NVIDIA DeepStream SDK

https://docs.nvidia.com/metropolis/deepstream/9.1/
Admin
NVIDIA DeepStream SDK is a complete streaming analytics toolkit for building AI-powered multi-sensor...
Tokens:2,714,999
Snippets:28,895
Trust Score:10
Update:2 weeks ago
Show doc for...
Tokens:
Raw
### start ()

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/classgstnvinferserver_1_1GstNvInferServerImpl.html

Starts the inference server by reading configuration and setting up processing context.

```APIDOC
## NvDsInferStatus start ()

### Description
Reads the configuration file and sets up processing context.

### Returns
- **NvDsInferStatus** - Status of the start operation.

```

--------------------------------

### Example 1: Video data transfer with nvdsucxserversink and nvdsucxclientsrc

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_plugin_gst-nvdsucx.html

Example 1: server and client pipelines for sending video data over RDMA using nvdsucxserversink and nvdsucxclientsrc. Start the server program before the client program.

```bash
gst-launch-1.0 uridecodebin uri="file:///sample_1080p.mp4" async-handling=1 name=src1 src1. ! \
queue ! nvvideoconvert ! 'video/x-raw(memory:NVMM),format=NV12,width=1920,height=1080' ! \
nvdsucxserversink addr=192.168.100.1 port=4000 buf-type=nvdsucx-buf-video
```

```bash
gst-launch-1.0 nvdsucxclientsrc addr=192.168.100.1 port=4000 nvbuf-memory-type=2 num-nvbuf=4 buf-type=nvdsucx-buf-video ! \
'video/x-raw(memory:NVMM),format=NV12,width=1920,height=1080,framerate=30/1' ! \
queue ! nvvideoconvert ! nvv4l2h264enc ! h264parse ! qtmux name=mux_0 ! \
filesink sync=1 async=0 qos=0 location=~/out_1080p.mp4
```

--------------------------------

### mg_start2

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/sources_2includes_2civetweb_8h.html

Initializes and starts a new server context with the given configuration.

```APIDOC
## mg_start2

### Description
Starts a CivetWeb server context with initialization data.

### Function Signature
```c
CIVETWEB_API struct mg_context *mg_start2(struct mg_init_data *init, struct mg_error_data *error);
```

### Parameters
- **init** (struct mg_init_data*) - Initialization data structure.
- **error** (struct mg_error_data*) - Structure to receive error information on failure.

### Return Value
Returns a pointer to the new context, or NULL on failure.
```

--------------------------------

### Get startup info with curl GET

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_RestServer.html

Gets startup info. The endpoint can be /api/v1/startup or /startup. The response indicates whether the service has started.

```bash
curl -XGET 'http://localhost:9000/api/v1/startup'
OR
curl -XGET 'http://localhost:9000/startup'

```

```json
{
     "startup-info" :
     {
             "ds-startup" : "YES"
     },
     "reason" : "GET_STARTUP_INFO_SUCCESS",
     "status" : "HTTP/1.1 200 OK"
}

```

--------------------------------

### Start Kafka server and create topic

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_3D_MultiModal_Lidar_Camera_BEVFusion.html

Starts a Kafka server on port 9092 and creates the 'quick-test' topic for messaging tests. Requires Kafka Quick Start setup.

```bash
$ docker run -p 9092:9092 apache/kafka:3.7.0 # start kafka server with port 9092
$ bin/kafka-topics.sh --create --topic quick-test --bootstrap-server localhost:9092 # creat topic 'quick-test'

```

--------------------------------

### Download and setup sample data

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/Pipetuner-guide.html

Unzip the sample file and run setup.sh to prepare sample data. Choose either 'deepstream' or 'metropolis' based on your usage; run only one of the two commands.

```bash
$ unzip pipe-tuner-sample.zip
$ cd pipe-tuner-sample/scripts

$ # DeepStream or Metropolis Microservices users should run only one of the following two commands based on their usage
$ bash setup.sh deepstream            # DeepStream users

$ bash setup.sh metropolis            # Metropolis Microservices users
```

--------------------------------

### Start and wait for a video player pipeline with pyservicemaker

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_service_maker_python_quick_start.html

