feat(tools): add docx extraction for minio ingestion

src/tools/builtin/minio-ingest.test.ts

@@ -32,6 +32,7 @@ describe('minio ingest internals', () => {
     expect(minioIngestInternals.isLikelyTextObject('notes/today.md')).toBe(true);
     expect(minioIngestInternals.isLikelyTextObject('logs/daemon.log')).toBe(true);
     expect(minioIngestInternals.isLikelyTextObject('manual.pdf')).toBe(true);
+    expect(minioIngestInternals.isLikelyTextObject('manual.docx')).toBe(true);
   });
 
   it('rejects likely binary extensions', () => {
@@ -79,7 +80,7 @@ describe('createMinioIngestTool', () => {
     const execRunner = vi.fn();
     const tool = createMinioIngestTool(makeBackupConfig(), store, { execRunner });
 
-    const result = await tool.execute({ object_key: 'knowledge/diagram.docx' });
+    const result = await tool.execute({ object_key: 'knowledge/diagram.exe' });
     expect(result.success).toBe(false);
     expect(result.error).toContain('Unsupported object type');
     expect(execRunner).not.toHaveBeenCalled();
@@ -121,6 +122,37 @@ describe('createMinioIngestTool', () => {
     );
   });
 
+  it('extracts DOCX text with pandoc', async () => {
+    const write = vi.fn();
+    const store = { write } as unknown as MemoryStore;
+    const execRunner = vi.fn(async (_file: string, args: string[]) => {
+      if (_file === 'mc' && args[0] === 'cp') {
+        return { stdout: '', stderr: '' };
+      }
+      if (_file === 'pandoc') {
+        return { stdout: 'Extracted DOCX text', stderr: '' };
+      }
+      return { stdout: '', stderr: '' };
+    });
+    const tool = createMinioIngestTool(makeBackupConfig(), store, { execRunner });
+
+    const result = await tool.execute({
+      object_key: 'knowledge/spec.docx',
+      mode: 'replace',
+    });
+    expect(result.success).toBe(true);
+    expect(write).toHaveBeenCalledWith(
+      'global/knowledge',
+      expect.stringContaining('Extracted DOCX text'),
+      'replace',
+    );
+    expect(execRunner).toHaveBeenCalledWith(
+      'pandoc',
+      expect.arrayContaining(['-t', 'plain']),
+      expect.any(Object),
+    );
+  });
+
   it('allows non-text extension when force=true', async () => {
     const write = vi.fn();
     const store = { write } as unknown as MemoryStore;

src/tools/builtin/minio-ingest.ts

@@ -34,6 +34,7 @@ const TEXT_EXTENSIONS = new Set([
 ]);
 const EXTRACTABLE_BINARY_EXTENSIONS = new Set([
   '.pdf',
+  '.docx',
 ]);
 
 export interface MinioIngestDeps {
@@ -61,6 +62,7 @@ async function readObjectText(
   objectKey: string,
   env: NodeJS.ProcessEnv,
 ): Promise<string> {
+  const ext = extname(objectKey).toLowerCase();
   if (!isExtractableBinaryObject(objectKey)) {
     const { stdout } = await runner('mc', ['cat', remotePath], { env, maxBuffer: 20 * 1024 * 1024 });
     return toText(stdout);
@@ -70,8 +72,26 @@ async function readObjectText(
   const localPath = join(tempDir, 'object.bin');
   try {
     await runner('mc', ['cp', remotePath, localPath], { env, maxBuffer: 20 * 1024 * 1024 });
-    const { stdout } = await runner('pdftotext', ['-q', localPath, '-'], { maxBuffer: 20 * 1024 * 1024 });
-    return toText(stdout);
+    if (ext === '.pdf') {
+      const { stdout } = await runner('pdftotext', ['-q', localPath, '-'], { maxBuffer: 20 * 1024 * 1024 });
+      return toText(stdout);
+    }
+    if (ext === '.docx') {
+      try {
+        const { stdout } = await runner('pandoc', [localPath, '-t', 'plain'], { maxBuffer: 20 * 1024 * 1024 });
+        return toText(stdout);
+      } catch (pandocError) {
+        try {
+          const { stdout } = await runner('docx2txt', [localPath, '-'], { maxBuffer: 20 * 1024 * 1024 });
+          return toText(stdout);
+        } catch (docx2txtError) {
+          const pErr = pandocError instanceof Error ? pandocError.message : String(pandocError);
+          const dErr = docx2txtError instanceof Error ? docx2txtError.message : String(docx2txtError);
+          throw new Error(`Failed to extract DOCX text (pandoc/docx2txt unavailable or failed): ${pErr}; ${dErr}`);
+        }
+      }
+    }
+    throw new Error(`Unsupported extractable object type: ${objectKey}`);
   } finally {
     rmSync(tempDir, { recursive: true, force: true });
   }

src/tools/builtin/minio-sync.test.ts

@@ -104,7 +104,7 @@ describe('createMinioSyncTool', () => {
     const store = { write } as unknown as MemoryStore;
     const execRunner = vi.fn(async (_file: string, args: string[]) => {
       if (args[0] === 'ls') {
-        return { stdout: '{"status":"success","type":"file","key":"knowledge/diagram.docx"}', stderr: '' };
+        return { stdout: '{"status":"success","type":"file","key":"knowledge/diagram.exe"}', stderr: '' };
       }
       return { stdout: 'fake text', stderr: '' };
     });
@@ -117,6 +117,33 @@ describe('createMinioSyncTool', () => {
     expect(write).not.toHaveBeenCalled();
   });
 
+  it('syncs DOCX objects via text extraction', async () => {
+    const write = vi.fn();
+    const store = { write } as unknown as MemoryStore;
+    const execRunner = vi.fn(async (_file: string, args: string[]) => {
+      if (_file === 'mc' && args[0] === 'ls') {
+        return { stdout: '{"status":"success","type":"file","key":"knowledge/spec.docx"}', stderr: '' };
+      }
+      if (_file === 'mc' && args[0] === 'cp') {
+        return { stdout: '', stderr: '' };
+      }
+      if (_file === 'pandoc') {
+        return { stdout: 'DOCX sync text', stderr: '' };
+      }
+      return { stdout: '', stderr: '' };
+    });
+
+    const tool = createMinioSyncTool(makeBackupConfig(), store, { execRunner });
+    const result = await tool.execute({ prefix: 'knowledge/' });
+    expect(result.success).toBe(true);
+    expect(result.output).toContain('Imported: 1');
+    expect(write).toHaveBeenCalledWith(
+      'global/knowledge/minio/knowledge/spec',
+      expect.stringContaining('DOCX sync text'),
+      'append',
+    );
+  });
+
   it('returns an error when minio is disabled', async () => {
     const write = vi.fn();
     const store = { write } as unknown as MemoryStore;