How it works
AiQL processes documents through four steps. Run them individually as tools, or run the whole sequence as a single managed ingestion.1
Convert
Normalize a source document into clean, machine-readable text.
2
Split
Break the converted text into smaller chunks.
3
Shape
Infer the schema of the data from the chunks.
4
Extract
Pull out the specific fields and values you need.
POST /queries using natural language.
Core concepts
Workspaces
Containers that hold your ingested data. Every ingestion and query operates on a single workspace.
Ingestions
Managed jobs that run the full
convert → split → shape → extract pipeline and load the result into a workspace.Queries
Natural-language requests that run over the data you have already ingested.
Tools
The individual pipeline steps, available on demand when you need a single operation.
Get started
Quickstart
Go from an API key to your first query in a few minutes.
Authentication
Authenticate requests with a Bearer token and workspace header.
SDKs
Use the typed Python and TypeScript clients instead of raw HTTP.
Pricing
Understand how credits are billed across tools, ingestions, and queries.