Equivalent to the previous example but uses Pipeline.start() and Pipeline.wait() instead of invoking the flow with '()'. This approach explicitly starts the pipeline and waits for it to finish.

```python
from pyservicemaker import Pipeline, Flow

pipeline = Pipeline("playback")
video_file = "/opt/nvidia/deepstream/deepstream/samples/streams/sample_1080p_h264.mp4"
Flow(pipeline).capture([video_file]).render()
pipeline.start()
pipeline.wait()

```

--------------------------------

### Example: run mosquitto with default config

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_plugin_gst-nvmsgbroker.html

Example of running mosquitto with the default configuration file located at /etc/mosquitto/mosquitto.conf.

```bash
mosquitto -c /etc/mosquitto/mosquitto.conf

```

--------------------------------

### REDIS adapter configuration example

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_plugin_gst-nvmsgbroker.html

Example configuration file entries for a REDIS protocol adapter, placed in a custom configuration file such as cfg_redis.txt. Adjust values for your redis server setup.

```ini
[message-broker]
hostname=localhost
port=6379
password=password
payloadkey=metadata
consumergroup=mygroup
consumername=myname
streamsize=10000
```

--------------------------------

### Start DeepStream Container with X11 and GPU

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_on_WSL2.html

Installs X11 utilities, enables X server access, and starts the DeepStream container with GPU, display, and audio support. The bind-mount destinations are for an x86_64 WSL2 host.

```bash
sudo apt install x11-xserver-utils
xhost +
sudo docker run -it --privileged --rm --name=docker --net=host --gpus all -e DISPLAY=$DISPLAY -e CUDA_CACHE_DISABLE=0  -e NVIDIA_DRIVER_CAPABILITIES=all --device /dev/snd -v /tmp/.X11-unix/:/tmp/.X11-unix -v /usr/lib/wsl/lib/libnvcuvid.so.1:/usr/lib/x86_64-linux-gnu/libnvcuvid.so.1:ro -v /usr/lib/wsl/lib/libnvidia-encode.so.1:/usr/lib/x86_64-linux-gnu/libnvidia-encode.so.1:ro nvcr.io/nvidia/deepstream:9.1-triton-multiarch
```

--------------------------------

### Gst-nvmsgbroker configuration file example

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_plugin_gst-nvmsgbroker.html

Example configuration file in gkey format with [message-broker] group, proto-cfg, consumer-group-id, and partition-key settings.

```ini
[message-broker]
proto-cfg="message.timeout.ms=2000;retries=5"
consumer-group-id = groupid
partition-key = keyid
```

--------------------------------

### mg_start

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/sources_2includes_2civetweb_8h_source.html

Starts the web server with the given callbacks and configuration options. It initializes the server context and returns a handle for further operations.

```APIDOC
## mg_start

### Description
Starts the web server with the given callbacks and configuration options. It initializes the server context and returns a handle for further operations.

### Method
Function call

### Endpoint
`struct mg_context *mg_start(const struct mg_callbacks *callbacks, void *user_data, const char **configuration_options)`

### Parameters
- **callbacks** (struct mg_callbacks*) - Required - User-defined callbacks structure.
- **user_data** (void*) - Required - User data pointer passed to callbacks.
- **configuration_options** (const char**) - Required - NULL terminated list of option_name, option_value pairs.

### Return Value
- **struct mg_context*** - Web server context, or NULL on error.

### Example
```c
const char *options[] = {
    "document_root", "/var/www",
    "listening_ports", "80,443s",
    NULL
};
struct mg_context *ctx = mg_start(&my_func, NULL, options);
```
```

--------------------------------

### install (std::function< void(Pipeline &, int key)> keyboard_listener)

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/classdeepstream_1_1Pipeline.html

Install a callback to capture keyboard events.

```APIDOC
## Pipeline & install (std::function< void(Pipeline &, int key)> keyboard_listener)

### Description
Install a callback to capture keyboard events.

### Method
install

### Parameters
- **keyboard_listener** (std::function<void(Pipeline&, int)>) - Required - Callback function for keyboard events.

### Returns
Reference to the Pipeline object.

