hermes - 💡(How to fix) Fix xAI TTS: missing speed parameter and voice discovery

Summary

The xAI TTS backend (_generate_xai_tts) is missing two features that the xAI API supports:

1. No speed parameter

xAI /v1/tts endpoint accepts a speed field (tested with 1.5x and 2.0x, both HTTP 200). Hermes does not pass it through, making the voice always speak at default speed.

Edge, OpenAI, MiniMax, and KittenTTS backends all support speed. xAI should too.

Fix: one-line addition in payload construction:

speed = float(xai_config.get("speed", 1.0))
payload["speed"] = speed

Config key: tts.xai.speed (float, default 1.0)

2. No voice catalog

xAI offers 5 distinct TTS voices: Ara, Eve, Leo, Rex, Sal (per xAI docs). Each has distinct personality/tone.

Hermes hardcodes DEFAULT_XAI_VOICE_ID = "eve" with no discovery mechanism. The /v1/tts/voices and /v1/voices endpoints return 404, so there may not be a list endpoint — but at minimum the 5 known voices should be documented and configurable.

Context

• xAI OAuth (SuperGrok) is the preferred auth path, and works for TTS
• All other major TTS backends support speed
• Voice switching via tts.xai.voice_id config works but users have no way to discover available voices