# Dust

> Look up your Dust agents, search your data sources, and keep documents and tables in Dust folders up to date.

App · 18 actions. Web page: https://docs.lodol.com/docs/integrations/dust

## Actions

### Delete document

Delete a document from a folder. It cannot be undone.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Document ID (`document_id`, text, required): The document's id, from 'List documents' or 'Upsert document'.

**Outputs**

- `deleted` (true/false): Always true once deleted
- `document_id` (text): The document deleted

**Example output**

```json
{
  "deleted": true,
  "document_id": "weekly-2026-10-26"
}
```

### Delete row

Delete one row of a table.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Table (`table_id`, text, required): The table. Pick a data source first, then choose from its tables.
- Row ID (`row_id`, text, required): The row's id, from 'List rows' or 'Upsert rows'.

**Outputs**

- `deleted` (true/false): Always true once deleted
- `row_id` (text): The row deleted

**Example output**

```json
{
  "deleted": true,
  "row_id": "row01"
}
```

### Delete table

Delete a table and all its rows. It cannot be undone.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Table (`table_id`, text, required): The table. Pick a data source first, then choose from its tables.

**Outputs**

- `deleted` (true/false): Always true once deleted
- `table_id` (text): The table deleted

**Example output**

```json
{
  "deleted": true,
  "table_id": "scores"
}
```

### Get agent

Get one agent with its description, instructions, model and tools.

**Inputs**

- Agent (`agent_id`, text, required): The Dust agent. Pick one, or give its id from an earlier step.
- Detail (`variant`, text): Summary, or full with tools and data sources. One of: `light` (Summary (default)), `full` (Full, with tools and data sources).

**Outputs**

- `agent` (object): The agent

**Example output**

```json
{
  "agent": {
    "description": "Answers questions from the company wiki.",
    "id": "xYzAbC1234",
    "instructions": "Answer briefly and cite the page you used.",
    "is_favorite": false,
    "model": "claude-sonnet-4-5",
    "model_provider": "anthropic",
    "name": "Research assistant",
    "picture_url": "https://dust.tt/static/emojis/bg-sky-200/books/1f4da",
    "scope": "visible",
    "status": "active",
    "tools": [
      "search"
    ],
    "version": 3
  }
}
```

### Get document

Get one document with its text.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Document ID (`document_id`, text, required): The document's id, from 'List documents' or 'Upsert document'.

**Outputs**

- `document` (object): The document

**Example output**

```json
{
  "document": {
    "chunk_count": 1,
    "created": 1761000000000,
    "id": "weekly-2026-10-26",
    "mime_type": "application/vnd.dust.folder.document",
    "source_url": null,
    "tags": [
      "notes"
    ],
    "text": "Refunds are accepted for 30 days.",
    "text_size": 120,
    "timestamp": 1761000000000,
    "title": "Weekly notes",
    "token_count": 25
  }
}
```

### Get row

Get one row of a table by its id. A row just added may take about 15 seconds to be found.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Table (`table_id`, text, required): The table. Pick a data source first, then choose from its tables.
- Row ID (`row_id`, text, required): The row's id, from 'List rows' or 'Upsert rows'.

**Outputs**

- `row` (object): The row

**Example output**

```json
{
  "row": {
    "id": "row01",
    "values": {
      "name": "Ada",
      "score": 3
    }
  }
}
```

### Get table

Get one table with its columns.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Table (`table_id`, text, required): The table. Pick a data source first, then choose from its tables.

**Outputs**

- `table` (object): The table

**Example output**

```json
{
  "table": {
    "columns": [
      {
        "name": "name",
        "type": "text"
      },
      {
        "name": "score",
        "type": "int"
      }
    ],
    "description": "Quiz scores",
    "id": "scores",
    "name": "scores",
    "tags": [],
    "timestamp": 1761000000000,
    "title": "Scores"
  }
}
```

### List agents

List the Dust agents you can talk to, with their descriptions, models and instructions.

**Inputs**

- Which agents (`view`, text): Defaults to the agents you can use. One of: `list` (Agents I can use (default)), `all` (All shared agents), `published` (Published agents), `global` (Dust's built-in agents), `favorites` (My favorites).