```

--------------------------------

### size_t getBufOffset(uint32_t batchIdx) const override

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/classnvdsinferserver_1_1BaseBatchBuffer.html

Gets the offset from start of the memory allocation to the buffer pointer.

```APIDOC
## size_t getBufOffset(uint32_t batchIdx) const override

### Description
Gets the offset from start of the memory allocation to the buffer pointer.

### Method
getBufOffset

### Parameters
- **batchIdx** (uint32_t) - The batch index.

### Returns
- **size_t** - The offset in bytes.

```

--------------------------------

### getBufOffset()

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/classnvdsinferserver_1_1BaseBatchBuffer.html

Gets the offset from the start of the memory allocation to the buffer pointer. Needed for sharing CUDA memory with Triton server.

```APIDOC
## getBufOffset()

### Description
Gets the offset from the start of the memory allocation to the buffer pointer. Needed for sharing CUDA memory with Triton server.

### Method
`size_t getBufOffset(uint32_t _batchIdx_) const`

### Parameters
- **_batchIdx_** (uint32_t) - Required - The batch index for which to get the offset.

### Returns
- **size_t** - The offset from the start of the memory allocation to the buffer pointer.

### Implementation Details
- Implements `nvdsinferserver::IBatchBuffer::getBufOffset`.
- Reimplemented in `nvdsinferserver::RefBatchBuffer`, `nvdsinferserver::BatchSurfaceBuffer`, and `nvdsinferserver::SurfaceBuffer`.
- Defined at line 105 of `infer_batch_buffer.h`.
```

--------------------------------

### Pipeline::install

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/9_81_2service-maker_2includes_2pipeline_8hpp_source.html

Installs a keyboard listener callback.

```APIDOC
## Pipeline::install

### Description
Installs a keyboard listener callback that will be invoked when a key is pressed while the pipeline is running.

### Method
`Pipeline& install(std::function<void(Pipeline&, int key)> keyboard_listener)`

### Parameters
- **keyboard_listener** (std::function<void(Pipeline&, int key)>) - Required - Callback function that receives the pipeline reference and the key code.

### Returns
- **Pipeline&** - Reference to the Pipeline object for chaining.

### Example
```cpp
pipeline.install([](Pipeline& p, int key) { /* handle key */ });
```
```

--------------------------------

### getBufOffset(uint32_t _batchIdx_) const

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/classnvdsinferserver_1_1SurfaceBuffer.html

Gets the offset from the start of the memory allocation to the buffer pointer for a given batch index. Needed for sharing CUDA memory with Triton server.

```APIDOC
## getBufOffset(uint32_t _batchIdx_) const

### Description
Gets the offset from the start of the memory allocation to the buffer pointer for the specified batch index. This is needed for sharing CUDA memory with Triton server.

### Method
`size_t getBufOffset(uint32_t _batchIdx_) const`

### Parameters
- `_batchIdx_` (uint32_t) - Required - The batch index for which to get the offset.

### Returns
- `size_t` - The offset from the start of the memory allocation to the buffer pointer.

### Notes
- This method is an inline override of `BaseBatchBuffer::getBufOffset`.
- It references `BaseBatchBuffer::getBatchSize()` and `NvBufSurface::isContiguous`.
```

--------------------------------

### setup

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/search/functions_13.html

Sets up the SensorFusionApp with configuration and window closed callback.

```APIDOC
## setup

### Description
Sets up the SensorFusionApp with a configuration path and a window closed callback.

### Method
setup

### Signature
`SensorFusionApp::setup(const std::string &configPath, std::function< void()> windowClosed)`

### Parameters
- **configPath** (const std::string&) - Required - Path to the configuration file.
- **windowClosed** (std::function<void()>) - Required - Callback function invoked when the window is closed.
```

--------------------------------

### size_t nvdsinferserver::BaseBatchBuffer::getBufOffset(uint32_t _batchIdx_) const

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/classnvdsinferserver_1_1DetectionOutput.html

