140 lines
4.5 KiB
Python
140 lines
4.5 KiB
Python
#
|
|
# Copyright (c) 2024-2025, Daily
|
|
#
|
|
# SPDX-License-Identifier: BSD 2-Clause License
|
|
#
|
|
|
|
import argparse
|
|
import os
|
|
import sys
|
|
from pathlib import Path
|
|
|
|
from dotenv import load_dotenv
|
|
from loguru import logger
|
|
from openai import audio
|
|
|
|
from pipecat.audio.vad.silero import SileroVADAnalyzer
|
|
from pipecat.frames.frames import Frame
|
|
from pipecat.pipeline.pipeline import Pipeline
|
|
from pipecat.pipeline.runner import PipelineRunner
|
|
from pipecat.pipeline.task import PipelineParams, PipelineTask
|
|
from pipecat.processors.aggregators.openai_llm_context import OpenAILLMContext
|
|
from pipecat.processors.frame_processor import FrameDirection, FrameProcessor
|
|
from pipecat.services.cartesia.tts import CartesiaTTSService
|
|
from pipecat.services.deepgram.stt import DeepgramSTTService
|
|
from pipecat.services.google.llm import GoogleLLMService, LLMSearchResponseFrame
|
|
from pipecat.transports.base_transport import TransportParams
|
|
from pipecat.transports.network.small_webrtc import SmallWebRTCTransport
|
|
from pipecat.transports.network.webrtc_connection import SmallWebRTCConnection
|
|
|
|
sys.path.append(str(Path(__file__).parent.parent))
|
|
|
|
load_dotenv(override=True)
|
|
|
|
|
|
# Function handlers for the LLM
|
|
search_tool = {"google_search_retrieval": {}}
|
|
tools = [search_tool]
|
|
|
|
system_instruction = """
|
|
You are an expert at providing the most recent news from any place. Your responses will be converted to audio, so avoid using special characters or overly complex formatting.
|
|
|
|
Always use the google search API to retrieve the latest news. You must also use it to check which day is today.
|
|
|
|
You can:
|
|
- Use the Google search API to check the current date.
|
|
- Provide the most recent and relevant news from any place by using the google search API.
|
|
- Answer any questions the user may have, ensuring your responses are accurate and concise.
|
|
|
|
Start each interaction by asking the user about which place they would like to know the information.
|
|
"""
|
|
|
|
|
|
class LLMSearchLoggerProcessor(FrameProcessor):
|
|
async def process_frame(self, frame: Frame, direction: FrameDirection):
|
|
await super().process_frame(frame, direction)
|
|
|
|
if isinstance(frame, LLMSearchResponseFrame):
|
|
print(f"LLMSearchLoggerProcessor: {frame}")
|
|
|
|
await self.push_frame(frame)
|
|
|
|
|
|
async def run_bot(webrtc_connection: SmallWebRTCConnection, _: argparse.Namespace):
|
|
logger.info(f"Starting bot")
|
|
|
|
transport = SmallWebRTCTransport(
|
|
webrtc_connection=webrtc_connection,
|
|
params=TransportParams(
|
|
audio_in_enabled=True,
|
|
audio_out_enabled=True,
|
|
vad_analyzer=SileroVADAnalyzer(),
|
|
),
|
|
)
|
|
|
|
stt = DeepgramSTTService(api_key=os.getenv("DEEPGRAM_API_KEY"))
|
|
|
|
tts = CartesiaTTSService(
|
|
api_key=os.getenv("CARTESIA_API_KEY"),
|
|
voice_id="71a7ad14-091c-4e8e-a314-022ece01c121", # British Reading Lady
|
|
)
|
|
|
|
# Initialize the Gemini Multimodal Live model
|
|
llm = GoogleLLMService(
|
|
api_key=os.getenv("GOOGLE_API_KEY"),
|
|
system_instruction=system_instruction,
|
|
tools=tools,
|
|
model="gemini-1.5-flash-002",
|
|
)
|
|
|
|
context = OpenAILLMContext(
|
|
[
|
|
{
|
|
"role": "user",
|
|
"content": "Start by greeting the user warmly, introducing yourself, and mentioning the current day. Be friendly and engaging to set a positive tone for the interaction.",
|
|
}
|
|
],
|
|
)
|
|
context_aggregator = llm.create_context_aggregator(context)
|
|
|
|
llm_search_logger = LLMSearchLoggerProcessor()
|
|
|
|
pipeline = Pipeline(
|
|
[
|
|
transport.input(),
|
|
stt,
|
|
context_aggregator.user(),
|
|
llm,
|
|
llm_search_logger,
|
|
tts,
|
|
transport.output(),
|
|
context_aggregator.assistant(),
|
|
]
|
|
)
|
|
|
|
task = PipelineTask(pipeline, params=PipelineParams(allow_interruptions=True))
|
|
|
|
@transport.event_handler("on_client_connected")
|
|
async def on_client_connected(transport, client):
|
|
logger.info(f"Client connected")
|
|
# Start conversation - empty prompt to let LLM follow system instructions
|
|
await task.queue_frames([context_aggregator.user().get_context_frame()])
|
|
|
|
@transport.event_handler("on_client_disconnected")
|
|
async def on_client_disconnected(transport, client):
|
|
logger.info(f"Client disconnected")
|
|
|
|
@transport.event_handler("on_client_closed")
|
|
async def on_client_closed(transport, client):
|
|
logger.info(f"Client closed connection")
|
|
await task.cancel()
|
|
|
|
runner = PipelineRunner(handle_sigint=False)
|
|
await runner.run(task)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
from run import main
|
|
|
|
main()
|