> ## Documentation Index
> Fetch the complete documentation index at: https://docs.veadk.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# LAS data lake

## Overview

`las` (`from veadk.tools.builtin_tools.las import las`) is built on the AI multimodal data lake service LAS, providing creation, preview, query/analysis, editing, and cleaning of multimodal datasets.

## Environment & prerequisites

<Warning>
  Requirements:

  1. Configure the API key for the agent's reasoning model.
  2. [Create a LAS general dataset](https://console.volcengine.com/las/region:las+cn-beijing/next/dataset/common/create) in advance and obtain its DatasetId.
  3. Configure the LAS service DatasetId and URL ([how to get the URL](https://www.volcengine.com/mcp-marketplace/detail?name=LAS%20MCP)).
</Warning>

Environment variables:

* `MODEL_AGENT_API_KEY`: API key for the agent's reasoning model
* `TOOL_LAS_URL`: LAS MCP Server service address
* `TOOL_LAS_DATASET_ID`: dataset ID used by the LAS service

## Usage

```python title="examples/tools/las/agent.py" lines theme={null}
import asyncio
import os

from veadk import Agent, Runner
from veadk.memory.short_term_memory import ShortTermMemory
from veadk.tools.builtin_tools.las import las

agent = Agent(
    name="las_agent",
    model_name="doubao-seed-2-1-pro-260628",
    description="Answer using data in LAS.",
    instruction=(
        f"You are a poet. First search dataset {os.environ['TOOL_LAS_DATASET_ID']} for "
        "relevant content, then write a poem based on the results."
    ),
    tools=[las],
)

runner = Runner(agent=agent, short_term_memory=ShortTermMemory())


async def main():
    response = await runner.run("Write a Chinese-style poem about wood and love")
    print(response)


if __name__ == "__main__":
    asyncio.run(main())
```

## Dataset and connection checks

Set the service URL before importing `las`. `TOOL_LAS_DATASET_ID` is not automatically applied as a filter to every tool call. Select the dataset explicitly in instructions or tool arguments, and enforce access permissions on the server. Ingest searchable content into the configured dataset before running the example.

This check only lists tools exposed by the service. If discovery succeeds but retrieval is empty, check the selected dataset, ingested content, and permissions. The endpoint comes from `TOOL_LAS_URL` and is not switched automatically by `CLOUD_PROVIDER`.

```python lines theme={null}
import asyncio

from veadk.tools.builtin_tools.las import las


async def main():
    try:
        tools = await las.get_tools()
        print([tool.name for tool in tools])
    finally:
        await las.close()


if __name__ == "__main__":
    asyncio.run(main())
```