Gets the offset from the start of the memory allocation to the buffer pointer. Needed for sharing CUDA memory with Triton server.

```APIDOC
## size_t nvdsinferserver::BaseBatchBuffer::getBufOffset(uint32_t _batchIdx_) const

### Description
Gets the offset from the start of the memory allocation to the buffer pointer. Needed for sharing CUDA memory with Triton server.

### Method
size_t nvdsinferserver::BaseBatchBuffer::getBufOffset(uint32_t _batchIdx_) const

### Parameters
- **uint32_t _batchIdx_** - The batch index for which to get the offset.

### Returns
size_t - The offset from the start of the memory allocation to the buffer pointer.

### Remarks
- Implements nvdsinferserver::IBatchBuffer
- Reimplemented in nvdsinferserver::RefBatchBuffer, nvdsinferserver::BatchSurfaceBuffer, and nvdsinferserver::SurfaceBuffer
- Definition at line 105 of file sources/libs/nvdsinferserver/infer_batch_buffer.h
```

--------------------------------

### parse_dsexample()

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/search/functions_10.html

Parses DS Example configuration from a GKeyFile.

```APIDOC
## parse_dsexample()

### Description
Parses DS Example configuration from a GKeyFile.

### Method
parse_dsexample

### Signature
- `parse_dsexample(NvDsDsExampleConfig *config, GKeyFile *key_file)`

### Parameters
- `config` (NvDsDsExampleConfig*) - Required - Output configuration structure.
- `key_file` (GKeyFile*) - Required - The GKeyFile to parse.

```

--------------------------------

### Create virtual environment and install dependencies

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_Inference_Builder.html

Set up a Python virtual environment and install required dependencies for Inference Builder.

```bash
cd inference_builder
git submodule update --init --recursive
python3 -m venv .venv
source .venv/bin/activate
pip3 install -r requirements.txt

```

--------------------------------

### Minimal L4T setup commands for Jetson

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_docker_containers.html

Commands to install Docker and NVIDIA container runtime on Jetson after flashing only the BSP from Jetpack. Run these in order after booting the flashed BSP.

```bash
Flash BSP from Jetpack and boot

Run "sudo apt update"

Run "sudo apt install docker.io"

Run "sudo apt install nvidia-container"

Run "sudo apt install nvidia-l4t-gstreamer"

Run "sudo service docker restart"
```

--------------------------------

### mg_start2()

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/globals_func_m.html

Starts the CivetWeb web server with additional options. Defined in the CivetWeb header file.

```APIDOC
## mg_start2()

### Description
Starts the CivetWeb web server with additional options. Defined in the CivetWeb header file.

### Source
- `9.1/sources/includes/civetweb.h`
- `sources/includes/civetweb.h`
```

--------------------------------

### CUDA GPG error output

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_Installation.html

Example error output during CUDA installation when the GPG public key is not available. Refer to the NVIDIA blog for updating the CUDA Linux GPG repository key.

```text
W: GPG error: https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64 InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY A4B469963BF863CC
E: The repository 'https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64 InRelease' is no longer signed.
```

--------------------------------

### Set up Triton Server and backends on Jetson

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_Quickstart.html

Set up Triton Server and backends on Jetson by running the setup script from the samples directory. The script downloads Triton Server Release 2.68.0 by default and installs backends into /opt/nvidia/deepstream/deepstream/lib/triton_backends. To use a different backend directory, update the infer_config with the shown model_repo block.

```bash
$ cd /opt/nvidia/deepstream/deepstream/samples/
$ sudo ./triton_backend_setup.sh
```

```text
model_repo {
 backend_dir: /opt/nvidia/tritonserver/backends/
}
```

--------------------------------

### start()

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/classds3d_1_1GuardDataRender.html

Starts the data render process with the given content and optional path. Returns an ErrCode indicating success or failure.

```APIDOC
## start()

### Description
Starts the data render process with the given content and optional path. Returns an ErrCode indicating success or failure.

### Method
start

### Signature
`ErrCode ds3d::GuardDataProcess<abiDataRender>::start(const std::string &_content_, const std::string &_path_ = "")`

### Parameters
- **content** (const std::string&) - Required - The content to be processed.
- **path** (const std::string&) - Optional - The path associated with the content. Defaults to empty string.

### Returns
- **ErrCode** - The result of the start operation.
```

