feat: OpenAI Realtime API pipeline

Replaced the 3-step sequential pipeline (Whisper STT → DeepSeek LLM
→ OpenAI TTS) with a single OpenAI Realtime API WebSocket using
gpt-4o-mini-realtime-preview.

- ~300-800ms latency vs 1-3s
- Server VAD for automatic turn detection
- Streaming audio chunks during playback
- Interruptions: user can speak over Kira mid-response
- Honcho memory still injected into session instructions
- Frontend captures PCM16 mono 24kHz via AudioContext
- Backend relays client ↔ OpenAI Realtime API
- Supports both voice (PCM16) and text input
This commit is contained in:
2026-06-04 13:32:39 -04:00
parent e64698b0ab
commit e2332af8d0
4 changed files with 551 additions and 251 deletions
+183 -172
View File
@@ -1,26 +1,21 @@
"""Kira — AI body double backend
Real-time speech-to-speech pipeline:
mic audio → Whisper API → text → DeepSeek LLM → response text → OpenAI TTS → audio
Honcho memory integration:
Cross-session user context injected into LLM prompts,
conversation exchanges stored for continuous learning.
User preferences (name, scene, outfit, accessory) persisted in peer metadata.
OpenAI Realtime API pipeline:
mic audio → [built-in STT → GPT-4o-mini → built-in TTS] speaker audio
Single WebSocket, ~300-800ms latency
"""
import json
import base64
import uuid
import logging
import asyncio
from fastapi import FastAPI, WebSocket, WebSocketDisconnect
from fastapi.middleware.cors import CORSMiddleware
from config import settings
from services.stt import transcribe_audio
from services.llm import get_kira_response
from services.tts import synthesize_speech
from services.realtime import RealtimeRelay
from services.memory import kira_memory
logging.basicConfig(level=logging.INFO)
@@ -36,24 +31,13 @@ app.add_middleware(
allow_headers=["*"],
)
# ─── Base system prompt (static part) ───
BASE_SYSTEM_PROMPT = (
"You are Kira, a warm, kind, and encouraging AI body double. "
"You speak in a friendly, girly-pop tone. You are helping someone with ADHD "
"stay focused and on task. Keep responses short, supportive, and uplifting. "
"Check in on them. Remind them to take breaks. Celebrate small wins. "
"Use occasional emoji but don't overdo it. Never be judgmental. "
"You remember things about them between conversations."
)
@app.on_event("startup")
async def startup():
"""Initialize Honcho memory on app startup."""
if kira_memory.init():
logger.info("Honcho memory initialized")
else:
logger.info("Honcho memory not configured — running without memory")
logger.info("Honcho memory not configured")
@app.get("/api/health")
@@ -62,61 +46,6 @@ async def health():
return {"status": "ok", "name": "kira", "memory": mem_status}
def build_system_prompt(user_id: str) -> dict:
"""Build system prompt with Honcho memory context injected."""
base = BASE_SYSTEM_PROMPT
if kira_memory.enabled:
try:
kira_memory.ensure_peers(user_id)
memory_suffix = kira_memory.build_system_prompt_suffix()
if memory_suffix:
base += memory_suffix
except Exception as e:
logger.warning(f"Failed to build memory context: {e}")
return {"role": "system", "content": base}
def handle_identify(msg: dict, session_id: str) -> dict | None:
"""Handle user identification. Returns user preferences or None."""
user_id = msg.get("user_id", "").strip()
if not user_id:
return {"type": "error", "message": "user_id is required"}
user_name = msg.get("name", "").strip()
if user_name:
kira_memory.set_user_preference(user_id, "name", user_name)
prefs = kira_memory.get_user_preferences(user_id)
logger.info(f"[{session_id}] Identified as {user_id} (name={user_name or prefs.get('name', '')})")
return {
"type": "identified",
"user_id": user_id,
"preferences": prefs,
}
def handle_set_preference(msg: dict, session_id: str, user_id: str) -> dict | None:
"""Handle preference update. Returns success status."""
