index_build

Instructions

Implementation Reference

• src/tools/doc.ts:58-76 (handler)

  The core handler function for indexBuild. It collects text files (.txt, .md, .html, .pdf) from the sandbox directory (or a given root), reads each file's content, creates a MiniSearch full-text index, and persists the index to .cache/index.json.

  export async function indexBuild(root?: string) {
    const base = root ? path.resolve(root) : CONFIG.sandboxDir;
    const files = await collectFiles(base);
    const docs: DocRecord[] = [];
    for (const p of files) {
      const { title, text } = await fileToText(p);
      docs.push({ id: p, path: p, title, text });
    }
    const mini = new MiniSearch({
      fields: ['title','text'],
      storeFields: ['path','title'],
      searchOptions: { boost: { title: 2 } }
    });
    mini.addAll(docs);
    const payload = { docs, index: mini.toJSON() };
    await fs.mkdir(path.dirname(INDEX_PATH), { recursive: true }).catch(()=>{});
    await fs.writeFile(INDEX_PATH, JSON.stringify(payload));
    return { ok: true, indexed: docs.length };
  }

• src/server.ts:142-142 (schema)

  Schema definition for index.build / index_build: accepts an optional 'root' string parameter.

  const indexBuildShape = { root: z.string().optional() };

• src/server.ts:143-149 (registration)

  Registration of the 'index.build' tool (dotted name) on the MCP server, mapping to the indexBuild handler.

  server.tool('index.build', 'Build MiniSearch index for documents in sandbox directory.',
    indexBuildShape, OPEN,
    async ({ root }) => {
      const res = await indexBuild(root);
      return { content: [{ type: 'text', text: JSON.stringify(res) }] };
    }
  );

• src/server.ts:150-156 (registration)

  Registration of the 'index_build' alias (underscore name) on the MCP server, also mapping to the indexBuild handler.

  server.tool('index_build', 'Alias of index.build',
    indexBuildShape, OPEN,
    async ({ root }) => {
      const res = await indexBuild(root);
      return { content: [{ type: 'text', text: JSON.stringify(res) }] };
    }
  );

• src/tools/doc.ts:27-39 (helper)

  Helper function collectFiles: recursively walks a directory and collects file paths matching .txt, .md, .html, or .pdf extensions.

  async function collectFiles(root: string): Promise<string[]> {
    const out: string[] = [];
    async function walk(dir: string) {
      const ents = await fs.readdir(dir, { withFileTypes: true });
      for (const e of ents) {
        const p = path.join(dir, e.name);
        if (e.isDirectory()) await walk(p);
        else if (/(\.txt|\.md|\.html?|\.pdf)$/i.test(e.name)) out.push(p);
      }
    }
    await walk(root);
    return out;
  }

• src/tools/doc.ts:41-56 (helper)

  Helper function fileToText: reads a file and extracts its title and text content, handling .pdf (via pdf-parse), .html (via JSDOM + Readability), and plain text files.

  async function fileToText(p: string): Promise<{ title: string, text: string }> {
    const buf = await fs.readFile(p);
    const name = path.basename(p);
    if (/\.pdf$/i.test(p)) {
      try { const parsed = await pdfParseLazy(buf as unknown as Buffer); return { title: name, text: parsed.text || '' }; }
      catch { return { title: name, text: '' }; }
    }
    const s = buf.toString('utf-8');
    if (/\.html?$/i.test(p)) {
      const dom = new JSDOM(s, { url: 'file://' + p });
      const reader = new Readability(dom.window.document);
      const art = reader.parse();
      return { title: art?.title || name, text: art?.textContent || dom.window.document.body.textContent || s };
    }
    return { title: name, text: s };
  }