Robo Pest app: voice -> STT relay (nightmare GPU) -> WorkBrain RAG -> lens display. STT relay service included. Verified end-to-end: speech->answer with citations in ~12s

apps/robo-pest/server/stt_relay.py

@@ -0,0 +1,45 @@
+#!/usr/bin/env python3
+"""Robo Pest STT relay — faster-whisper on nightmare's RTX 4080 SUPER.
+
+POST /transcribe  (body: raw PCM 16kHz s16le mono bytes)  -> {"text": "..."}
+GET  /health -> {"status": "ok", "model": ...}
+"""
+import io
+import os
+import struct
+import numpy as np
+from fastapi import FastAPI, Request
+from faster_whisper import WhisperModel
+
+MODEL_NAME = os.environ.get("STT_MODEL", "small.en")
+DEVICE = os.environ.get("STT_DEVICE", "cuda")
+
+app = FastAPI()
+model = WhisperModel(MODEL_NAME, device=DEVICE, compute_type="float16")
+
+
+def pcm16_to_float32(data: bytes) -> np.ndarray:
+    n = len(data) // 2
+    ints = struct.unpack(f"<{n}h", data[: n * 2])
+    return np.array(ints, dtype=np.float32) / 32768.0
+
+
+@app.get("/health")
+def health():
+    return {"status": "ok", "model": MODEL_NAME, "device": DEVICE}
+
+
+@app.post("/transcribe")
+async def transcribe(request: Request):
+    body = await request.body()
+    if not body:
+        return {"text": ""}
+    audio = pcm16_to_float32(body)
+    segments, info = model.transcribe(audio, language="en", beam_size=1, vad_filter=True)
+    text = " ".join(s.text.strip() for s in segments).strip()
+    return {"text": text, "duration": round(info.duration, 2)}
+
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=8765)