (approximate)
-  const pricing: Record<string, { input: number; output: number }> = {
-    "gemini": { input: 0.075, output: 0.30 },
-    "azure-openai": { input: 0.10, output: 0.40 },
-    "openai": { input: 0.15, output: 0.60 },
-    "anthropic": { input: 3.00, output: 15.00 },
-  };
-  const p = pricing[config.provider] || pricing["gemini"];
-  return (_usage.promptTokens / 1_000_000) * p.input + (_usage.completionTokens / 1_000_000) * p.output;
-}
-
-export function printUsageSummary(): void {
-  const u = _usage;
-  const cost = getEstimatedCost();
-  const provider = _llmConfig?.provider || "unknown";
-  const model = _llmConfig?.model || "unknown";
-
-  console.log(`\x1b[34m📊 LLM 사용량 (${provider}/${model}):\x1b[0m`);
-  console.log(`  호출 횟수:     ${u.totalCalls}회`);
-  console.log(`  입력 토큰:     ${u.promptTokens.toLocaleString()}`);
-  console.log(`  출력 토큰:     ${u.completionTokens.toLocaleString()}`);
-  console.log(`  총 토큰:       ${u.totalTokens.toLocaleString()}`);
-  console.log(`  예상 비용:     ~$${cost.toFixed(4)}`);
-}
-
 // ── Provider implementations ──
 
-async function geminiComplete(system: string, userMessage: string, maxTokens: number): Promise<{ text: string; usage?: any }> {
-  const config = getLLMConfig();
-  const url = `https://generativelanguage.googleapis.com/v1beta/models/${config.model}:generateContent?key=${config.api_key}`;
+async function geminiComplete(config: LLMConfig, system: string, userMessage: string, maxTokens: number): Promise<{ text: string; usage?: { prompt_tokens: number; completion_tokens: number; total_tokens: number } }> {
+  const url = `https://generativelanguage.googleapis.com/v1beta/models/${config.model}:generateContent`;
 
   const resp = await fetch(url, {
     method: "POST",
-    headers: { "Content-Type": "application/json" },
+    headers: { "Content-Type": "application/json", "x-goog-api-key": config.api_key },
     body: JSON.stringify({
       system_instruction: { parts: [{ text: system }] },
       contents: [{ parts: [{ text: userMessage }] }],
@@ -87,9 +28,10 @@ async function geminiComplete(system: string, userMessage: string, maxTokens: nu
     throw new Error(`Gemini API error (${resp.status}): ${err.slice(0, 200)}`);
   }
 
-  const data = await resp.json();
-  const text = data.candidates?.[0]?.content?.parts?.[0]?.text || "";
-  const usage = data.usageMetadata;
+  const data = await resp.json() as Record<string, unknown>;
+  const candidates = data.candidates as Array<{ content: { parts: Array<{ text: string }> } }> | undefined;
+  const text = candidates?.[0]?.content?.parts?.[0]?.text || "";
+  const usage = data.usageMetadata as { promptTokenCount?: number; candidatesTokenCount?: number; totalTokenCount?: number } | undefined;
   return {
     text,
     usage: usage ? {
@@ -100,78 +42,213 @@ async function geminiComplete(system: string, userMessage: string, maxTokens: nu
   };
 }
 
