how to implement call recording in my livekit agent. ?

Thanks for reaching out, Saurav. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

so once recording is started, will it automatically stop when call end?

Absolutely! LiveKit can automatically stop recording when a call ends, depending on how you set up the egress: • RoomComposite Egress: The recording will stop on its own once everyone leaves the room. • Participant Egress: The recording stops automatically when that specific participant leaves the room. Just a quick tip: make sure the room is properly closed when the call wraps up to keep everything running smoothly. If you have any questions or want me to explain anything further, just let me know—I’m here to help! Sources: Composite and web recordings | LiveKit Docs | Recording Participants | LiveKit Docs

refactor this code to save recordings to supabase storage

req = api.RoomCompositeEgressRequest(

room_name="my-room",

layout="speaker",

preset=api.EncodingOptionsPreset.H264_720P_30,

audio_only=False,

segment_outputs=[api.SegmentedFileOutput(

filename_prefix="my-output",

playlist_name="my-playlist.m3u8",

live_playlist_name="my-live-playlist.m3u8",

segment_duration=5,

gcp=api.GCPUpload(

credentials=file_contents,

bucket="<my-bucket>",

),

)],

)

To save recordings to Supabase Storage, you can set up the

RoomCompositeEgressRequest

to use the S3-compatible storage option. Here’s a handy refactor of the code for you: Sources: protobufs/livekit_egress.proto | README.md | livekit-protocol/livekit/protocol/egress.py

so how file will be saved like with what names?

Just to give you a quick rundown on how the file naming works for your recordings: it all comes down to the

filename_prefix

and the way you set up the

SegmentedFileOutput

. Here’s the scoop: • Filename Prefix: The

filename_prefix

you choose (like

recording_file_name_prefix

) becomes the base name for your files. • Segmented Files: Since you’re using segmented outputs, your recordings will be split into multiple files with names in order, such as

recording_file_name_prefix_00001.ts

,

recording_file_name_prefix_00002.ts

, and so forth. • Directory Structure: If your prefix includes a folder path (for example,

recordings/my-room

), the files will be saved right there in that folder inside your S3 bucket. If you’d like to get creative and customize the naming even more—maybe add the room name or timestamps—just give me a shout! I’m happy to help you set that up.

and is this enough.? do i need to pass rec_req to somewhere? rec_req = api.RoomCompositeEgressRequest( room_name=agents.get_job_context().room.name, layout="speaker", preset=api.EncodingOptionsPreset.H264_720P_30, audio_only=True, segment_outputs=[ api.SegmentedFileOutput( filename_prefix=recording_file_name_prefix, segment_duration=5, s3=s3_upload_config, ) ], )

Once you’ve created the

RoomCompositeEgressRequest

, the next step is to send it over to the LiveKit egress service to kick off the recording. Just having the request object sets things up, but it won’t start recording by itself. Here’s a simple way to get the egress going:

lkapi = api.LiveKitAPI()
res = await lkapi.egress.start_room_composite_egress(rec_req)

await lkapi.aclose()

This will begin recording based on what you’ve configured. You’ll get back an

egress_id

in the response, which is handy for keeping track of or managing your recording session. If you have any questions or want a hand with this, just let me know—I’m here to help! Sources: Session recording and transcripts | LiveKit Docs | Universal Egress | protobufs/livekit_egress.proto

i want to save the recording url in database. how to do that?

Here’s a simple way to save the recording URL in your database right after you start the recording: 1. Start the Recording: Just call the

start_room_composite_egress

method to kick things off. This will give you an

EgressInfo

object. 2. Grab the Recording URL: From that

EgressInfo

, pull out the

egress_id

and build the recording URL using your output settings. For example:

recording_url = f"https://{req.segment_outputs[0].s3.bucket}.s3.{req.segment_outputs[0].s3.region}.<http://amazonaws.com/{req.segment_outputs[0].filename_prefix}|amazonaws.com/{req.segment_outputs[0].filename_prefix}>"

3. Save It in Your Database: Pop the

egress_id

and

recording_url

into your database so you can easily find them later. Like this:

save_recording_info(egress_id, recording_url)

Just a heads-up: make sure to catch any errors along the way and double-check the URL before saving. If you want, I’m here to help with the database part! Sources: Session recording and transcripts | LiveKit Docs | Egress API | LiveKit Docs | protobufs/livekit_egress.proto | cmd/lk/egress.go

if i want to save the recording in a file, this needs to be a call recording so ofc the format will not be mp4. so give me code for that.