--------------------------------

### PipeTuner console output during tuning

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/Pipetuner-guide.html

Example console output from a PipeTuner run. It shows dependency installation, BBO client/server launch, progress updates (0% to 100%), and a message to press CTRL+C to stop tuning. Note that TensorRT engine files are generated during the first iteration, which may take 10 to 30 minutes longer.

```text
Installing dependencies...
Installing dependencies (1/2)
Installing dependencies (2/2)
Launch BBO client...
Launch BBO server...
PipeTuner started successfully!

!!!!! To stop tuning process in the middle, press CTRL+C !!!!!

adding: DsAppServer (deflated 63%)
2024-03-19 08:09:52,108 root           INFO    seq_list: ['Retail_Synthetic_Cam01', 'Retail_Synthetic_Cam02', 'Retail_Synthetic_Cam03', 'Retail_Synthetic_Cam04', 'Retail_Synthetic_Cam05', 'Retail_Synthetic_Cam06', 'Retail_Synthetic_Cam07', 'Retail_Synthetic_Cam08']
2024-03-19 08:09:52,169 root           INFO    Writing configs to <path>/output/SDG_sample_PeopleNet-ResNet34_NvDCF-ResNet50_MOT_test.yml_output/results/configs_03-19-2024_08-09-52
2024-03-19 08:09:52,169 root           INFO    send backend init
2024-03-19 08:09:52,170 root           INFO    creating optimizers...
2024-03-19 08:09:52,170 root           INFO    done. created 2
 * Serving Flask app 'ds_bbo_frontend_server'
 * Debug mode: on
2024-03-19 08:09:52,186 root           INFO    init jobs done
2024-03-19 08:09:52,187 root           INFO    progress: 0% (0/4)
2024-03-19 08:09:52,193 root           INFO    wait backend ready
...
2024-03-19 08:12:15,232 root           INFO    progress: 25% (1/4) ETA 00:07:09
2024-03-19 08:12:25,447 root           INFO    progress: 25% (1/4) ETA 00:07:39
...
2024-03-19 08:13:47,190 root           INFO    progress: 50% (2/4) ETA 00:03:55
2024-03-19 08:13:57,413 root           INFO    progress: 75% (3/4) ETA 00:01:21
...
2024-03-19 08:14:58,719 root           INFO    progress: 100% (4/4) ETA 00:00:00
2024-03-19 08:14:58,720 root           INFO    OPTIMIZATION DONE!
…
stopping workers...
number of workers to stop: 2
number of workers to stop: 0
done.

!!!!! Press CTRL+C key to end PipeTuner !!!!!

```

--------------------------------

### Pipeline::install

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/9_81_2service-maker_2includes_2pipeline_8hpp_source.html

Installs a callback to capture keyboard events.

```APIDOC
## Pipeline::install

### Description
Install a callback to capture keyboard events.

### Method
install

### Signature
`Pipeline & install(std::function< void(Pipeline &, int key)> keyboard_listener)`

### Parameters
- **keyboard_listener** (std::function<void(Pipeline&, int)>) - Required - Callback function to handle keyboard events.

### Returns
- **Pipeline&** - Reference to the pipeline for chaining.

### Definition
9.1/service-maker/includes/pipeline.hpp:276
```

--------------------------------

### Install prerequisite packages

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_Installation.html

Installs the prerequisite packages required before installing the DeepStream SDK on Ubuntu.

```bash
$ sudo apt install \
libssl3 \
libssl-dev \
libcurl4-openssl-dev \
libgles2-mesa-dev \
libgstreamer1.0-0 \
gstreamer1.0-tools \
gstreamer1.0-plugins-good \
gstreamer1.0-plugins-bad \
gstreamer1.0-plugins-ugly \
gstreamer1.0-libav \
libgstreamer-plugins-base1.0-dev \
libgstrtspserver-1.0-0 \
libjansson4 \
libyaml-cpp-dev \
libjsoncpp-dev \
protobuf-compiler \
libmosquitto1 \
gcc \
make \
git \
python3
```

--------------------------------

### Basic Pipeline Setup with nvdsdynamicsrcbin

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_plugin_gst-nvdsdynamicsrcbin.html

