Record and export sequences tutorial

viam machines part add-resource --part=$VIAM_PART_ID \
  --name=test-camera --model-name=fake --resource-subtype=camera
viam machines part add-resource --part=$VIAM_PART_ID \
  --name=test-sensor --model-name=fake --resource-subtype=sensor

Wait a few seconds for the machine to pick up the change, then confirm the sensor returns readings:

viam machines part run --part=$VIAM_PART_ID \
  --component=test-sensor --method=GetReadings

The CLI can’t add a registry module to a machine. viam machines part add-resource adds the sensor’s entry but not the viam:capture-control module entry, so viam-server can’t build the sensor. You can follow the Viam app instructions, or use the Viam MCP server instead, which adds the module for you. Ask your MCP client something like:

On my machine <machine-name>, add a sensor named my-capture-sensor with the model viam:capture-control:capture-control-sensor and these attributes:

{
  "resources": [
    { "resource_name": "test-camera", "method": "GetImages" },
    { "resource_name": "test-sensor", "method": "Readings" }
  ]
}

The client calls the add_machine_config_item tool. Its result includes a module_added entry for viam:capture-control.

Add the data management service, then set its attributes. The service’s API is rdk:service:data_manager, which --resource-subtype doesn’t accept, so pass --api:

viam machines part add-resource --part=$VIAM_PART_ID \
  --name=data-manager --api=rdk:service:data_manager --model-name=builtin
viam resource update --part=$VIAM_PART_ID --resource-name=data-manager \
  --config '{"sync_interval_mins": 0.1, "capture_control_sensor": {"name": "my-capture-sensor", "key": "overrides"}}'

If your machine already has a data management service under another name, remove it first. A machine can have only one.

for tag in demo-1 demo-2 demo-3; do
  viam machines part run --part=$VIAM_PART_ID \
    --component=my-capture-sensor --method=DoCommand \
    --data="{\"command\": {\"start_capture\": true, \"frequency_hz\": 2, \"tags\": [\"$tag\"]}}"
  sleep 10
  viam machines part run --part=$VIAM_PART_ID \
    --component=my-capture-sensor --method=DoCommand \
    --data='{"command": {"stop_capture": true}}'
  sleep 2
done

The CLI can’t list sequences yet, so this step uses the Python SDK. Save this as list_sequences.py. It prints each matching sequence’s details, then its ID on standard output:

import asyncio
import os
import sys

from viam.app.viam_client import ViamClient
from viam.rpc.dial import DialOptions

TAGS = {"demo-1", "demo-2", "demo-3"}


async def main():
    client = await ViamClient.create_from_dial_options(
        DialOptions.with_api_key(
            os.environ["VIAM_API_KEY"], os.environ["VIAM_API_KEY_ID"]
        )
    )
    data = client.data_client

    page_token = None
    while True:
        sequences, next_page_token = await data.list_sequences(
            organization_id=os.environ["VIAM_ORG_ID"], page_token=page_token
        )
        for s in sequences:
            if s.part_id != os.environ["VIAM_PART_ID"]:
                continue
            if not TAGS.intersection(s.sequence_tags):
                continue
            resources = ", ".join(
                f"{r.resource_name} · {r.method_name}" for r in s.resources
            )
            print(
                ",".join(s.sequence_tags),
                s.start_time.ToDatetime().isoformat(),
                s.end_time.ToDatetime().isoformat(),
                resources,
                file=sys.stderr,
            )
            print(s.id)
        if not next_page_token:
            break
        page_token = next_page_token

    client.close()


asyncio.run(main())

Run it and keep the IDs for the next step:

SEQUENCE_IDS=$(python list_sequences.py)
echo $SEQUENCE_IDS

You should see three sequences, one for each tag, each about 15 seconds long, with test-camera · GetImages and test-sensor · Readings as resources. The script matches on tags, so if you run the tutorial again, it also lists the sequences from earlier runs.

viam dataset create can’t set a dataset’s type yet, so this step uses the Python SDK. Save this as create_dataset.py. It creates a sequence dataset, adds the sequences you pass it, and prints the dataset’s ID. The dataset is named demos unless you set DATASET_NAME:

import asyncio
import os
import sys

from viam.app.viam_client import ViamClient
from viam.proto.app.dataset import DatasetType
from viam.rpc.dial import DialOptions


async def main():
    # Split the arguments too, because zsh passes $SEQUENCE_IDS as one argument.
    sequence_ids = " ".join(sys.argv[1:]).split()
    client = await ViamClient.create_from_dial_options(
        DialOptions.with_api_key(
            os.environ["VIAM_API_KEY"], os.environ["VIAM_API_KEY_ID"]
        )
    )
    data = client.data_client
    dataset_id = await data.create_dataset(
        name=os.environ.get("DATASET_NAME", "demos"),
        organization_id=os.environ["VIAM_ORG_ID"],
        type=DatasetType.DATASET_TYPE_SEQUENCE_DATA,
    )
    await data.add_sequences_to_dataset(
        dataset_id=dataset_id, sequence_ids=sequence_ids
    )
    print(dataset_id)
    client.close()


asyncio.run(main())

Run it with the IDs from step 5:

DATASET_ID=$(python create_dataset.py $SEQUENCE_IDS)
echo $DATASET_ID

Dataset names must be unique in your organization. If demos already exists, pick another name, for example export DATASET_NAME=demos-2, and run the script again.