If you want to save your call recordings as audio files in the OGG format, here’s a simple way to set up the `RoomCompositeEgressRequest`:

req = api.RoomCompositeEgressRequest(
    room_name="my-room",
    audio_only=True,  # For audio-only recordings
    file_outputs=[api.EncodedFileOutput(
        file_type=api.EncodedFileType.OGG,  # Specify OGG format
        filepath="recordings/{room_name}-{time}.ogg",
        s3=api.S3Upload(
            bucket="your-bucket-name",
            region="your-region",
            access_key="your-access-key",
            secret="your-secret-key",
        ),
    )],
)

Just swap out the placeholders with your own details, and you’ll have your recording saved as an OGG file in your S3 bucket. If you have any questions or want me to walk you through it, just let me know—I’m here to help!

i want to save recording to a file

If you'd like to save the recording to a local file, you can set up the

RoomCompositeEgressRequest

with a local file path. Here's a quick example to help you out:

req = api.RoomCompositeEgressRequest(
    room_name="my-room",
    audio_only=True,  # For audio-only recordings
    file_outputs=[api.EncodedFileOutput(
        file_type=api.EncodedFileType.OGG,  # Specify the file format
        filepath="/path/to/recording.ogg",  # Local file path
    )],
)

Just swap out

/path/to/recording.ogg

with wherever you'd like the file to be saved on your computer. Make sure that folder exists and that your app has permission to write there. If you run into any issues or have questions, feel free to ask—I'm here to help!

i am getting this error.

2025-08-06 16:50:06.644 | ERROR    | agents.assistant_agent:on_enter:114 - Failed to start call recording egress: TwirpError(code=invalid_argument, message=request has missing or invalid field: output, status=400)

here is the code m using

recording_file_name = (
                    f"call-recording-{self.assistant.get('id')}-{datetime.now().strftime('%Y-%m-%d-%H-%M-%S')}.ogg"
                )
                rec_req = api.RoomCompositeEgressRequest(
                    room_name=agents.get_job_context().room.name,
                    file_outputs=[
                        api.EncodedFileOutput(
                            file_type=api.EncodedFileType.OGG,
                            filepath=recording_file_name,
                        )
                    ],
                    audio_only=True,
                    # segment_outputs=[
                    #     api.SegmentedFileOutput(
                    #         filename_prefix=recording_file_name_prefix,
                    #         segment_duration=5,
                    #         s3=s3_upload_config,
                    #     )
                    # ],
                )
                # Use a short-lived LiveKitAPI client so the underlying aiohttp.ClientSession
                # is properly closed after the request finishes. Otherwise we leak the session
                # and `asyncio` will complain with an "Unclosed client session" warning which
                # we observed in the logs when starting call recording. Using an async context
                # manager automatically invokes `aclose()` on exit and prevents the leak.
                try:
                    async with api.LiveKitAPI() as lkapi:
                        info = await lkapi.egress.start_room_composite_egress(rec_req)
                        <http://logger.info|logger.info>(f"Started egress (id={info.egress_id}) for call recording")
                except Exception as e:
                    logger.error(f"Failed to start call recording egress: {e}")

Hey there! The error

request has missing or invalid field: output

pops up because the egress request is missing a properly set output destination. No worries though—here’s how you can fix it: 1. Make Sure Output is Specified: Every

RoomCompositeEgressRequest

needs at least one valid output. Looks like you’re using

file_outputs

, which is exactly right. 2. Double-Check the Output Setup: For

EncodedFileOutput

, just confirm these: ◦

file_type

is set properly (like

api.EncodedFileType.OGG

). ◦

filepath

points to a valid location (for example,

recordings/my-room.ogg

). 1. Use the Right Field Format: You’re already on track using

file_outputs

, which is perfect. Just avoid older fields like

file

. Here’s a tweaked version of your code to guide you:

recording_file_name = (
    f"call-recording-{self.assistant.get('id')}-{datetime.now().strftime('%Y-%m-%d-%H-%M-%S')}.ogg"
)
rec_req = api.RoomCompositeEgressRequest(
    room_name=agents.get_job_context().room.name,
    file_outputs=[
        api.EncodedFileOutput(
            file_type=api.EncodedFileType.OGG,
            filepath=f"recordings/{recording_file_name}",
        )
    ],
    audio_only=True,
)

Just a quick heads-up: make sure the

recordings/

folder exists and you have permission to write to it. If you hit any more snags or want me to walk you through anything else, just give me a shout! Sources: packages/livekit-server-sdk/src/EgressClient.ts | livekit-protocol/src/livekit.rs | livekit-protocol/livekit/protocol/egress.pyi | src/main/kotlin/io/livekit/server/EgressServiceClient.kt | pkg/errors/errors.go

i still getting this error

2025-08-06 16:59:43.926 | INFO     | __mp_main__:entrypoint:155 - Agent entrypoint finished.
2025-08-06 16:59:43,926 - DEBUG livekit.agents - using audio io: `RoomIO` -> `AgentSession` -> `TranscriptSynchronizer` -> `RoomIO` {"pid": 28825, "job_id": "AJ_MUCk5DEFBgUn"}
2025-08-06 16:59:43,926 - DEBUG livekit.agents - using transcript io: `AgentSession` -> `TranscriptSynchronizer` -> `RoomIO` {"pid": 28825, "job_id": "AJ_MUCk5DEFBgUn"}
2025-08-06 16:59:44.447 | ERROR    | agents.assistant_agent:on_enter:118 - Failed to start call recording egress: TwirpError(code=invalid_argument, message=request has missing or invalid field: output, status=400)

