v26: voice __init__.py + requirements-voice.txt

packages/voice/__init__.py

@@ -1,10 +1,4 @@
-"""packages/voice — Ultron V4 Voice Layer.
-
-Modules:
-  stt.py      — Whisper STT via Groq API (push-to-talk Discord audio -> text)
-  tts.py      — Kokoro TTS (text -> WAV bytes, CPU, ~0.3s for 100 tokens)
-  pipeline.py — Discord audio pipeline: receive opus -> decode -> STT -> Brain -> TTS -> send WAV
-
-Target latency: <1.5s end-to-end.
-Separate HF Space (ultron-voice) in Phase 6.
-"""
+# packages/voice/__init__.py
+# Ultron V4 — Voice package
+# Server: packages/voice/server.py (FastAPI, Groq Whisper STT + Kokoro TTS)
+# Deploy: docker/ultron-voice.Dockerfile

requirements-voice.txt

@@ -0,0 +1,27 @@
+# requirements-voice.txt
+# Ultron V4 — Voice Space dependencies
+# Deploy: docker/ultron-voice.Dockerfile (separate HF Space from Brain)
+# Install: pip install -r requirements-voice.txt
+
+# Web framework
+fastapi>=0.110.0
+uvicorn[standard]>=0.29.0
+
+# HTTP client (Brain /infer calls, health checks)
+httpx>=0.27.0
+
+# Groq SDK (Whisper STT — POST /stt)
+groq>=0.5.0
+
+# Kokoro TTS (POST /tts)
+# PyPI: https://pypi.org/project/kokoro/
+kokoro>=0.9.2
+
+# Audio processing (numpy for raw PCM ops if needed by Kokoro)
+numpy>=1.26.0
+
+# Pydantic v2 (FastAPI models)
+pydantic>=2.6.0
+
+# python-multipart (FastAPI file upload — /stt receives audio bytes)
+python-multipart>=0.0.9