Creates a GStreamer pipeline with nvdsdynamicsrcbin as the source element and a fakesink, adds and links them, then sets the pipeline to PLAYING state.

```c
// Create pipeline elements
GstElement *pipeline = gst_pipeline_new("dynamicsrcbin-test");
GstElement *src = gst_element_factory_make("nvdsdynamicsrcbin", "src");
GstElement *sink = gst_element_factory_make("fakesink", "sink");

// Add elements to pipeline and link them
gst_bin_add_many(GST_BIN(pipeline), src, sink, NULL);
gst_element_link_many(src, sink, NULL);

// Set pipeline to PLAYING state
gst_element_set_state(pipeline, GST_STATE_PLAYING);
```

--------------------------------

### Extract video frames with MediaExtractor and custom BufferProvider

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_service_maker_python_advanced_features.html

Use MediaExtractor to decode video frames from multiple MediaChunk sources and feed them into a custom BufferProvider for rendering. Only applicable on x86 systems. The example extracts 1-second chunks starting at 40 seconds from a sample video, using a ThreadPoolExecutor to render each chunk's queue with a renderer pipeline.

```python
from pyservicemaker import BufferProvider, Buffer, Flow, Pipeline, ColorFormat, RenderMode
from pyservicemaker.utils import MediaExtractor, MediaChunk
from concurrent.futures import ThreadPoolExecutor
import queue

VIDEO_FILE = "/opt/nvidia/deepstream/deepstream/samples/streams/sample_720p.mp4"
N_CHUNKS = 8
# Enable display if you have one on your system and want to see the frames
ENABLE_DISPLAY = False

class MyBufferProvider(BufferProvider):
    def __init__(self, queue, width, height):
        print("MyBufferProvider")
        super().__init__()
        self._queue = queue
        self.format = "RGB"
        self.width = width
        self.height = height
        self.framerate = 30
        self.device = 'gpu'
        self.frames = 0

    def generate(self, size):
        try:
            frame = self._queue.get(timeout=5)
            if frame is None:
                # Done with the frame queue
                return Buffer()
            self.frames += 1
            return frame.tensor.wrap(ColorFormat.RGB)
        except queue.Empty:
            print("Buffer empty")
            return Buffer()
# extract 1 second video frame starting from 40 seconds
chunks = [MediaChunk(f, start_pts=int(40*1e9), duration=int(1e9)) for f in [VIDEO_FILE]*N_CHUNKS]
with MediaExtractor(chunks=chunks, batch_size=N_CHUNKS) as extractor:
    qs = extractor()
    with ThreadPoolExecutor(max_workers=N_CHUNKS) as exe:
        render_mode = RenderMode.DISPLAY if ENABLE_DISPLAY else RenderMode.DISCARD
        exe.map(lambda q: Flow(Pipeline("renderer")).inject([MyBufferProvider(q, 1280, 720)]).render(render_mode, sync=False)(), qs)
```

--------------------------------

### MQTT adapter configuration file (cfg_mqtt.txt)

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_plugin_gst-nvmsgbroker.html

Example configuration file entries for the MQTT protocol adapter. Includes username, password, client-id, and commented TLS options. Uncomment share-connection to generate a connection signature.

```ini
[message-broker]
username = user
password = password
client-id = uniqueID
#enable-tls = 1
#tls-cafile =
#tls-capath =
#tls-certfile =
#tls-keyfile =
#share-connection = 1
#loop-timeout = 2000
#keep-alive = 60
```

--------------------------------

### Compile and install Gst-nvreplay

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_plugin_gst-nvreplay.html

Compile and install the Gst-nvreplay plugin from its source directory. Requires sudo for the install step.

```bash
cd deepstream/sources/gst-plugins/gst-nvreplay/
make && sudo -E make install

```

--------------------------------

### mg_start_domain()

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/9_81_2sources_2includes_2civetweb_8h.html

Starts a domain (virtual host) on an existing server context.

```APIDOC
## mg_start_domain()

### Description
Starts a domain (virtual host) on an existing server context.

### Method
Function call

### Signature
`CIVETWEB_API int mg_start_domain(struct mg_context *_ctx_, const char **_configuration_options_)`

### Parameters
- **`_ctx_`** (struct mg_context*) - Required - Pointer to the server context.
- **`_configuration_options_`** (const char**) - Required - Array of configuration options.

### Return Value
- **int** - 0 on success, non-zero on failure.
```