It looks like the error is happening because the

EncodedFileOutput

is missing a storage destination. Besides specifying the

file_type

and

filepath

, you’ll need to add at least one storage destination like S3, GCP, Azure, or AliOSS. Here’s a quick example of how to save the file locally:

rec_req = api.RoomCompositeEgressRequest(
    room_name=agents.get_job_context().room.name,
    file_outputs=[
        api.EncodedFileOutput(
            file_type=api.EncodedFileType.OGG,
            filepath="/path/to/recordings/call-recording.ogg",
        )
    ],
    audio_only=True,
)

If you’re using cloud storage (like S3), you’ll want to include the storage details like this:

rec_req = api.RoomCompositeEgressRequest(
    room_name=agents.get_job_context().room.name,
    file_outputs=[
        api.EncodedFileOutput(
            file_type=api.EncodedFileType.OGG,
            filepath="recordings/call-recording.ogg",
            s3=api.S3Upload(
                bucket="your-bucket-name",
                region="your-region",
                access_key="your-access-key",
                secret="your-secret-key",
            ),
        )
    ],
    audio_only=True,
)

Just double-check that the folder or storage bucket exists and you have permission to write to it. If you’re still running into trouble or want a hand with anything else, feel free to reach out—I’m here to help! Sources: Egress API | LiveKit Docs | livekit/livekit_egress.pb.go

not working

s3_upload_config = api.S3Upload(
    access_key=settings.SUPABASE_S3_ACCESS_KEY.get_secret_value(),
    secret=settings.SUPABASE_S3_SECRET_KEY.get_secret_value(),
    region=settings.SUPABASE_S3_REGION.get_secret_value(),
    endpoint=settings.SUPABASE_S3_STORAGE_URL.get_secret_value(),
    bucket=settings.SUPABASE_S3_BUCKET_NAME.get_secret_value(),
)

recording_file_name = (
                    f"call-recording-{self.assistant.get('id')}-{datetime.now().strftime('%Y-%m-%d-%H-%M-%S')}.ogg"
                )
                filepath = os.path.join(os.getcwd(), recording_file_name)
                <http://logger.info|logger.info>(f"Recording file path: {filepath}")
                rec_req = api.RoomCompositeEgressRequest(
                    room_name=agents.get_job_context().room.name,
                    layout="speaker",
                    file_outputs=[
                        api.EncodedFileOutput(
                            file_type=api.EncodedFileType.OGG,
                            filepath=filepath,
                            s3=s3_upload_config,
                        )
                    ],
                    audio_only=True,
                )
                try:
                    async with api.LiveKitAPI() as lkapi:
                        info = await lkapi.egress.start_room_composite_egress(rec_req)
                        <http://logger.info|logger.info>(f"Started egress (id={info.egress_id}) for call recording")
                except Exception as e:
                    logger.error(f"Failed to start call recording egress: {e}")

Hey! The issue happens because

EncodedFileOutput

can’t handle both local file paths and S3 settings at the same time. Just so you know, the

filepath

field is actually a filename template for where the file goes in your cloud storage (like S3), not a local file path. Here’s a simple way to fix your configuration:

rec_req = api.RoomCompositeEgressRequest(
    room_name=agents.get_job_context().room.name,
    file_outputs=[
        api.EncodedFileOutput(
            file_type=api.EncodedFileType.OGG,
            filepath="recordings/call-recording.ogg",  # Path within S3 bucket
            s3=api.S3Upload(
                bucket=settings.SUPABASE_S3_BUCKET_NAME.get_secret_value(),
                region=settings.SUPABASE_S3_REGION.get_secret_value(),
                access_key=settings.SUPABASE_S3_ACCESS_KEY.get_secret_value(),
                secret=settings.SUPABASE_S3_SECRET_KEY.get_secret_value(),
                endpoint=settings.SUPABASE_S3_STORAGE_URL.get_secret_value(),
            ),
        )
    ],
    audio_only=True,
)

With this setup, your recording will be saved right to your S3 bucket at the path you specified. If you want to get the file locally afterward, you can just download it from S3 once the recording’s done. Just shout if you want a hand with that! Sources: Egress API | LiveKit Docs | livekit/livekit_egress.pb.go | Session recording and transcripts | LiveKit Docs