if not user_id or user_id == "default-user":
return {"type": "error", "message": "Must identify first"}
key = msg.get("key", "").strip()
value = msg.get("value", "").strip()
if not key:
return {"type": "error", "message": "key is required"}
ok = kira_memory.set_user_preference(user_id, key, value)
return {
"type": "preference_saved",
"key": key,
"success": ok,
}
@app.websocket("/api/ws")
async def conversation_ws(websocket: WebSocket):
await websocket.accept()
@@ -125,8 +54,102 @@ async def conversation_ws(websocket: WebSocket):
identified = False
logger.info(f"[{session_id}] WebSocket connected")
audio_buffer = bytearray()
conversation_history: list[dict] = []
# Track conversation for Honcho
pending_transcripts: list[tuple[str, str]] = []
# Will be set when Realtime relay is ready
relay_ready = asyncio.Event()
relay: RealtimeRelay | None = None
relay_task: asyncio.Task | None = None
audio_queue: asyncio.Queue[bytes] = asyncio.Queue()
text_queue: asyncio.Queue[str] = asyncio.Queue()
async def on_ready():
relay_ready.set()
logger.info(f"[{session_id}] Realtime relay ready")
async def on_audio_delta(audio_bytes: bytes):
"""Forward audio chunks from OpenAI to the client."""
try:
audio_b64 = base64.b64encode(audio_bytes).decode("utf-8")
await websocket.send_json({
"type": "audio",
"data": audio_b64,
})
except Exception:
pass
async def on_transcript(text: str):
"""Store transcripts for Honcho."""
pending_transcripts.append(("transcript", text))
role, content = text.split(": ", 1)
logger.info(f"[{session_id}] {role}: {content}")
await websocket.send_json({
"type": "transcript",
"role": role,
"text": content,
})
async def on_speech_started():
"""Kira started speaking."""
await websocket.send_json({"type": "speaking_start"})
async def on_speech_stopped():
"""Kira finished speaking."""
await websocket.send_json({"type": "speaking_end"})
async def on_interruption():
"""User interrupted — Kira stops speaking."""
await websocket.send_json({"type": "interruption"})
async def on_error(msg: str):
await websocket.send_json({"type": "error", "message": msg})
# ── Create and start the Realtime relay ──
relay = RealtimeRelay(
on_audio_delta=on_audio_delta,
on_transcript=on_transcript,
on_speech_started=on_speech_started,
on_speech_stopped=on_speech_stopped,
on_interruption=on_interruption,
on_error=on_error,
on_ready=on_ready,
)
relay_task = asyncio.create_task(relay.connect())
# Wait for relay to be ready
try:
await asyncio.wait_for(relay_ready.wait(), timeout=15)
except asyncio.TimeoutError:
logger.error(f"[{session_id}] Realtime relay failed to connect")
await websocket.send_json({"type": "error", "message": "Failed to connect to AI"})
relay_task.cancel()
return
# ── Forward audio/text from client to relay ──
async def forward_audio():
while relay and relay._connected:
try:
pcm16 = await asyncio.wait_for(audio_queue.get(), timeout=1)
await relay.send_audio(pcm16)
except asyncio.TimeoutError:
continue
except Exception:
break
async def forward_text():
while relay and relay._connected:
try:
text = await asyncio.wait_for(text_queue.get(), timeout=1)
await relay.send_text(text)
except asyncio.TimeoutError:
continue
except Exception:
break
fwd_audio_task = asyncio.create_task(forward_audio())
fwd_text_task = asyncio.create_task(forward_text())
try:
while True:
@@ -134,110 +157,98 @@ async def conversation_ws(websocket: WebSocket):
msg = json.loads(raw)
msg_type = msg.get("type", "")
# ── Identity & Preferences ──
# ── Identity ──
if msg_type == "identify":
response = handle_identify(msg, session_id)
if response:
await websocket.send_json(response)
if response["type"] == "identified":
user_id = response["user_id"]
identified = True
# Set up Honcho for this user
if kira_memory.enabled:
try:
kira_memory.ensure_peers(user_id)
kira_memory.ensure_session(session_id)
logger.info(f"[{session_id}] Honcho session ready for {user_id}")
except Exception as e:
logger.warning(f"[{session_id}] Honcho setup failed: {e}")
user_id = msg.get("user_id", "").strip()
user_name = msg.get("name", "").strip()
if user_name and user_id:
kira_memory.set_user_preference(user_id, "name", user_name)
prefs = kira_memory.get_user_preferences(user_id)
identified = True
if kira_memory.enabled:
kira_memory.ensure_peers(user_id)
kira_memory.ensure_session(session_id)
# Inject Honcho context into the Realtime session instructions
memory_suffix = ""
if kira_memory.enabled:
try:
ctx = kira_memory.build_system_prompt_suffix()
if ctx:
memory_suffix = ctx
except Exception:
pass
if relay and relay._connected and memory_suffix:
await relay._send({
"type": "session.update",
"session": {
"instructions": (
"You are Kira, a warm, kind, and encouraging AI body double. "
"Speak in a friendly, girly-pop tone. Help someone with ADHD "
"stay focused. Keep responses short and supportive. "
"Check in, remind breaks, celebrate wins. Never judgmental."
