Sqliqs lets you ask questions about your database in plain English — no SQL required. This guide walks through how to get the best results, how to keep your data safe, and the limits worth knowing before you rely on it.

Getting started

You have two ways in:

Playground — try Sqliqs instantly against a hosted sample database. No login, no key, no setup. Great for seeing how it works.
Your own database — open the dashboard, create a project, and paste a connection string for PostgreSQL, MySQL, SQLite, or MongoDB. No sign-up required, and your connection details stay in your browser.

Once connected, open a project and start asking questions.

Pick a model and add your key

Sqliqs is bring-your-own-key (BYOK). In a project's settings you choose:

A provider (OpenRouter, Anthropic, OpenAI, Google, xAI, or DeepSeek), and
A model from that provider, paired with your own API key.

You can also start with the built-in free model — no key needed — but it's rate-limited and slower (see Limitations). For real work, add your own key.

Better model, better answers. More capable models write more accurate queries, handle complex schemas, and reason over your data far better. If answers feel shallow or a query looks wrong, switch to a stronger model.

Asking good questions

The clearer your question, the better the query Sqliqs writes:

Be specific. "What was total revenue by month in 2025?" beats "show me revenue."
Name things when you can. Referencing the table or column you mean helps the model target the right data.
Ask for what you want to see. Say "show me the rows / give me a table" to get a table, or "chart this" to get a visualization.
Follow up. Chat keeps context, so you can drill in: "now break that down by region."

The four surfaces

Chat — ask questions and get plain-English answers, with the query and result table shown alongside.
Visualization — describe what you want to see; Sqliqs picks the chart type and fills it with your data.
Reports — ask for a report in one sentence and get a formatted document with narrative, tables, and charts.
ER Diagram — an auto-generated map of your tables, columns, and relationships, exportable as SVG or PNG.

Keeping your data safe

Read-only by design. Every query Sqliqs generates is read-only — it cannot insert, update, or delete your data.
Use a read-only database user. For defense in depth, connect with a database role that only has SELECT permission. Then even a bug can't touch your data.
Your keys stay with you. API keys and connection strings live only in your browser and are sent over an encrypted connection per request — never stored on our servers. See the Privacy Policy.
Don't share your keys. Treat your API keys like passwords. Rotate them with your provider if one is exposed.

Limitations

A few things to understand before you depend on Sqliqs:

Storage is local to your browser. Projects, chat history, settings, and keys are saved in this browser's local storage (IndexedDB). There is no cloud storage and no sync.
No cross-browser or cross-device access. Open Sqliqs in a different browser, profile, or device and your projects won't be there. Clearing your browser data erases them permanently.
The free model is limited. The built-in free model has a small daily request cap and slower responses. It's for trying things out — bring your own key for sustained or heavy use.
Results are capped. Queries return a limited number of rows, so for large datasets Sqliqs prefers summaries and aggregates over dumping raw rows.
AI can be wrong. Models occasionally write a query that doesn't match your intent. The query is always shown — glance at it, and verify anything important.

Tips for the best experience

Back up anything you want to keep — export reports and diagrams, since history isn't synced.
Start a new chat for a new topic to keep context focused.
For analysis over big tables, ask for aggregates ("average", "count by…") rather than every row.
If a response is cut off or errors out, try again or switch to a model with more capacity.

User Guide