LAS provides a wide range of intelligent data processing operators, including operators for video understanding, text generation, and image generation based on the VolcanoEngineModelArk model, as well as multimodal data processing operators for audio, video, and documents that are enhanced by LAS's model capabilities. These operators enable more complex scenarios, higher quality results, and more convenient data processing.

LAS offers both "online" and "offline" operators, covering a variety of multimodal data processing scenarios such as audio, video, and documents.

LAS provides the online experience feature and offers a certain amount of free quota. No configuration is required; you can quickly process sample documents, audio, video, and other files through the online experience portal to try out the multimodal data processing capabilities provided by LAS.

The free quota limits for various operators supported by LAS in the online experience are as follows. Please monitor your usage during the trial to avoid incurring charges for exceeding the quota.

In addition to using the online trial feature, you can quickly initiate a data processing request through a simple API call to learn about the workflow for using online operators. You can refer to the process below, complete the preparation steps, and practice step by step to understand the workflows and key points for different scenarios.

Preparation #

• When using LAS operators for data processing, you need to generate an API Key for operator authentication. Because API Key information is sensitive, leaking your API Key may result in your model usage being consumed by others, causing potential losses. Therefore, after obtaining the API Key, it is recommended to configure it as an environment variable to ensure safe and proper use of the API Key. For details, see: Obtain and configure the API Key.
• When calling online operators, you first need to determine the BaseURL for operator calls based on the region where your current LAS service is deployed, which is used to configure the operator call path parameters. For details, see: Obtain the Base URL.

Hands-on practice #

• You can quickly initiate a data processing request through a simple API call to learn how to call the interface and set key parameters for LAS data processing operators. For details, see: Quick start.
• At the same time, LAS also provides practical examples of data processing in image and video scenarios, demonstrating common architectural approaches, debugging code, batch task configuration, and other end-to-end operational practices in multimodal data processing scenarios. For more details, see:
  • Hands-on practice: Video data ingestion and processing
  • Hands-on practice: Image data ingestion and processing

LAS skills & LAS CLI #

• Based on LAS's extensive data processing operators, LAS also offers more convenient multimodal data processing skills. You can easily install and use LAS skills in scenarios such as OpenClaw and IDE tools to process multimodal data such as audio, video, and documents. For the currently supported official LAS skills, see: LAS official skills list.
• LAS CLI is the official command-line tool for the AI Data Lake Service (LAS), providing commands for Key features such as authentication for LAS calls, multimodal data processing, and intelligent task orchestration. Combined with AI agent applications, you can easily integrate and invoke LAS's multimodal data processing capabilities with simple task descriptions while consuming fewer tokens. For details on installing and configuring LAS CLI and the list of operators currently supported by LAS CLI, see: LAS CLI command reference.

Datasets & Lance #

• LAS provides dataset features for data lakes designed for multimodal data, supporting more efficient and cost-effective data version management, data sharing, and more. For more details, see Dataset management.
• Lance is a modern columnar data format designed specifically for large-scale AI datasets. It is similar to Parquet, but has been deeply optimized for the storage and access of multimodal data such as images, videos, audio, point clouds, and traditional tabular data. Lance supports POSIX-compliant Linux file systems as well as mainstream cloud object storage (Volcano Engine TOS), making it an ideal foundation for building a unified AI data lake. LAS datasets support creating datasets in the Lance format and allow you to quickly convert other types of datasets to Lance datasets for easy data ingestion into the lake. For more details, see: Practical application: Instructions for using veDaft-based datasets.

Development machine & VeDaft #

The LAS development machine is a professional development environment tailored for algorithm developers, featuring high efficiency and convenience. Developers can use it to quickly start writing, debugging, and running data processing tasks.

• In terms of resource support, the development machine provides flexible computing resource configurations, covering both CPU and GPU, to meet the computing power requirements of tasks with varying complexity. At the same time, it supports mounting storage systems such as TOS and vePFS, providing stable support for data storage and reading.
• For the development environment, the development machine comes pre-integrated with a wide range of open-source images, including Daft, Ray, and Spark, significantly reducing environment setup time. In addition, it is equipped with persistent cloud disk storage services dedicated to storing data during development, ensuring data security and traceability.

This integrated design enables a truly "out-of-the-box" development model, effectively improving development efficiency and allowing algorithm developers to focus more on core data processing and algorithm optimization. For more information, see: Development machine overview.

Chat API & Responses API #

LAS also provides general-purpose data plane OpenAPI, including the Chat API and Responses API. Model capabilities for data processing can also be invoked through the general-purpose OpenAPI. For details, see: Chat API, Create model response.