--------------------------------

### Start a DS3D dataloader manually

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_3D_Custom_Manual.html

Manually start a DS3D dataloader after it has been created. The dataloader can also be started automatically by gst-pipeline.

```cpp
GuardDataLoader dataloader = loaderSrc.customProcessor;
ErrCode c = dataloader.start();

```

--------------------------------

### DsExampleCtxInit

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/9_81_2sources_2gst-plugins_2gst-dsexample-cuda_2dsexample__lib_2dsexample__lib_8h.html

Initializes the DsExample context with the provided initialization parameters.

```APIDOC
## DsExampleCtxInit

### Description
Initializes the DsExample context with the provided initialization parameters.

### Syntax
```c
DsExampleCtx * DsExampleCtxInit (DsExampleInitParams *init_params)
```

### Parameters
- **init_params** (DsExampleInitParams*) - Required - Pointer to the initialization parameters.

### Returns
- **DsExampleCtx*** - Pointer to the initialized context, or NULL on failure.

```

--------------------------------

### Verify installed WSL distributions

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_on_WSL2.html

Verifies that Ubuntu is installed properly after a potential restart. If the distribution is not listed, re-run the install command.

```bash
$ wsl --list

```

--------------------------------

### mg_start

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/9_81_2sources_2includes_2civetweb_8h_source.html

Starts the CivetWeb web server with user-defined callbacks and configuration options.

```APIDOC
## mg_start

### Description
Starts the web server with the given callbacks and configuration options. Returns a context handle or NULL on error.

### Method
Function call

### Signature
```c
CIVETWEB_API struct mg_context *mg_start(const struct mg_callbacks *callbacks, void *user_data, const char **configuration_options);
```

### Parameters
- **callbacks** (const struct mg_callbacks*) - Required - User-defined callbacks structure.
- **user_data** (void*) - Required - User data pointer passed to callbacks.
- **configuration_options** (const char**) - Required - NULL terminated list of option_name, option_value pairs.

### Return Value
- **struct mg_context*** - Web server context, or NULL on error.

### Example
```c
const char *options[] = {
  "document_root", "/var/www",
  "listening_ports", "80,443s",
  NULL
};
struct mg_context *ctx = mg_start(&my_func, NULL, options);
```
```

--------------------------------

### Remove previous DeepStream installations

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_Installation.html

Removes all previous DeepStream 3.0 or prior installations. Run these commands before installing the DeepStream SDK on Ubuntu.

```bash
$ sudo rm -rf /usr/local/deepstream /usr/lib/x86_64-linux-gnu/gstreamer-1.0/libgstnv* /usr/bin/deepstream* /usr/lib/x86_64-linux-gnu/gstreamer-1.0/libnvdsgst*
/usr/lib/x86_64-linux-gnu/gstreamer-1.0/deepstream*
/opt/nvidia/deepstream/deepstream*
$ sudo rm -rf /usr/lib/x86_64-linux-gnu/libv41/plugins/libcuvidv4l2_plugin.so
```

--------------------------------

### DsExampleCtxInit

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/9_81_2sources_2gst-plugins_2gst-dsexample-cuda_2dsexample__lib_2dsexample__lib_8h.html

Initializes a DsExampleCtx context with the given initialization parameters and returns a pointer to the new context.

```APIDOC
## DsExampleCtxInit

### Description
Initializes a DsExampleCtx context with the given initialization parameters and returns a pointer to the new context.

### Function Signature
```c
DsExampleCtx* DsExampleCtxInit(DsExampleInitParams* _init_params_)
```

### Parameters
- `_init_params_` (DsExampleInitParams*) - Required - Pointer to initialization parameters.

### Returns
- `DsExampleCtx*` - Pointer to the newly created context, or NULL on failure.
```

--------------------------------

### Install librdkafka-dev

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_Installation.html

Installs the librdkafka development package, enabling the Kafka protocol adaptor for message broker integration. Run after installing confluent-platform.

```bash
sudo apt-get install librdkafka-dev
```

--------------------------------

### Run sample_app.py with a media URI

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_service_maker_python_intro_to_pipeline_api.html

Command to run the sample application with a media file URI.

```bash
$ python3 sample_app.py file:///opt/nvidia/deepstream/deepstream/samples/streams/sample_720p.mp4
```

