Stream Files to Indexes

Starting a Stream

To start a stream through the SDK, use the steps below :

Create a Stream Object

ms = Stream("your-index-name")

Discover Available Sources

sources = ms.sources()  # returns a list of device names
print(sources)

Example console output would look like the following:

2 Sources Detected
==================================================
Device Camera or Screen
--------------------------------------------------
0 FaceTime HD Camera
1 Color LCD (1728 x 1117 @ 120 Hz)
==================================================

Select a Source

ms.set(1)  # choose the “Color LCD” screen capture from the list above

If no source is set explicitly, index 0 is used by default.

Start Streaming

ms.start()  # stream indefinitely (Ctrl-C to stop)

# OR limit the session to N seconds
ms.start(30)  # stream for 30 seconds

Once started, the stream will begin capturing video from the selected source functions can be run on the stream from its respective index. For example:

resp = moonshine.search(
    index="your-index-name",
    query="some prompt here"
)

Best Practices

Some best practices for streaming:

Verify devices first with ms.sources()— ID ordering can change between reboots.
Combine live Stream sessions with batch upload() to cover both real-time and historical footage.
Keep streams short when testing (ms.start(10)) to conserve local and network resources.
Use search or inquire during a live session to build interactive dashboards or alerts.

Stream() Constructor

string

required

Index that receives the live feed.Must have been created with performance=True

Stream() methods

.sources()

Returns List[str]—the display-ready names of cameras or screens detected on your machine.

.set()

int

required

The integer ID shown by .sources()

.start()

int

Optional. Duration to stream in seconds.
Omit or pass Noneto run until manually stopped.

Get Started

Ingestion

Running Functions

Workflows

Starting a Stream

Best Practices

​Starting a Stream

​Best Practices

Starting a Stream

Best Practices