LAS CLI is the official command-line tool for the AI Data Lake Service (LAS), providing commands for LAS invocation authentication, multimodal data processing, intelligent task orchestration, and other Key features. By integrating with AI agents, you can conveniently leverage LAS's multimodal data processing capabilities through simple task descriptions while consuming fewer tokens. This document introduces the installation and key usage instructions for LAS CLI commands.

Environment requirements #

• Node.js >= 16.0.0 (if you choose the npm installation method)
• LAS service has been enabledVolcanoEngine, and you have obtained the LAS API Key orVolcanoEngine the account's AK/SK information for subsequent authentication configuration when using the LAS service.

Install LAS CLI #

You can choose any of the following installation methods based on your operating system and environment.

npm install -g @las-ai/cli@latest

curl -fsSL https://lf3-static.bytednsdoc.com/obj/eden-cn/0eh7lpqbog/foeh7bfbvo/install-cli.sh | sudo bash

• After installation, you can run las-cli --version to verify the installation result. If the version number is output successfully, the installation is complete.
• You can send the following installation commands directly to the Agent to have the Agent install LAS CLI for you, or execute the installation commands in the terminal. The following are installation examples.

Uninstall LAS CLI #

• npm installation method: npm uninstall -g @las-ai/cli
• Binary installation method: rm -f /usr/local/bin/las-cli (or the actual path prompted during installation)

Step 1: Initialize LAS CLI #

After installing LAS CLI for the first time, you need to run the init command to initialize LAS configuration, mainly to complete the authentication configuration for invoking LAS services.

1. Obtain authentication information for LAS services. Currently, authentication for invoking LAS services supports using the LAS API Key. You can refer to the following method to obtain LAS API Key information.
   • LAS API Key: After logging in to LAS Console , create a LAS API Key on the Resource management > API key management page and obtain the corresponding information. For details, see API key management.
2. Open the LAS initialization configuration file ~/.las-cli/config.json, configure the authentication parameters, and then run the init command to initialize LAS CLI.

   warning

   Authentication information is highly sensitive. If leaked, it may result in your model usage being consumed by others, resulting in losses. Therefore, please store and initialize configuration in a trusted environment.

   • You can send the obtained LAS API Key information to the Agent to have the Agent complete the initialization configuration for you, or directly modify the configuration file in the terminal and run the init command to perform the initialization configuration. The following are configuration examples for both scenarios.
   • To obtain the complete init command documentation, run las-cli init --help, or refer to the init command reference section below.

Step 2: Invoke LAS services for multimodal data processing #

1. Learn about the currently supported multimodal data processing operators available via LAS CLI, as well as the parameters for each operator.

2. Use LAS CLI to call operator services for multimodal data processing.
   The following demonstrates how to use the PDF Parse Doubao - Doubao AI PDF parsing operator to parse a sample PDF document, showing you how to perform multimodal data processing with LAS CLI.

Step 3: Query and manage LAS data processing tasks #

By using the las-cli task list command, you can view the list of executed data processing tasks. Additionally, the current invocation methods for LAS multimodal data processing operators support both sync (synchronous) and async (asynchronous) modes. For asynchronous tasks, you can also use this command to check task status.

Step 4: Intelligent task orchestration #

1. LAS CLI provides native intelligent task orchestration capabilities. You can submit tasks using natural language via las-cli task create "task description", and the CLI will invoke a large model to orchestrate operators for the task.
   
2. After submitting a task, use las-cli task list to view the task list and see the current status of the task:
   
3. Once the task is completed, use las-cli task status task-ID to view the execution result:
   

Invocation method description #

• General format: las-cli command

init #

Command format #

las-cli init

Parameter description #

When executing the init command, an interactive LAS initialization parameter configuration interface will be launched. You can enter the configuration results for the following parameters in the interactive interface.

region #

Command format #

las-cli region [region]

Parameter description #

• To reset the region for LAS services, you can set new region information.
• If region information is not specified, the current region setting for LAS services will be returned.

Examples #

• View the current region for LAS services.

  las-cli region
  

• Modify the current region for LAS services.

  las-cli region cn-shanghai
  

config #

Command format: list #

Use the list subcommand to view the current configuration results for parameters related to large models.

las-cli config list

Command format: model #

Use the model subcommand to reset the large model parameters for LAS services.

las-cli config model

Parameter description #

• When setting model parameters with the model subcommand, you need to configure the following parameters.

operators #

Command format: list #

Use the subcommand list to view the current list of supported operators, parameter descriptions for each operator, Examples, and other help content.

las-cli operators list

Command format: call #

Use the subcommand call to invoke a specified operator for multimodal data processing.

las-cli operators call <operatorId> [--param1 value1 --param2 value2 ...]

The details are as follows:

• operatorId: Replace with the operatorId of the LAS operator to be invoked.
• Set operator parameters using --param according to the parameter requirements for operator invocation.

Supported operators and operatorId #

Description of operator invocation methods #

Operators on the LAS platform support two invocation modes, and the CLI provides a unified invocation experience for both modes.

