feat: add per-modality default model configuration

Users can now set default models in config.json for each modality,
avoiding the need to pass --model every time.

Config keys (hyphen aliases accepted):
  - default_text_model   (default-text-model)
  - default_speech_model (default-speech-model)
  - default_video_model  (default-video-model)
  - default_music_model  (default-music-model)

Resolution priority:
  1. Explicit --model flag
  2. Config file default
  3. Hardcoded fallback

Music cover is special: defaultMusicModel only applies if it's a
valid cover model name (music-cover or music-cover-free), otherwise
the key-type-based default is preserved.

Tests added for config set/show, text chat, speech synthesize,
video generate, music generate, and model resolution logic.
97 tests pass.

Author: raylanlin

speech_2026-04-11-17-52-51.mp3

@@ -6,14 +6,22 @@ import { readConfigFile, writeConfigFile } from '../../config/loader';
 import type { Config } from '../../config/schema';
 import type { GlobalFlags } from '../../types/flags';
 
-const VALID_KEYS = ['region', 'base_url', 'output', 'timeout', 'api_key'];
+const VALID_KEYS = ['region', 'base_url', 'output', 'timeout', 'api_key', 'default_text_model', 'default_speech_model', 'default_video_model', 'default_music_model'];
+
+// Allow hyphen-style keys (e.g. default-text-model → default_text_model)
+const KEY_ALIASES: Record<string, string> = {
+  'default-text-model': 'default_text_model',
+  'default-speech-model': 'default_speech_model',
+  'default-video-model': 'default_video_model',
+  'default-music-model': 'default_music_model',
+};
 
 export default defineCommand({
   name: 'config set',
   description: 'Set a config value',
   usage: 'mmx config set --key <key> --value <value>',
   options: [
-    { flag: '--key <key>', description: 'Config key (region, base_url, output, timeout, api_key)' },
+    { flag: '--key <key>', description: 'Config key (region, base_url, output, timeout, api_key, default_text_model, default_speech_model, default_video_model, default_music_model)' },
     { flag: '--value <value>', description: 'Value to set' },
   ],
   examples: [
@@ -25,6 +33,9 @@ export default defineCommand({
     const key = flags.key as string | undefined;
     const value = flags.value as string | undefined;
 
+    // Resolve hyphen aliases to underscore keys
+    const resolvedKey = KEY_ALIASES[key] || key;
+
     if (!key || value === undefined) {
       throw new CLIError(
         '--key and --value are required.',
@@ -33,29 +44,29 @@ export default defineCommand({
       );
     }
 
-    if (!VALID_KEYS.includes(key)) {
+    if (!VALID_KEYS.includes(resolvedKey)) {
       throw new CLIError(
         `Invalid config key "${key}". Valid keys: ${VALID_KEYS.join(', ')}`,
         ExitCode.USAGE,
       );
     }
 
     // Validate specific values
-    if (key === 'region' && !['global', 'cn'].includes(value)) {
+    if (resolvedKey === 'region' && !['global', 'cn'].includes(value)) {
       throw new CLIError(
         `Invalid region "${value}". Valid values: global, cn`,
         ExitCode.USAGE,
       );
     }
 
-    if (key === 'output' && !['text', 'json'].includes(value)) {
+    if (resolvedKey === 'output' && !['text', 'json'].includes(value)) {
       throw new CLIError(
         `Invalid output format "${value}". Valid values: text, json`,
         ExitCode.USAGE,
       );
     }
 
-    if (key === 'timeout') {
+    if (resolvedKey === 'timeout') {
       const num = Number(value);
       if (isNaN(num) || num <= 0) {
         throw new CLIError(
@@ -68,16 +79,16 @@ export default defineCommand({
     const format = detectOutputFormat(config.output);
 
     if (config.dryRun) {
-      console.log(formatOutput({ would_set: { [key]: value } }, format));
+      console.log(formatOutput({ would_set: { [resolvedKey]: value } }, format));
       return;
     }
 
     const existing = readConfigFile() as Record<string, unknown>;
-    existing[key] = key === 'timeout' ? Number(value) : value;
+    existing[resolvedKey] = resolvedKey === 'timeout' ? Number(value) : value;
     await writeConfigFile(existing);
 
     if (!config.quiet) {
-      console.log(formatOutput({ [key]: existing[key] }, format));
+      console.log(formatOutput({ [resolvedKey]: existing[resolvedKey] }, format));
     }
   },
 });

speech_2026-04-11-17-53-11.mp3

@@ -31,6 +31,12 @@ export default defineCommand({
       result.api_key = maskToken(file.api_key);
     }
 
+    // Default models
+    if (file.default_text_model) result.default_text_model = file.default_text_model;
+    if (file.default_speech_model) result.default_speech_model = file.default_speech_model;
+    if (file.default_video_model) result.default_video_model = file.default_video_model;
+    if (file.default_music_model) result.default_music_model = file.default_music_model;
+
     console.log(formatOutput(result, format));
   },
 });

speech_2026-04-11-17-53-17.mp3

@@ -10,9 +10,17 @@ export function isCodingPlan(config: Config): boolean {
 }
 
 export function musicGenerateModel(config: Config): string {
+  // Config default > key-type-based default
+  if (config.defaultMusicModel) return config.defaultMusicModel;
   return isCodingPlan(config) ? 'music-2.6' : 'music-2.6-free';
 }
 
+const VALID_COVER_MODELS = new Set(['music-cover', 'music-cover-free']);
+
 export function musicCoverModel(config: Config): string {
+  // Config default (only if it's a valid cover model) > key-type-based default
+  if (config.defaultMusicModel && VALID_COVER_MODELS.has(config.defaultMusicModel)) {
+    return config.defaultMusicModel;
+  }
   return isCodingPlan(config) ? 'music-cover' : 'music-cover-free';
 }

speech_2026-04-11-17-55-45.mp3

@@ -55,7 +55,9 @@ export default defineCommand({
       );
     }
 
-    const model = (flags.model as string) || 'speech-2.8-hd';
+    const model = (flags.model as string)
+      || config.defaultSpeechModel
+      || 'speech-2.8-hd';
     const voice = (flags.voice as string) || 'English_expressive_narrator';
     const ts = new Date().toISOString().slice(0, 19).replace(/[T:]/g, '-');
     const ext = (flags.format as string) || 'mp3';

src/commands/config/set.ts

@@ -117,7 +117,9 @@ export default defineCommand({
       }
     }
 
-    const model = (flags.model as string) || 'MiniMax-M2.7';
+    const model = (flags.model as string)
+      || config.defaultTextModel
+      || 'MiniMax-M2.7';
     const shouldStream = flags.stream === true || (flags.stream === undefined && process.stdout.isTTY);
     const format = detectOutputFormat(config.output);
 

src/commands/config/show.ts

@@ -71,13 +71,17 @@ export default defineCommand({
       );
     }
 
-    // Determine model: explicit --model overrides auto-switch
+    // Determine model: explicit --model > auto-switch > config default > hardcoded
     const explicitModel = flags.model as string | undefined;
-    let model = explicitModel || 'MiniMax-Hailuo-2.3';
-    if (!explicitModel && flags.lastFrame) {
+    let model: string;
+    if (explicitModel) {
+      model = explicitModel;
+    } else if (flags.lastFrame) {
       model = 'MiniMax-Hailuo-02';
-    } else if (!explicitModel && flags.subjectImage) {
+    } else if (flags.subjectImage) {
       model = 'S2V-01';
+    } else {
+      model = config.defaultVideoModel || 'MiniMax-Hailuo-2.3';
     }
     const format = detectOutputFormat(config.output);