**Outputs**

- `agents` (list): The agents
- `count` (number): How many

**Example output**

```json
{
  "agents": [
    {
      "description": "Answers questions from the company wiki.",
      "id": "xYzAbC1234",
      "instructions": "Answer briefly and cite the page you used.",
      "is_favorite": false,
      "model": "claude-sonnet-4-5",
      "model_provider": "anthropic",
      "name": "Research assistant",
      "picture_url": "https://dust.tt/static/emojis/bg-sky-200/books/1f4da",
      "scope": "visible",
      "status": "active",
      "tools": [
        "search"
      ],
      "version": 3
    }
  ],
  "count": 1
}
```

### List data sources

List a space's data sources: folders, websites and connected apps. Only folders can be written to.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.

**Outputs**

- `count` (number): How many
- `data_sources` (list): The data sources
- `folders` (list): Only the folders, the data sources that can be written to

**Example output**

```json
{
  "count": 1,
  "data_sources": [
    {
      "connector_provider": null,
      "created_at": 1761000000000,
      "description": "Notes written by Lodol",
      "id": "dts_abc123",
      "is_folder": true,
      "name": "Meeting notes"
    }
  ],
  "folders": [
    {
      "connector_provider": null,
      "created_at": 1761000000000,
      "description": "Notes written by Lodol",
      "id": "dts_abc123",
      "is_folder": true,
      "name": "Meeting notes"
    }
  ]
}
```

### List documents

List the documents in a data source, one page at a time, without their text.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Page size (`limit`, number): How many to return, 1 to 200. From 1 to 200.
- Skip (`offset`, number): How many to skip, to get a later page. At least 0.

**Outputs**

- `count` (number): How many on this page
- `documents` (list): The documents
- `total` (number): How many in all

**Example output**

```json
{
  "count": 1,
  "documents": [
    {
      "chunk_count": 1,
      "created": 1761000000000,
      "id": "weekly-2026-10-26",
      "mime_type": "application/vnd.dust.folder.document",
      "source_url": null,
      "tags": [
        "notes"
      ],
      "text_size": 120,
      "timestamp": 1761000000000,
      "title": "Weekly notes",
      "token_count": 25
    }
  ],
  "total": 1
}
```

### List rows

List a table's rows, one page at a time. Rows just added may take about 15 seconds to show.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Table (`table_id`, text, required): The table. Pick a data source first, then choose from its tables.
- Page size (`limit`, number): How many to return, 1 to 200. From 1 to 200.
- Skip (`offset`, number): How many to skip, to get a later page. At least 0.

**Outputs**

- `count` (number): How many on this page
- `rows` (list): The rows
- `total` (number): How many in all

**Example output**

```json
{
  "count": 1,
  "rows": [
    {
      "id": "row01",
      "values": {
        "name": "Ada",
        "score": 3
      }
    }
  ],
  "total": 1
}
```

### List spaces

List the spaces your key can see, with their ids and kinds.

**Inputs**: None.

**Outputs**

- `count` (number): How many
- `spaces` (list): The spaces

**Example output**

```json
{
  "count": 1,
  "spaces": [
    {
      "id": "vlt_abc123",
      "is_restricted": false,
      "kind": "global",
      "name": "Company Data"
    }
  ]
}
```

### List tables

List the tables in a data source, with their columns.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.

**Outputs**

- `count` (number): How many
- `tables` (list): The tables

**Example output**

```json
{
  "count": 1,
  "tables": [
    {
      "columns": [
        {
          "name": "name",
          "type": "text"
        },
        {
          "name": "score",
          "type": "int"
        }
      ],
      "description": "Quiz scores",
      "id": "scores",
      "name": "scores",
      "tags": [],
      "timestamp": 1761000000000,
      "title": "Scores"
    }
  ]
}
```

### Search agents

Find agents whose name contains some text, ignoring case.

**Inputs**

- Name contains (`query`, text, required): Part of the agent's name.

