For everyone

Ask Trino#

Ask Trino turns a plain-English question into Trino SQL you can review and run. It's an assistant, not an autopilot: it only ever produces SQL text — it never connects to your clusters or data itself.

How it works#

  1. Open Ask Trino from the sidebar and set the context on the left rail — the cluster, catalog, and schema you want the query to target, plus an optional persona that tunes the assistant's style.
  2. Type what you want in plain language (or start from a suggested prompt).
  3. The assistant replies with a SQL statement and a short explanation.
  4. Review the SQL, then send it to the SQL editor to run it like any other query — with the same cluster, catalog, and result limits.

Safety#

  • The assistant only emits SQL as text. It has no connection to Trino or your data; nothing runs until you choose to run it.
  • Generated SQL is validated as read-only — a single SELECT statement — before it can run through the normal query path. Statements that try to write or run multiple commands are rejected.
  • Always read the SQL before running it. Treat the assistant as a fast first draft, not a guarantee of correctness.

Configuration (admins)#

Ask Trino calls an LLM provider from the control plane to draft SQL. It's off until you give it an API key. The key lives in the server environment — it is never logged and never returned to the browser — so it can't be set from the UI. If the feature is unconfigured, the assistant is unavailable and the rest of TrinoHub is unaffected.

Enable it#

  1. Get an API key. By default Ask Trino talks to OpenRouter — create an account and an API key there. (To use a different OpenAI-compatible provider, see the overrides below.)
  2. Give the key to the control plane. Set OPENROUTER_API_KEY (or the alias ASK_TRINO_API_KEY) in the server environment. The simplest way is a .env file in the project root — the app loads it automatically on start:

dotenv # .env (project root; never commit this file) OPENROUTER_API_KEY=sk-or-your-key-here

Alternatively, export it wherever the control plane runs (systemd unit, container env, etc.). See .env.example for a template. 3. Restart the control plane so it picks up the new environment. 4. Open Ask Trino from the sidebar — it's now live.

On an AWS (CloudFormation) deploy#

The stack in deploy/aws/cloudformation.yaml can wire the key in for you — pass one of these parameters (both optional; you can also add them later with a stack update):

  • OpenRouterApiKey — the key itself. Quickest, but the value is stored in the instance's EC2 user-data, so anyone with ec2:DescribeInstanceAttribute in the account can read it. Fine for trials; prefer the secret below for production.
  • OpenRouterSecretArn — the ARN of a Secrets Manager secret holding the key (a raw string, or JSON with an OPENROUTER_API_KEY field). The key never appears in user-data, and the stack grants the control-plane role read access to exactly that secret. Takes precedence if both are set.

Either way, boot writes the key to the control plane's .env and the assistant comes up enabled. To create the secret first:

aws secretsmanager create-secret --name trinohub/openrouter \
  --secret-string sk-or-your-key-here
# then pass its ARN as OpenRouterSecretArn

Model and provider (optional)#

  • Model — the default is openai/gpt-4o-mini. Change it in Settings → Ask Trino (paste any model id from openrouter.ai/models, e.g. anthropic/claude-sonnet-4.5), or set ASK_TRINO_MODEL in the environment.
  • Provider — point at any OpenAI-compatible chat-completions endpoint with ASK_TRINO_API_BASE (default: https://openrouter.ai/api/v1/chat/completions).

If a request fails with "Ask Trino is not configured", the key isn't set in the server environment; if the provider "rejected the API key," check the key value and, if you overrode it, the model id and API base.

Edit this page on GitHub