fix(speech): serialize pronunciation dictionary correctly

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: kartikkabadi

src/commands/speech/synthesize.ts

@@ -92,10 +92,17 @@ export default defineCommand({
     if (flags.subtitles) body.subtitle_enable = true;  // Correct API parameter name
 
     if (flags.pronunciation) {
-      body.pronunciation_dict = (flags.pronunciation as string[]).map(p => {
-        const [from, to] = p.split('/');
-        return { tone: to || from!, text: from! };
-      });
+      const tones = flags.pronunciation as string[];
+      for (const p of tones) {
+        if (!p.includes('/')) {
+          throw new CLIError(
+            `Invalid --pronunciation value ${JSON.stringify(p)}: expected the form word/(pronunciation), e.g. 重/(chong2).`,
+            ExitCode.USAGE,
+            'mmx speech synthesize --text "我要去重庆出差。" --pronunciation "重/(chong2)"',
+          );
+        }
+      }
+      body.pronunciation_dict = { tone: tones };
     }
 
     if (dryRun(config, body)) return;

src/types/api.ts

@@ -105,7 +105,7 @@ export interface SpeechRequest {
     channel?: number;
   };
   language_boost?: string;
-  pronunciation_dict?: Array<{ tone: string; text: string }>;
+  pronunciation_dict?: { tone: string[] };
   output_format?: 'url' | 'hex';
   stream?: boolean;
   subtitle_enable?: boolean;  // Correct API parameter name (not 'subtitle')

test/commands/speech/synthesize.test.ts

@@ -202,6 +202,178 @@ describe('speech synthesize command', () => {
       console.log = originalLog;
     }
   });
+
+  it('--pronunciation uses the API-compatible request shape', async () => {
+    const config = {
+      apiKey: 'test-key',
+      region: 'global' as const,
+      baseUrl: 'https://api.mmx.io',
+      output: 'json' as const,
+      timeout: 10,
+      verbose: false,
+      quiet: false,
+      noColor: true,
+      yes: false,
+      dryRun: true,
+      nonInteractive: true,
+      async: false,
+    };
+
+    const originalLog = console.log;
+    let output = '';
+    console.log = (msg: string) => { output += msg; };
+
+    try {
+      await synthesizeCommand.execute(config, {
+        text: '我要去重庆出差。',
+        pronunciation: [
+          '重/(chong2)',
+          '行长/(hang2)(zhang3)',
+        ],
+        quiet: false,
+        verbose: false,
+        noColor: true,
+        yes: false,
+        dryRun: true,
+        help: false,
+        nonInteractive: true,
+        async: false,
+      });
+
+      const parsed = JSON.parse(output);
+
+      expect(parsed.request.pronunciation_dict).toEqual({
+        tone: [
+          '重/(chong2)',
+          '行长/(hang2)(zhang3)',
+        ],
+      });
+    } finally {
+      console.log = originalLog;
+    }
+  });
+
+  it('--pronunciation with a single value serializes under tone', async () => {
+    const config = {
+      apiKey: 'test-key',
+      region: 'global' as const,
+      baseUrl: 'https://api.mmx.io',
+      output: 'json' as const,
+      timeout: 10,
+      verbose: false,
+      quiet: false,
+      noColor: true,
+      yes: false,
+      dryRun: true,
+      nonInteractive: true,
+      async: false,
+    };
+
+    const originalLog = console.log;
+    let output = '';
+    console.log = (msg: string) => { output += msg; };
+
+    try {
+      await synthesizeCommand.execute(config, {
+        text: '我要去重庆出差。',
+        pronunciation: ['重/(chong2)'],
+        quiet: false,
+        verbose: false,
+        noColor: true,
+        yes: false,
+        dryRun: true,
+        help: false,
+        nonInteractive: true,
+        async: false,
+      });
+
+      const parsed = JSON.parse(output);
+
+      expect(parsed.request.pronunciation_dict).toEqual({
+        tone: ['重/(chong2)'],
+      });
+      // Must NOT be the old array-of-objects shape.
+      expect(Array.isArray(parsed.request.pronunciation_dict)).toBe(false);
+    } finally {
+      console.log = originalLog;
+    }
+  });
+
+  it('--pronunciation preserves multi-character word values verbatim', async () => {
+    const config = {
+      apiKey: 'test-key',
+      region: 'global' as const,
+      baseUrl: 'https://api.mmx.io',
+      output: 'json' as const,
+      timeout: 10,
+      verbose: false,
+      quiet: false,
+      noColor: true,
+      yes: false,
+      dryRun: true,
+      nonInteractive: true,
+      async: false,
+    };
+
+    const originalLog = console.log;
+    let output = '';
+    console.log = (msg: string) => { output += msg; };
+
+    try {
+      await synthesizeCommand.execute(config, {
+        text: '行长',
+        pronunciation: ['行长/(hang2)(zhang3)'],
+        quiet: false,
+        verbose: false,
+        noColor: true,
+        yes: false,
+        dryRun: true,
+        help: false,
+        nonInteractive: true,
+        async: false,
+      });
+
+      const parsed = JSON.parse(output);
+
+      expect(parsed.request.pronunciation_dict.tone).toEqual([
+        '行长/(hang2)(zhang3)',
+      ]);
+    } finally {
+      console.log = originalLog;
+    }
+  });
+
+  it('--pronunciation rejects malformed values without a separator', async () => {
+    const config = {
+      apiKey: 'test-key',
+      region: 'global' as const,
+      baseUrl: 'https://api.mmx.io',
+      output: 'json' as const,
+      timeout: 10,
+      verbose: false,
+      quiet: false,
+      noColor: true,
+      yes: false,
+      dryRun: true,
+      nonInteractive: true,
+      async: false,
+    };
+
+    await expect(
+      synthesizeCommand.execute(config, {
+        text: 'Hello',
+        pronunciation: ['chong2'],
+        quiet: false,
+        verbose: false,
+        noColor: true,
+        yes: false,
+        dryRun: true,
+        help: false,
+        nonInteractive: true,
+        async: false,
+      }),
+    ).rejects.toThrow('Invalid --pronunciation value "chong2"');
+  });
 });
 
 describe('speech synthesize format validation', () => {