Mini DBA AI Assistant works by sending your question plus selected Mini DBA context to an AI provider. This page explains what context can be included, how detail pages add extra evidence, how the default free token allowance works, and how to reduce sensitive data exposure.
What Data May Be Sent
The exact prompt depends on where you open the AI Assistant and which context switches are enabled. Assume the request may include:
- Your typed question.
- Server name and database platform.
- Page name and selected live or historical time range.
- Connection status and refresh time.
- Performance metrics such as CPU, memory, waits, I/O, sessions, batch requests, buffer cache, service tier, host metrics, and database properties.
- Session and request rows such as login, host, program, wait type, command, and query text where Mini DBA has that data.
- Active alert names, severities, detected values, thresholds, and descriptions.
- Query text and execution plan context from detail pages where available.
- Enterprise summary information when using estate-wide AI context.
Do not type secrets into the prompt. Avoid pasting passwords, API keys, connection strings, private customer data, personal data, or confidential payload values. Use approved masking or internal policy before sending context to any AI provider.
Optional Context Switches
The page-level assistant lets you choose Metrics, Sessions, and Alerts.
Metrics is usually the lowest-risk useful option because it focuses on operational counters. Sessions may include login, host, application, query, command, or wait details. Alerts may include names, thresholds, and descriptions that reveal operational policies or customer-specific issues.
Use the smallest context that answers the question:
- Metrics only for broad capacity questions.
- Metrics plus Sessions for blocking, live workload, long-running queries, and concurrency.
- Metrics plus Alerts for alert explanation and incident prioritization.
- Plan context only when analyzing a specific query or execution plan.
Detail Page Context
Some Mini DBA detail pages send additional context to improve answer quality. Execution plan pages can pass plan XML, plan text, SQL text, plan tree output, JSON plans, or parsed plan operators depending on the database platform. Other pages can pass page-specific summaries such as Azure elastic pool context.
Additional context is appended to the live snapshot before the request is sent. This gives the assistant the exact object you are asking about and reduces generic answers.
Default Free Daily Token Limit
The default Mini DBA AI provider includes a limited free daily token allowance per user. The limit is fixed in the product at 100,000 tokens per user per day for the default provider.
When the default provider is used:
- The AI panel shows daily tokens used and the daily limit.
- Token use is tracked in the browser.
- If the threshold is reached, Mini DBA blocks further default-provider requests until the next day.
- The message links to BYO AI settings so you can continue using your own provider.
The token count may come from the AI provider response. If the provider does not return token usage, Mini DBA estimates usage from the system prompt, question, context, conversation history, and response length.
BYO AI And Token Limits
BYO AI uses your configured provider and subscription. It is not limited by the default Mini DBA free daily token quota. Your provider may still enforce its own rate limits, context limits, billing limits, regional restrictions, and model-specific token windows.
If cost matters, reduce token use by asking narrower questions, disabling unnecessary context switches, clearing old conversation history, and using plan context only when the selected plan is relevant.
Conversation History
Mini DBA keeps conversation history for the current server or page context during the user session. This lets you ask follow-up questions without repeating all context. Conversation history can also consume additional tokens because previous messages may be sent with later requests.
Use Clear when:
- You are switching from one issue to another.
- Old assumptions may confuse the next answer.
- You want to reduce token usage.
- You have moved from broad triage to a focused execution plan or alert.
Security Checklist
Before enabling AI broadly, confirm:
- Which AI provider is allowed by policy.
- Whether database names, server names, query text, alert text, and plan XML can leave the environment.
- Whether BYO AI should use a private endpoint, managed identity, or centrally managed service credential.
- Whether browser local storage is acceptable for per-user BYO AI settings.
- Which teams are allowed to use AI on production customer data.
- Whether screenshots, exported responses, and support tickets need masking.
Troubleshooting
If no answer is returned, check AI Provider Settings, provider status, model or deployment name, API key, endpoint URL, API version, network path, and Web Proxy Settings. If Mini DBA says no live data snapshot is available, wait for the next server refresh or open a page that has current model data.
If the default quota bar does not appear, the default provider may not be configured, BYO AI may be enabled, or the browser may not have returned token usage yet.
Related Pages
- Mini DBA AI Assistant
- Configure BYO AI
- Performance Tuning With AI
- AI Provider Settings
- Web Proxy Settings
- MCP Server Settings