-async function azureOpenAIComplete(system: string, userMessage: string, maxTokens: number): Promise<{ text: string; usage?: any }> {
-  const config = getLLMConfig();
-
-  // Try loading from ~/keys/openai.azure.com/ if no api_key in config
-  let apiKey = config.api_key;
-  let endpoint = config.endpoint;
-  let model = config.model;
-
-  if (!apiKey) {
-    try {
-      const keyFile = `${process.env.HOME}/keys/openai.azure.com/${config.model}.json`;
-      const raw = require("fs").readFileSync(keyFile, "utf-8");
-      const keyConfig = JSON.parse(raw)[0];
-      apiKey = keyConfig.key;
-      endpoint = keyConfig.endpoint.split("/openai/")[0];
-      model = keyConfig.deployment;
-    } catch {
-      throw new Error("Azure OpenAI API key not configured");
+// ── Class-based LLM client ──
+
+type ProviderResult = { text: string; usage?: { prompt_tokens: number; completion_tokens: number; total_tokens: number } };
+
+export class LLMClient {
+  private config: LLMConfig;
+  private usage: UsageStats = { totalCalls: 0, promptTokens: 0, completionTokens: 0, totalTokens: 0 };
+  private _openaiClient: InstanceType<typeof import("openai").default> | null = null;
+  private _anthropicClient: InstanceType<typeof import("@anthropic-ai/sdk").default> | null = null;
+  private _azureClient: InstanceType<typeof import("openai").AzureOpenAI> | null = null;
+
+  constructor(config: LLMConfig) {
+    this.config = config;
+  }
+
+  private async azureComplete(system: string, userMessage: string, maxTokens: number): Promise<ProviderResult> {
+    let apiKey = this.config.api_key;
+    let endpoint = this.config.endpoint;
+    let model = this.config.model;
+
+    if (!apiKey) {
+      try {
+        const keyFile = `${process.env.HOME}/keys/openai.azure.com/${this.config.model}.json`;
+        const raw = require("fs").readFileSync(keyFile, "utf-8");
+        const keyConfig = JSON.parse(raw)[0] as { key: string; endpoint: string; deployment: string };
+        apiKey = keyConfig.key;
+        endpoint = keyConfig.endpoint.split("/openai/")[0];
+        model = keyConfig.deployment;
+      } catch {
+        throw new Error("Azure OpenAI API key not configured");
+      }
+    }
+
+    if (!this._azureClient) {
+      const { AzureOpenAI } = await import("openai");
+      this._azureClient = new AzureOpenAI({ endpoint, apiKey, deployment: model, apiVersion: "2024-12-01-preview" });
     }
+
+    const resp = await this._azureClient.chat.completions.create({
+      model: model,
+      max_completion_tokens: maxTokens,
+      messages: [
+        { role: "system", content: system },
+        { role: "user", content: userMessage },
+      ],
+    });
+
+    return {
+      text: resp.choices[0]?.message?.content || "",
+      usage: resp.usage ? {
+        prompt_tokens: resp.usage.prompt_tokens || 0,
+        completion_tokens: resp.usage.completion_tokens || 0,
+        total_tokens: resp.usage.total_tokens || 0,
+      } : undefined,
+    };
   }
 
-  const { AzureOpenAI } = await import("openai");
-  const client = new AzureOpenAI({ endpoint, apiKey, deployment: model, apiVersion: "2024-12-01-preview" });
+  private async openaiComplete(system: string, userMessage: string, maxTokens: number): Promise<ProviderResult> {
+    const { default: OpenAI } = await import("openai");
+    if (!this._openaiClient) {
+      this._openaiClient = new OpenAI({ apiKey: this.config.api_key });
+    }
+    const resp = await this._openaiClient.chat.completions.create({
+      model: this.config.model || "gpt-4o",
+      messages: [
+        { role: "system", content: system },
+        { role: "user", content: userMessage },
+      ],
+      max_tokens: maxTokens,
+    });
+    return {
+      text: resp.choices[0]?.message?.content || "",
+      usage: resp.usage ? {
+        prompt_tokens: resp.usage.prompt_tokens || 0,
+        completion_tokens: resp.usage.completion_tokens || 0,
+        total_tokens: resp.usage.total_tokens || 0,
+      } : undefined,
+    };
+  }
 
