Build Your First Voice Agent using Pipecat

1
import os
2
from dotenv import load_dotenv
3
from loguru import logger
4
from pipecat.frames.frames import LLMRunFrame
5
from pipecat.pipeline.pipeline import Pipeline
6
from pipecat.pipeline.runner import PipelineRunner
7
from pipecat.pipeline.task import PipelineTask
8
from pipecat.processors.aggregators.llm_context import LLMContext
9
from pipecat.processors.aggregators.llm_response_universal import (
10
    LLMContextAggregatorPair,
11
)
12
from pipecat.runner.types import RunnerArguments
13
from pipecat.runner.utils import create_transport
14
from pipecat.services.sarvam.stt import SarvamSTTService
15
from pipecat.services.sarvam.tts import SarvamTTSService
16
from pipecat.services.openai.llm import OpenAILLMService
17
from pipecat.transports.base_transport import TransportParams
18
from pipecat.transports.daily.transport import DailyParams
19
20
load_dotenv(override=True)
21
22
async def bot(runner_args: RunnerArguments):
23
    """Main bot entry point."""
24
    
25
    # Create transport (supports both Daily and WebRTC)
26
    transport = await create_transport(
27
        runner_args,
28
        {
29
            "daily": lambda: DailyParams(audio_in_enabled=True, audio_out_enabled=True),
30
            "webrtc": lambda: TransportParams(
31
                audio_in_enabled=True, audio_out_enabled=True
32
            ),
33
        },
34
    )
35
36
    # Initialize AI services
37
    stt = SarvamSTTService(api_key=os.getenv("SARVAM_API_KEY"))
38
    tts = SarvamTTSService(api_key=os.getenv("SARVAM_API_KEY"))
39
    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4o-mini")
40
41
    # Set up conversation context
42
    messages = [
43
        {
44
            "role": "system",
45
            "content": "You are a friendly AI assistant. Keep your responses brief and conversational.",
46
        },
47
    ]
48
    context = LLMContext(messages)
49
    context_aggregator = LLMContextAggregatorPair(context)
50
51
    # Build pipeline
52
    pipeline = Pipeline(
53
        [
54
            transport.input(),
55
            stt,
56
            context_aggregator.user(),
57
            llm,
58
            tts,
59
            transport.output(),
60
            context_aggregator.assistant(),
61
        ]
62
    )
63
64
    task = PipelineTask(pipeline)
65
66
    @transport.event_handler("on_client_connected")
67
    async def on_client_connected(transport, client):
68
        logger.info("Client connected")
69
        messages.append(
70
            {"role": "system", "content": "Say hello and briefly introduce yourself."}
71
        )
72
        await task.queue_frames([LLMRunFrame()])
73
74
    @transport.event_handler("on_client_disconnected")
75
    async def on_client_disconnected(transport, client):
76
        logger.info("Client disconnected")
77
        await task.cancel()
78
79
    runner = PipelineRunner(handle_sigint=runner_args.handle_sigint)
80
    await runner.run(task)
81
82
if __name__ == "__main__":
83
    from pipecat.runner.run import main
84
    main()

1
# Initialize AI services with Hindi support
2
stt = SarvamSTTService(
3
    api_key=os.getenv("SARVAM_API_KEY"),
4
    language="hi-IN",  # Hindi
5
    model="saarika:v2.5"
6
)
7
8
tts = SarvamTTSService(
9
    api_key=os.getenv("SARVAM_API_KEY"),
10
    target_language_code="hi-IN",
11
    model="bulbul:v2",
12
    speaker="manisha"  # Or: anushka, vidya, arya, abhilash, karun, hitesh
13
)
14
15
llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4o-mini")

1
stt = SarvamSTTService(
2
    api_key=os.getenv("SARVAM_API_KEY"),
3
    language="ta-IN",
4
    model="saarika:v2.5"
5
)
6
7
tts = SarvamTTSService(
8
    api_key=os.getenv("SARVAM_API_KEY"),
9
    target_language_code="ta-IN",
10
    model="bulbul:v2",
11
    speaker="anushka"
12
)
13
14
llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4o-mini")

1
# Auto-detect the user's language
2
stt = SarvamSTTService(
3
    api_key=os.getenv("SARVAM_API_KEY"),
4
    language="unknown",  # Auto-detects language
5
    model="saarika:v2.5"
6
)
7
8
tts = SarvamTTSService(
9
    api_key=os.getenv("SARVAM_API_KEY"),
10
    target_language_code="en-IN",
11
    model="bulbul:v2",
12
    speaker="karun"
13
)
14
15
llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4o-mini")

1
# User speaks Hindi → Saaras converts to English → LLM processes → Responds in English
2
3
stt = SarvamSTTService(
4
    api_key=os.getenv("SARVAM_API_KEY"),
5
    model="saaras:v2.5"  # Speech-to-English translation
6
)
7
8
tts = SarvamTTSService(
9
    api_key=os.getenv("SARVAM_API_KEY"),
10
    target_language_code="en-IN",
11
    model="bulbul:v2",
12
    speaker="abhilash"
13
)
14
15
llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4o-mini")

1
tts = SarvamTTSService(
2
    api_key=os.getenv("SARVAM_API_KEY"),
3
    target_language_code="en-IN",
4
    model="bulbul:v2",
5
    speaker="anushka",
6
    pitch=0.0,           # Range: -1.0 to 1.0
7
    pace=1.0,            # Range: 0.5 to 2.0
8
    loudness=1.5,        # Range: 0.5 to 2.0
9
    speech_sample_rate=16000  # 8000, 16000, or 24000 Hz
10
)