xfyun-sdk Roadmap

This document tracks planned features that are not yet implemented in the current release.

🚧 Items listed here are user-facing commitments, not guarantees. They represent intended direction. ETA is omitted to avoid implying a release date that may slip.

---

🚧 Planned (Not Yet Implemented)

WeChat Mini Program Support

Status: Not implemented Effort estimate: ~3-5 days Why deferred: WebSocket + MediaRecorder APIs in mini-program environment differ significantly from browser (uses wx.connectSocket + RecorderManager). Requires a parallel adapter layer in src/platforms/miniprogram/.

What it would entail:

• PlatformAdapter interface abstraction in core (currently hardcoded to browser globals)
• BrowserAdapter (current implementation) + MiniProgramAdapter (new)
• WebSocket polyfill using wx.connectSocket (signature-compatible with browser WebSocket)
• Audio capture using wx.getRecorderManager() instead of MediaRecorder
• Audio playback using wx.createInnerAudioContext() instead of HTMLAudioElement
• Build pipeline: separate rollup config for mini-program output (CommonJS only, no UMD)

Current alternative: Users targeting mini-program can use the core classes (XfyunASR / XfyunTTS / XfyunTranslator) if they provide a WebSocket-compatible transport. The recording and audio playback parts still require mini-program-specific adapters.

---

Vue 3 Composable Distribution

Status: Examples provided, no published package Effort estimate: ~1 day

The Vue composables (useSpeechRecognizer, useSpeechSynthesizer, useTranslator) live in examples/vue-demo/src/composables/. They are functional and tested manually, but:

• Not published to npm
• No version pinning (changes with the SDK may break them)
• No separate test coverage

Future plan: Either (a) move them to a separate xfyun-sdk-vue package, or (b) document them as copy-paste patterns with stability guarantees. Decision deferred to user feedback.

---

TTS Audio Playback Helper

Status: Not implemented Effort estimate: ~0.5 day

Currently SpeechSynthesizer returns audio chunks via onAudioData callback. Users have to wire up their own <audio> element / Audio() instance to play.

A built-in playAudio() method (or auto-play option) on XfyunTTS would streamline the most common use case. Tradeoff: locks the SDK to browser Audio API (which already works fine).

---

Stable Major Version (v2.0.0)

Status: Blocked on error code migration completion Target: After one release cycle with legacyCode field

v2.0 will:

• Remove legacyCode field (breaking change completion)
• Remove XFUN_ERROR_CODES alias (use XFUN_SERVER_ERROR_CODES only)
• Tighten Translator.tsx domain type to union type
• Drop Node.js 14 / 16 support (target Node 18+)

---

✅ Recently Shipped

See changelog.md for the full history.