+ memory_suffix
),
},
})
await websocket.send_json({
"type": "identified",
"user_id": user_id,
"preferences": prefs,
})
continue
# ── Preferences ──
if msg_type == "set_preference":
response = handle_set_preference(msg, session_id, user_id)
if response:
await websocket.send_json(response)
key = msg.get("key", "").strip()
value = msg.get("value", "").strip()
if key and user_id and user_id != "default-user":
kira_memory.set_user_preference(user_id, key, value)
continue
# ── Conversation ──
# ── Audio from frontend (PCM16) ──
if msg_type == "audio":
audio_b64 = msg.get("data", "")
if audio_b64:
pcm16 = base64.b64decode(audio_b64)
await audio_queue.put(pcm16)
continue
system_prompt = build_system_prompt(user_id)
# ── Text input ──
if msg_type == "conversation_text":
text = msg.get("text", "").strip()
if text:
await text_queue.put(text)
# Also store in Honcho immediately
if kira_memory.enabled and identified:
kira_memory.store_user_message(text)
continue
if msg_type == "audio_chunk":
chunk = base64.b64decode(msg["data"])
audio_buffer.extend(chunk)
elif msg_type == "transcribe":
if not audio_buffer:
await websocket.send_json({"type": "error", "message": "No audio data"})
continue
logger.info(f"[{session_id}] Transcribing {len(audio_buffer)} bytes...")
transcript = await transcribe_audio(bytes(audio_buffer))
audio_buffer.clear()
if not transcript:
await websocket.send_json({"type": "error", "message": "Could not transcribe audio"})
continue
await websocket.send_json({"type": "transcript", "text": transcript})
logger.info(f"[{session_id}] User: {transcript}")
conversation_history.append({"role": "user", "content": transcript})
messages = [system_prompt] + conversation_history[-10:]
kira_text = await get_kira_response(messages)
conversation_history.append({"role": "assistant", "content": kira_text})
logger.info(f"[{session_id}] Kira: {kira_text}")
if kira_memory.enabled and identified:
try:
kira_memory.store_messages(transcript, kira_text)
except Exception as e:
logger.warning(f"[{session_id}] Failed to store messages: {e}")
await websocket.send_json({"type": "speaking_start", "text": kira_text})
audio_bytes = await synthesize_speech(kira_text)
audio_b64 = base64.b64encode(audio_bytes).decode("utf-8")
await websocket.send_json({"type": "audio", "data": audio_b64, "text": kira_text})
await websocket.send_json({"type": "speaking_end"})
elif msg_type == "ping":
if msg_type == "ping":
await websocket.send_json({"type": "pong"})
elif msg_type == "conversation_text":
user_text = msg.get("text", "").strip()
if not user_text:
continue
logger.info(f"[{session_id}] User (text): {user_text}")
conversation_history.append({"role": "user", "content": user_text})
messages = [system_prompt] + conversation_history[-10:]
kira_text = await get_kira_response(messages)
conversation_history.append({"role": "assistant", "content": kira_text})
logger.info(f"[{session_id}] Kira: {kira_text}")
if kira_memory.enabled and identified:
try:
kira_memory.store_messages(user_text, kira_text)
except Exception as e:
logger.warning(f"[{session_id}] Failed to store messages: {e}")
await websocket.send_json({"type": "speaking_start", "text": kira_text})
audio_bytes = await synthesize_speech(kira_text)
audio_b64 = base64.b64encode(audio_bytes).decode("utf-8")
await websocket.send_json({"type": "audio", "data": audio_b64, "text": kira_text})
await websocket.send_json({"type": "speaking_end"})
except WebSocketDisconnect:
logger.info(f"[{session_id}] Disconnected")
except Exception as e:
logger.error(f"[{session_id}] Error: {e}")
try:
await websocket.send_json({"type": "error", "message": str(e)})
except Exception:
pass
finally:
# Store pending transcripts in Honcho
if kira_memory.enabled and identified:
for _, transcript_text in pending_transcripts:
if transcript_text.startswith("user: "):
content = transcript_text[6:]
kira_memory.store_user_message(content)
elif transcript_text.startswith("assistant: "):
content = transcript_text[11:]
kira_memory.store_kira_message(content)
fwd_audio_task.cancel()
fwd_text_task.cancel()
if relay:
await relay.disconnect()
if relay_task:
relay_task.cancel()