-  const resp = await client.chat.completions.create({
-    model,
-    max_completion_tokens: maxTokens,
-    messages: [
-      { role: "system", content: system },
-      { role: "user", content: userMessage },
-    ],
-  });
+  private async anthropicComplete(system: string, userMessage: string, maxTokens: number): Promise<ProviderResult> {
+    const { default: Anthropic } = await import("@anthropic-ai/sdk");
+    if (!this._anthropicClient) {
+      this._anthropicClient = new Anthropic({ apiKey: this.config.api_key });
+    }
+    const resp = await this._anthropicClient.messages.create({
+      model: this.config.model || "claude-sonnet-4-20250514",
+      max_tokens: maxTokens,
+      system: system,
+      messages: [{ role: "user", content: userMessage }],
+    });
+    const content = resp.content[0]?.type === "text" ? resp.content[0].text : "";
+    return {
+      text: content,
+      usage: resp.usage ? {
+        prompt_tokens: resp.usage.input_tokens || 0,
+        completion_tokens: resp.usage.output_tokens || 0,
+        total_tokens: (resp.usage.input_tokens || 0) + (resp.usage.output_tokens || 0),
+      } : undefined,
+    };
+  }
 
-  return {
-    text: resp.choices[0]?.message?.content || "",
-    usage: resp.usage ? {
-      prompt_tokens: resp.usage.prompt_tokens || 0,
-      completion_tokens: resp.usage.completion_tokens || 0,
-      total_tokens: resp.usage.total_tokens || 0,
-    } : undefined,
-  };
+  async chatComplete(system: string, userMessage: string, maxTokens = 8192): Promise<string> {
+    let result: ProviderResult;
+
+    switch (this.config.provider) {
+      case "gemini":
+        result = await geminiComplete(this.config, system, userMessage, maxTokens);
+        break;
+      case "azure-openai":
+        result = await this.azureComplete(system, userMessage, maxTokens);
+        break;
+      case "openai":
+        result = await this.openaiComplete(system, userMessage, maxTokens);
+        break;
+      case "anthropic":
+        result = await this.anthropicComplete(system, userMessage, maxTokens);
+        break;
+      default:
+        throw new Error(`Unknown LLM provider: ${this.config.provider}`);
+    }
+
+    // Track usage
+    if (result.usage) {
+      this.usage.totalCalls++;
+      this.usage.promptTokens += result.usage.prompt_tokens || 0;
+      this.usage.completionTokens += result.usage.completion_tokens || 0;
+      this.usage.totalTokens += result.usage.total_tokens || 0;
+    }
+
+    return result.text;
+  }
+
+  getUsageStats(): UsageStats {
+    return { ...this.usage };
+  }
+
+  resetUsageStats(): void {
+    this.usage.totalCalls = 0;
+    this.usage.promptTokens = 0;
+    this.usage.completionTokens = 0;
+    this.usage.totalTokens = 0;
+  }
+
+  getEstimatedCost(): number {
+    // Pricing per 1M tokens (approximate)
+    const pricing: Record<string, { input: number; output: number }> = {
+      "gemini": { input: 0.075, output: 0.30 },
+      "azure-openai": { input: 0.10, output: 0.40 },
+      "openai": { input: 2.50, output: 10.00 },
+      "anthropic": { input: 3.00, output: 15.00 },
+    };
+    const p = pricing[this.config.provider] || pricing["gemini"];
+    return (this.usage.promptTokens / 1_000_000) * p.input + (this.usage.completionTokens / 1_000_000) * p.output;
+  }
+
+  printUsageSummary(): void {
+    const u = this.usage;
+    const cost = this.getEstimatedCost();
+
+    console.log(`\x1b[34m📊 LLM 사용량 (${this.config.provider}/${this.config.model}):\x1b[0m`);
+    console.log(`  호출 횟수:     ${u.totalCalls}회`);
+    console.log(`  입력 토큰:     ${u.promptTokens.toLocaleString()}`);
+    console.log(`  출력 토큰:     ${u.completionTokens.toLocaleString()}`);
+    console.log(`  총 토큰:       ${u.totalTokens.toLocaleString()}`);
+    console.log(`  예상 비용:     ~$${cost.toFixed(4)}`);
+  }
 }
 
-// ── Main interface ──
+// ── Deprecated global state wrappers (for backward compatibility) ──
+
+/** @deprecated Use LLMClient class instead */
+let _globalClient: LLMClient | null = null;
 
