You decide what AI can do.
Keep AI advisory, allow read-only investigation, or approve each write. AI activity records what was asked. Per-connection notes can carry schema conventions the assistant should follow.
DBNox Intelligence
DBNox AI writes and explains queries in the editor, reads execution plans, and can investigate with typed database tools. Writes wait for your approval. A local model can stay on the machine.
Why is this query taking 12 seconds?
Investigating
Finding
orders scans 8.7M rows while returning 214.
No existing index covers store_id + status + created_at
Production mode: Advisory only
AI Assist and /ai turn a plain-language request into SQL, or into a Redis command on a Redis connection. Search engines are asked in that engine's query form.
Explain the statement under the cursor, or a notebook cell, without executing it.
Visual EXPLAIN can send the plan to the model and show the analysis beside the operators.
The investigator can call EXPLAIN, statistics, indexes, locks and sessions, then explain what those tools returned.
Describe what you want inside a container. DBNox asks the model for one shell command and shows the prompt before you run it.
Run Qwen through the bundled llama.cpp server, and start it when DBNox opens, or use Ollama. Cloud providers stay optional.
Keep AI advisory, allow read-only investigation, or approve each write. AI activity records what was asked. Per-connection notes can carry schema conventions the assistant should follow.
API keys stay in the operating-system keychain. A local Llama or Ollama runtime needs no cloud key. Choosing a model does not grant it database permissions.