Pipeline tools - AiQL Docs

The pipeline tools mirror the API tools. An agent can call them individually or chain them as convert → split → shape → extract.

convert

Normalizes a source document into clean text and pages.

Parameter	Type	Description
`source`	`string`	URL or storage path of the document to convert.

Returns the normalized text along with the page count. See the Convert cookbook for details.

Breaks converted text into smaller chunks that the later steps can process reliably.

Parameter	Type	Description
`text`	`string`	Converted document text to split.
`max_words`	`integer`	Maximum number of words per chunk. Defaults to `500`.

Returns an ordered list of chunks. See the Split cookbook for details.

Inspects chunks and infers the schema of the data they contain.

Parameter	Type	Description
`chunks`	`string[]`	Ordered list of text chunks to inspect.

Returns the inferred fields and their types. See the Shape cookbook for details.

Pulls the requested fields out of the text and returns them as structured values.

Parameter	Type	Description
`chunks`	`string[]`	Ordered list of text chunks to extract from.
`fields`	`string[]`	Names of the fields to pull out of the text.

Returns a record per chunk with the requested fields populated. See the Extract cookbook for details.

Tools called over MCP are billed the same as the equivalent API tools. See Pricing for credit rates.