Natural Language to AQL Quick Start

Ask a question in plain language and get an executable AQL query back, using AQLizer

Beta

Prerequisites

An ArangoDB database (note its name).
An API key for OpenAI or an OpenAI-compatible LLM provider.
For the API: a bearer token from POST https://<EXTERNAL_ENDPOINT>:8529/_open/auth.

Generate your first query

Open AQLizer

In the Contextual Data Platform, select your database and open the Query Editor. Click the AQLizer button at the bottom of the query tab.

Start the service

Provide your OpenAI API key, select a model (for example, gpt-5.4), and click Start Service.

Ask a question

Type a natural language question - for example, “Find all users who made purchases in the last month” - and click Ask (or press Return).

Run the generated query

Click Open in Editor to load the generated AQL into a new tab, then click Run query to execute it against your database.

Deploy the service

POST https://<EXTERNAL_ENDPOINT>:8529/_platform/acp/v1/graphrag

{
  "env": {
    "db_name": "<your_database_name>",
    "chat_api_provider": "openai",
    "chat_api_key": "<your-openai-api-key>",
    "chat_model": "gpt-5.4"
  }
}

Save the serviceIdPostfix (the trailing segment of serviceId) from the response.

Generate an AQL query

POST https://<EXTERNAL_ENDPOINT>:8529/graph-rag/:serviceIdPostfix/v1/process_text_stream

{
  "input_text": "Find all users who made purchases in the last month",
  "mode": "aqlizer",
  "response_instruction": "Return concise, executable AQL."
}

With "mode": "aqlizer", the endpoint streams back schema-aware AQL generated from your question. Copy the generated query into the Query Editor to run it against your database.

You now have a service that translates plain language into executable AQL and runs it against your database.

Next steps

Setup: Detailed deployment and configuration.
Web Interface: The full guided experience.
API Reference: All endpoints and options.