--------------------------------

### Get readiness info with curl GET

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_RestServer.html

Gets readiness info. The endpoint can be /api/v1/ready or /ready. The response indicates whether the service is ready.

```bash
curl -XGET 'http://localhost:9000/api/v1/ready'
OR
curl -XGET 'http://localhost:9000/ready'

```

```json
{
     "ready-info" :
     {
             "ds-ready" : "YES"
     },
     "reason" : "GET_READY_INFO_SUCCESS",
     "status" : "HTTP/1.1 200 OK"
}

```

--------------------------------

### mg_start2

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/9_81_2sources_2includes_2civetweb_8h_source.html

Initializes the CivetWeb server context with the provided initialization data and error handling.

```APIDOC
## mg_start2

### Description
Initializes the CivetWeb server context with the provided initialization data and error handling.

### Function Signature
```c
CIVETWEB_API struct mg_context *mg_start2(struct mg_init_data *init, struct mg_error_data *error);
```

### Parameters
- **init** (struct mg_init_data*) - Required - Pointer to initialization data containing callbacks, user data, and configuration options.
- **error** (struct mg_error_data*) - Required - Pointer to error data structure to receive error details on failure.

### Return Value
- **struct mg_context*** - Pointer to the created server context, or NULL on failure.

### Error Codes
- **MG_ERROR_DATA_CODE_INVALID_PARAM** - Caller provided invalid parameter.
- **MG_ERROR_DATA_CODE_INVALID_OPTION** - "configuration_option" contains invalid element.
- **MG_ERROR_DATA_CODE_INIT_TLS_FAILED** - Initializing TLS/SSL library failed.
- **MG_ERROR_DATA_CODE_MISSING_OPTION** - Mandatory "configuration_option" missing.
- **MG_ERROR_DATA_CODE_DUPLICATE_DOMAIN** - Duplicate "authentication_domain" option.
- **MG_ERROR_DATA_CODE_OUT_OF_MEMORY** - Not enough memory.
- **MG_ERROR_DATA_CODE_SERVER_STOPPED** - Server already stopped.
- **MG_ERROR_DATA_CODE_INIT_LIBRARY_FAILED** - mg_init_library must be called first.
- **MG_ERROR_DATA_CODE_OS_ERROR** - Operating system function failed.
- **MG_ERROR_DATA_CODE_INIT_PORTS_FAILED** - Failed to bind to server ports.
- **MG_ERROR_DATA_CODE_INIT_USER_FAILED** - Failed to switch user (option "run_as_user").
- **MG_ERROR_DATA_CODE_INIT_ACL_FAILED** - Access Control List error.
- **MG_ERROR_DATA_CODE_INVALID_PASS_FILE** - Global password file error.
- **MG_ERROR_DATA_CODE_SCRIPT_ERROR** - Lua background script init error.
```

--------------------------------

### start()

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/sdk-api/classds3d_1_1gst_1_1PipelineContext.html

Starts the pipeline context by playing the pipeline and running the main loop. The provided callback is invoked when the loop quits.

```APIDOC
## start()

### Description
Starts the pipeline context by playing the pipeline and running the main loop. The provided callback is invoked when the loop quits.

### Method
start

### Signature
`virtual ErrCode ds3d::gst::PipelineContext::start(std::function<void()> _loopQuitCb_)`

### Parameters
- `_loopQuitCb_` (std::function<void()>) - Required - Callback invoked when the main loop quits.

### Return Value
- `ErrCode` - Returns `ds3d::kGood` on success, or an error code on failure.

### Remarks
- Calls `playPipeline()` and `runMainLoop()`.
- Defined in `sources/libs/ds3d/gst/nvds3d_pipeline_context.h`.
```

--------------------------------

### Get liveness info with curl GET

Source: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS_RestServer.html

Gets liveness info. The endpoint can be /api/v1/live or /live. The response indicates whether the service is live.

```bash
curl -XGET 'http://localhost:9000/api/v1/live'
OR
curl -XGET 'http://localhost:9000/live'

```

```json
{
     "live-info" :
     {
             "ds-liveness" : "YES"
     },
     "reason" : "GET_LIVE_INFO_SUCCESS",
     "status" : "HTTP/1.1 200 OK"
}

```
Tokens:
Raw