Parameter description: ASR Pro (las_asr_pro) #

Enhanced speech recognition operator, supporting audio preprocessing, noise reduction, and extended analysis features.
Invocation mode: async

Examples:

las-cli operators call las_asr_pro \
  --audio_url "https://example.com/audio.wav" \
  --audio_format wav \
  --enable_punc true \
  --enable_speaker_info true \
  --language zh

Parameter description: Audio Convert (las_audio_convert) #

Convert audio files between different formats.
Invocation mode: sync

Examples:

las-cli operators call las_audio_convert \
  --input_path "tos://bucket/input.mp3" \
  --output_path "tos://bucket/output.wav" \
  --output_format wav

Parameter description: Audio Extract & Split (las_audio_extract_and_split) #

Extract audio from video and split it into segments based on the specified duration.
Invocation mode: sync

Examples:

las-cli operators call las_audio_extract_and_split \
  --input_path "tos://bucket/video.mp4" \
  --output_path_template "tos://bucket/output/segment_%03d.wav" \
  --split_duration 60 \
  --output_format wav

Parameter description: Image Resample (las_image_resample) #

Resample and scale images.
Invocation mode: sync

Examples:

las-cli operators call las_image_resample \
  --image "https://example.com/photo.jpg" \
  --tos_dir "tos://bucket/output/" \
  --target_size "[800, 600]" \
  --method lanczos

Parameter description: PDF Parse Doubao (las_pdf_parse_doubao) #

Use Doubao AI to parse PDF documents, supporting both standard and detailed parsing modes.
Invocation mode: async

Examples:

las-cli operators call las_pdf_parse_doubao \
  --url "https://example.com/document.pdf" \
  --parse_mode detail \
  --start_page 1 \
  --num_pages 10

Parameter description: Video Edit (las_video_edit) #

Edit and process videos using AI.
Invocation mode: async

Examples:

las-cli operators call las_video_edit \
  --video_url "https://example.com/video.mp4" \
  --output_tos_path "tos://bucket/output/edited" \
  --task_description "将视频剪辑为精彩片段" \
  --mode simple

Parameter description: Video Inpaint (las_video_inpaint) #

Intelligently remove objects such as watermarks and subtitles from videos.
Invocation mode: async

Examples:

las-cli operators call las_video_inpaint \
  --video_url "https://example.com/video.mp4" \
  --output_tos_path "tos://bucket/output/inpainted" \
  --targets '["watermark", "subtitle"]' \
  --detection_precise_mode true

Parameter description: Video Resize (las_video_resize) #

Adjust video resolution and encoding parameters.
Invocation mode: async

Examples:

las-cli operators call las_video_resize \
  --video_path "tos://bucket/input.mp4" \
  --output_tos_dir "tos://bucket/output/" \
  --output_file_name "resized.mp4" \
  --min_width 1280 --max_width 1920 \
  --min_height 720 --max_height 1080

Parameter description: VLM Video (las_vlm_video) #

Use a visual language model to understand and analyze video content.
Invocation mode: sync

Examples:

las-cli operators call las_vlm_video \
  --video_url "https://example.com/video.mp4" \
  --text "请描述这个视频的内容" \
  --temperature 0.7

task #

Command format: create #

Use the task create "<description>" command to create an intelligent orchestration task with a natural language description. The CLI will invoke a pre-configured large model, decompose your text description into a series of specific operator invocation steps, and execute them automatically.

# Example: Let AI help you analyze video and extract audio
las-cli task create "请分析这个视频 tos://bucket/video.mp4，告诉我里面有什么内容，并把音频提取出来"

Command format: list #

The task list command is used to view all locally recorded task history. This includes intelligent orchestration tasks generated by task create and asynchronous tasks produced by asynchronous operator invocation.

las-cli task list

Command format: status #

task status <task_id> command is used to query the current status and result of a specified task. The CLI first searches the local orchestration task history. If the task is not found, it queries the LAS server as a normal asynchronous task.

las-cli task status task-1678886400000-abcd

Command format: wait #

task wait <task_id> command blocks the current terminal until the specified asynchronous task is completed. This is very useful in scenarios where a script needs to wait for the result of a preceding task.

# Wait for an ASR task to complete and save its result to the result.json file
las-cli task wait <asr-task-id> --operator las_asr_pro --out result.json

Command format: kill #

task kill <task_id> command sends a request to the LAS server to terminate a running asynchronous task.

las-cli task kill <your-task-id>

Command format: delete #

task delete <task_id> is used to delete a specified task entry from the local task history, while task clear clears all local task history. This only affects local records and does not affect tasks already existing on the server.

# Delete a single local record
las-cli task delete task-1678886400000-abcd

# Clear all local records
las-cli task clear

help #

Command format #

las-cli help [command]

Examples #

# View help information for the config command
las-cli help config

# View help information for the operators call command
las-cli help operators call

Q: What should I do when using LAS CLI for the first time?
A: It is strongly recommended to run las-cli init directly. This interactive command guides you through all necessary configurations, including authentication credentials and default region, and is the fastest way to get started.
Q: I installed it on Windows. Why does running las-cli prompt that the command cannot be found?
A: If you installed via a PowerShell script, the script automatically updates the system PATH environment variable. However, you need to restart your terminal (such as PowerShell, CMD, or Windows Terminal) for the new environment variable to take effect.
Q: How do I set a global region to avoid specifying the --region parameter every time?
A: You can set a global default region using the las-cli region <region_name> command. For example, after running las-cli region cn-shanghai, all subsequent operator invocations will default to the Shanghai region unless you explicitly override it with --region.