FAQ & troubleshooting

Yes. Data Explorer is a management and exploration tool, not a database server. You need at least one running database to connect to. The app itself also needs a separate PostgreSQL database for users, connections, saved queries, dashboards, and other internal metadata.

The app currently supports PostgreSQL, MySQL, SQL Server, MongoDB, and MongoDB Atlas (SRV). The SQL workspace and NoSQL workspace are separate surfaces, but they share the same auth, connections, AI, dashboards, and audit foundations.

Not in the sense of dumping your whole database to the cloud. When you use AI, the app can send prompts, schema context, or the context you explicitly request to the configured AI provider. The app does not render raw HTML from AI and does not automatically send full result sets unless that flow really needs them.

No longer Gemini-only. The app now supports AI routing across Gemini, OpenRouter, Groq, Cerebras, and the Beeknoee explicit lane. The routing mode in the UI sets the broad preference, but explicit model/provider selection remains the strongest control. Each response also shows the actual provider/model so the lane stays transparent.

Because the mode is only one part of the router. Auto balances broadly, Fast / Cheap leans toward lighter lanes, Best Quality leans toward stronger lanes, and Gemini Only locks into Gemini. But the final result also depends on any explicit model/provider selection, whether the prompt contains an image, and whether the request is detected as current-information work.

The current session model uses a refresh token in an httpOnly cookie and an access token in memory. If you still get logged out, check FRONTEND_URL, HTTPS or cookie policy in production, and whether JWT_SECRET or REFRESH_TOKEN_SECRET changed between deploys. Rotating those secrets will invalidate older sessions.

Saved queries are no longer local-only. They are stored in the backend metadata database and support private, team, and workspace visibility. Tab content still has client-side state for UX recovery, but the saved-query flow itself is now server-backed.

Use `npx prisma migrate deploy` in staging and production so schema history remains auditable. Reserve `db push` for disposable local prototypes, never production data.

Many networks block database ports such as 5432, 3306, or 27017 directly. In that case, both local development and the deployed app can lose reachability to cloud or private databases. The most practical workaround is to use a local database for development, or tunnels, VPNs, or agents on heavily restricted networks.

Not by default. Insert into editor only inserts SQL into the editor. Run suggestion is the action that executes it, and the app now shows a confirmation step before running AI-generated SQL to reduce the risk of accidental execution.

Connections now carry policy flags such as readOnly, allowSchemaChanges, allowImportExport, and allowQueryExecution. The backend enforces these guardrails even if the frontend is bypassed. When a statement is blocked, the UI shows a clear reason instead of failing vaguely.

Your privacy is our priority. Data Explorer has removed automatic query sharing based on public email domains (like @gmail.com, @outlook.com) to prevent data leaks between unrelated users. To share, create an Organization and invite your colleagues.

SshTunnelService now applies strict SSRF validation. The system blocks connections to localhost, 127.0.0.1, and RFC1918 internal IP ranges through tunnels to protect your internal network from reconnaissance or SSRF attacks.