+/** @deprecated Use `new LLMClient(config)` instead */
+export function setLLMConfig(config: LLMConfig): void {
+  _globalClient = new LLMClient(config);
+}
+
+/** @deprecated Use LLMClient instance methods instead */
+export function getUsageStats(): UsageStats {
+  if (!_globalClient) return { totalCalls: 0, promptTokens: 0, completionTokens: 0, totalTokens: 0 };
+  return _globalClient.getUsageStats();
+}
+
+/** @deprecated Use LLMClient instance methods instead */
+export function resetUsageStats(): void {
+  if (_globalClient) _globalClient.resetUsageStats();
+}
+
+/** @deprecated Use LLMClient instance methods instead */
+export function getEstimatedCost(): number {
+  if (!_globalClient) return 0;
+  return _globalClient.getEstimatedCost();
+}
+
+/** @deprecated Use LLMClient instance methods instead */
+export function printUsageSummary(): void {
+  if (_globalClient) _globalClient.printUsageSummary();
+}
+
+/** @deprecated Use LLMClient instance methods instead */
 export async function chatComplete(
   system: string,
   userMessage: string,
   maxTokens = 8192
 ): Promise<string> {
-  const config = getLLMConfig();
-
-  let result: { text: string; usage?: any };
-
-  switch (config.provider) {
-    case "gemini":
-      result = await geminiComplete(system, userMessage, maxTokens);
-      break;
-    case "azure-openai":
-      result = await azureOpenAIComplete(system, userMessage, maxTokens);
-      break;
-    default:
-      throw new Error(`Unknown LLM provider: ${config.provider}`);
-  }
-
-  // Track usage
-  if (result.usage) {
-    _usage.totalCalls++;
-    _usage.promptTokens += result.usage.prompt_tokens || 0;
-    _usage.completionTokens += result.usage.completion_tokens || 0;
-    _usage.totalTokens += result.usage.total_tokens || 0;
-  }
-
-  return result.text;
+  if (!_globalClient) throw new Error("LLM config not set. Call setLLMConfig() first.");
+  return _globalClient.chatComplete(system, userMessage, maxTokens);
 }

package/src/pipeline/chunker.test.ts

@@ -0,0 +1,42 @@
+import { expect, test, describe } from "bun:test";
+import { slugify, cleanTitle } from "./chunker";
+
+describe("slugify", () => {
+  test("영어 텍스트", () => {
+    expect(slugify("Hello World")).toBe("hello-world");
+  });
+  test("한국어 텍스트", () => {
+    expect(slugify("양자역학")).toBe("양자역학");
+  });
+  test("한영 혼합", () => {
+    expect(slugify("Chapter 3 양자역학")).toBe("chapter-3-양자역학");
+  });
+  test("특수문자 제거", () => {
+    expect(slugify("Hello! @World#")).toBe("hello-world");
+  });
+  test("빈 문자열", () => {
+    expect(slugify("")).toBe("");
+  });
+  test("연속 공백/하이픈", () => {
+    expect(slugify("hello   world---test")).toBe("hello-world-test");
+  });
+  test("80자 제한", () => {
+    const long = "a".repeat(100);
+    expect(slugify(long).length).toBeLessThanOrEqual(80);
+  });
+  test("한글 자모", () => {
+    expect(slugify("ㅋㅋㅋ 테스트")).toBe("ㅋㅋㅋ-테스트");
+  });
+});
+
+describe("cleanTitle", () => {
+  test("Chapter 번호 제거", () => {
+    expect(cleanTitle("Chapter 3 Quantum Mechanics")).toBe("Quantum Mechanics");
+  });
+  test("숫자 접두사 제거", () => {
+    expect(cleanTitle("3.2.1 Angular Momentum")).toBe("Angular Momentum");
+  });
+  test("일반 제목 유지", () => {
+    expect(cleanTitle("Quantum Mechanics")).toBe("Quantum Mechanics");
+  });
+});