python-agents-examples/docs/examples/transcriber at main · AnsImran/python-agents-examples · GitHub

History

Name		Name	Last commit message	Last commit date
parent directory ..
README.md		README.md
transcriber.py		transcriber.py

README.md

title

Transcriber

Prerequisites

A .env at the repo root with your LiveKit credentials:

LIVEKIT_URL=your_livekit_url
LIVEKIT_API_KEY=your_api_key
LIVEKIT_API_SECRET=your_api_secret

Install dependencies:

pip install python-dotenv "livekit-agents[silero]"

Load configuration and create the AgentServer

Import the necessary modules and load environment variables. Create an AgentServer to handle incoming sessions.

import datetime
from dotenv import load_dotenv
from livekit.agents import JobContext, AgentServer, cli, Agent, AgentSession, inference

load_dotenv()

server = AgentServer()

Create an STT-only agent session

Start an AgentSession with only STT configured. The Agent is lightweight with just instructions—no TTS or LLM needed for pure transcription.

session = AgentSession(
    stt=inference.STT(model="deepgram/nova-3-general"),
)

await session.start(
    agent=Agent(instructions="You are a helpful assistant that transcribes user speech to text."),
    room=ctx.room
)

Listen for final transcripts

Subscribe to user_input_transcribed and append each final transcript to user_speech_log.txt with a timestamp.

@session.on("user_input_transcribed")
def on_transcript(transcript):
    if transcript.is_final:
        timestamp = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
        with open("user_speech_log.txt", "a") as f:
            f.write(f"[{timestamp}] {transcript.transcript}\n")

Create the RTC session entrypoint

Wire it all together in the entrypoint so the agent begins listening immediately when the session starts.

@server.rtc_session()
async def entrypoint(ctx: JobContext):
    ctx.log_context_fields = {"room": ctx.room.name}

    session = AgentSession(
        stt=inference.STT(model="deepgram/nova-3-general"),
    )

    @session.on("user_input_transcribed")
    def on_transcript(transcript):
        if transcript.is_final:
            timestamp = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
            with open("user_speech_log.txt", "a") as f:
                f.write(f"[{timestamp}] {transcript.transcript}\n")

    await session.start(
        agent=Agent(instructions="You are a helpful assistant that transcribes user speech to text."),
        room=ctx.room
    )
    await ctx.connect()

Run it

python transcriber.py console

The agent starts listening right away and logs transcriptions to user_speech_log.txt.

How it works

Deepgram STT streams audio and emits user_input_transcribed events.
Each final transcript is timestamped and appended to a log file.
Because there is no LLM/TTS, the agent never speaks; it only records.
The rest of the session lifecycle is handled by AgentSession.

Log file format

[2024-01-15 14:30:45] Hello, this is my first transcription
[2024-01-15 14:30:52] Testing the speech to text functionality

Full example

import datetime
from dotenv import load_dotenv
from livekit.agents import JobContext, AgentServer, cli, Agent, AgentSession, inference

load_dotenv()

server = AgentServer()

@server.rtc_session()
async def entrypoint(ctx: JobContext):
    ctx.log_context_fields = {"room": ctx.room.name}

    session = AgentSession(
        stt=inference.STT(model="deepgram/nova-3-general"),
    )

    @session.on("user_input_transcribed")
    def on_transcript(transcript):
        if transcript.is_final:
            timestamp = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
            with open("user_speech_log.txt", "a") as f:
                f.write(f"[{timestamp}] {transcript.transcript}\n")

    await session.start(
        agent=Agent(instructions="You are a helpful assistant that transcribes user speech to text."),
        room=ctx.room
    )
    await ctx.connect()

if __name__ == "__main__":
    cli.run_app(server)

python-agents-examples/docs/examples/transcriber at main · AnsImran/python-agents-examples · GitHub

Expand file tree

Breadcrumbs

transcriber

Directory actions

More options