**Outputs**

- `agents` (list): The matching agents
- `count` (number): How many

**Example output**

```json
{
  "agents": [
    {
      "description": "Answers questions from the company wiki.",
      "id": "xYzAbC1234",
      "instructions": "Answer briefly and cite the page you used.",
      "is_favorite": false,
      "model": "claude-sonnet-4-5",
      "model_provider": "anthropic",
      "name": "Research assistant",
      "picture_url": "https://dust.tt/static/emojis/bg-sky-200/books/1f4da",
      "scope": "visible",
      "status": "active",
      "tools": [
        "search"
      ],
      "version": 3
    }
  ],
  "count": 1
}
```

### Search data source

Search a data source's documents by meaning and get the best matches with their text.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Search for (`query`, text, required): What to look for, in plain words.
- Results (`top_k`, number): How many matches to return, 1 to 100. Default 10. From 1 to 100.
- Return whole documents (`full_text`, true/false): Off: the matching passages. On: each document's full text. Default: `false`.
- Only documents tagged (`tags`, list): Documents with any of these tags.

**Outputs**

- `count` (number): How many
- `documents` (list): The best matches

**Example output**

```json
{
  "count": 1,
  "documents": [
    {
      "content": "Refunds are accepted for 30 days.",
      "id": "weekly-2026-10-26",
      "score": 0.82,
      "tags": [
        "notes"
      ],
      "timestamp": 1761000000000,
      "title": "Weekly notes"
    }
  ]
}
```

### Upsert document

Add a text document to a folder, or replace the one with the same id, so agents can search it.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Document ID (`document_id`, text, required): Your own id for the document. Using the same id again replaces it.
- Text (`text`, text, required): The document's text.
- Title (`title`, text): Shown in Dust and in search results.
- Tags (`tags`, list): Labels to filter searches by.
- Source link (`source_url`, text): A link back to where the text came from.
- Document date (`timestamp`, date and time): When the document was written. Defaults to now.

**Outputs**

- `document` (object): The document

**Example output**

```json
{
  "document": {
    "chunk_count": 1,
    "created": 1761000000000,
    "id": "weekly-2026-10-26",
    "mime_type": "application/vnd.dust.folder.document",
    "source_url": null,
    "tags": [
      "notes"
    ],
    "text_size": 120,
    "timestamp": 1761000000000,
    "title": "Weekly notes",
    "token_count": 25
  }
}
```

### Upsert rows

Add rows to a table. Each row is an object of column values; to replace a row, give its row_id and a value object. Dust makes new rows readable after about 15 seconds.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Table (`table_id`, text, required): The table. Pick a data source first, then choose from its tables.
- Rows (`rows`, list, required): A list of rows, each an object of column values, up to 500. Column names use lowercase letters, digits and underscores. Values are text, numbers or true/false.
- Replace all existing rows (`truncate`, true/false): On: delete every row already in the table first. Default: `false`.

**Outputs**

- `row_ids` (list): The id of each row, in order
- `rows_upserted` (number): How many rows were sent
- `table_id` (text): The table

**Example output**

```json
{
  "row_ids": [
    "row01",
    "row02"
  ],
  "rows_upserted": 2,
  "table_id": "scores"
}
```

### Upsert table

Create a table in a folder, or update the one with the same id. Its columns come from the rows you add to it.

**Inputs**

- Space (`space_id`, text, required): The space that holds the data source. Pick one, or give its id.
- Data source (`data_source_id`, text, required): The data source (a folder, for anything that writes). Pick a space first, then choose from its data sources.
- Table ID (`table_id`, text, required): Your own id for the table. Using the same id again updates it.
- Name (`name`, text): The table's name for queries. Defaults to the id.
- Title (`title`, text): Shown in Dust. Defaults to the name.
- Description (`description`, text): What the table holds, so agents know when to use it.
- Tags (`tags`, list): Labels for the table.

**Outputs**

- `table` (object): The table

**Example output**

```json
{
  "table": {
    "columns": [
      {
        "name": "name",
        "type": "text"
      },
      {
        "name": "score",
        "type": "int"
      }
    ],
    "description": "Quiz scores",
    "id": "scores",
    "name": "scores",
    "tags": [],
    "timestamp": 1761000000000,
    "title": "Scores"
  }
}
```
