Frequently asked questions
AI features in DbVisualizer FAQ
DbVisualizer includes some strictly opt-in AI features to help users understand their schemas, generate and optimize SQL and understand errors. We've gathered the most common questions and answers regarding these features below.
The AI Assistant helps users:
- Generate SQL from natural language
- Improve existing scripts
- Convert SQL between dialects
- Explain errors and obtain database guidance
Users can choose to include database metadata as context.
The Assistant provides suggestions. It does not autonomously execute SQL; a human user always initiates execution. Users must review generated statements before running them.
No. The underlying AI models are supplied by external providers, whose brands and model names are shown in the interface. DbVis develops and operates the DbVisualizer integration and backend.
The backend prepares the customer's prompt and selected context, reduces schema context where appropriate and adds application instructions, such as requesting SQL when the user works in the SQL editor. DbVis does not add a proprietary knowledge corpus. These integration functions do not mean that DbVis developed the underlying model.
The selected external provider and model are identified in the AI Assistant. Users can choose from the models made available in their installation and service configuration.
The subprocessors page lists current AI providers services. That page is a contractual inventory and does not establish which exact models are enabled.
No, not currently. Customer-provided AI services, API keys and internally hosted models are not currently supported. Bring-your-own AI is on the roadmap, with no committed availability date.
Selecting one of the models offered by DbVis does not connect the Assistant to a customer's own AI account.
AI use is governed by the applicable customer agreement, EULA and AI Addendum. The DbVis Data Processing Agreement governs personal data processing within its scope. The current DPA and subprocessor list are available at https://www.dbvis.com/dpa and https://www.dbvis.com/subprocessors/.
DbVis has contractual arrangements with its AI suppliers. The applicable provider terms and DPA must be checked for the service actually used.
Yes, these questions are covered in the Data Processing Addendum of the AI Provider. The applicable DPAs can be found here: https://openai.com/policies/data-processing-addendum, https://www.dbvis.com/dpa/.
Use your caution when you decide what to submit to the AI.
DbVisualizer processes the content while handling the request. Its backend receives the prompt and selected context, prepares the provider request and relays the response. DbVis Software AB does not retain AI prompt, context or response content.
DbVis Software AB retains operational metadata recording that an AI API call occurred, its date and time, the selected model, token consumption and moderation flags. These records describe AI requests, not SQL execution.
External-provider retention follows the applicable provider DPA and service configuration, typically ~30 days.
The optional database context is metadata, such as schema structure, not table rows. A request can nevertheless contain database values or personal data if users include them in prompts, pasted SQL, examples or error messages. Earlier conversation content can also be part of a request.
Users should review the information they submit and follow their organisation's rules for sharing database information.
Optional context can include the database engine and version, driver information and schema structure, including tables, columns and relationships. Users can select the context they share.
The client shows selected context. The backend may reduce schema content using vector-based selection and adds application instructions. A view of selected client context should therefore not be described as a byte-for-byte preview of the final provider request.
Compliance is currently not required since it has been determined that the AI assistant in DbVisualizer falls outside of Article 3 in the EU AI Act. In the language of this law, DbVis Software is a distributor of the AI feature, and Open AI, Anthropic and, and potentially others are providers.
No, and there are no current plans for certification, based on the current limited scope of the AI assistant in DbVisualizer. But we are monitoring our AI Provider(s) for both ISO certification and EU AI act compliance. We are working with ISO 27001 standards.
There is an AI balance included per user, set by your license type, at no additional cost.
For an initial limited time only, DbVis will top up your balance free of charge. We do however reserve the right to deny additional balance if we deem usage abusive. When your AI balance runs out, the AI feature will stop processing prompts.
Please note that you need an internet connection to run the AI feature. An active subscription of the Pro version or an Eval license is required to use the AI feature.
We are responsible for embedding and transferring your prompts safely to the AI, but not for how the AI handles your data, nor can we be held liable for the quality of the results or any consequential damages. The same caution should be applied when using the AI feature in DbVisualizer as when using a commercially available AI service directly, including the prompts you write, the data you submit, and how you scrutinize the results.
Customers control deployment, user permissions and what information may be sent to AI. Users must assess generated SQL for correctness, appropriate access, performance and possible changes to data or schemas before they initiate execution. AI output may be inaccurate or unsuitable.
Yes. AI is opt-in and disabled by default. Administrators can disable AI for all users of an installation by adding this property to dbvis-custom.prefs in DBVIS-HOME/resources:
dbvis.-AiDisabled=true
This installation property requires an active Pro licence. AI use itself also requires an active eligible licence.
Customers should centrally deploy the configured installation and make the relevant configuration folder read-only for users. They must also govern which installations users may run. A separately downloaded, unmanaged installation can allow AI use when the user has an active eligible licence.
No, the AI feature is strictly user initiated.
- The user signs in through the DbVisualizer Authorization Server. The client receives a short-lived access token.
- The client sends the prompt, conversation history and selected optional context to the DbVis backend over HTTPS.
- The backend uses vector-based selection to reduce unnecessary schema context and token usage. It adds application instructions, such as requesting SQL for an editor request. It does not add a proprietary DbVis knowledge corpus.
- The backend sends the prepared request to the user's selected external provider and model. The response returns through the backend to the client.
- DbVis retains the operational metadata described in question 7, but not AI content. SQL execution remains a separate action initiated by a human.
Yes. Prompts, SQL literals, object names, error messages and conversation history can contain personal data, confidential information or credentials. Submit only information your organisation permits. Remove unnecessary identifiers and secrets before sending.
Do not assume that the Assistant automatically masks or removes sensitive content.
The DPA sets out confidentiality and personnel-access requirements. Backend non-retention of content does not, on its own, establish that transient content can never be accessed. Usage metadata is retained separately. https://www.dbvis.com/dpa
DbVis records that an AI API call occurred, its date and time, the selected model, token consumption and moderation flags. It does not retain prompt or response content in the backend.
These usage records do not prove which SQL a user executed. When a user copies generated SQL elsewhere, it is raw text without an AI-origin trace. Database execution auditing is a separate control.
AI requests pass through DbVis-operated services and the selected external provider. The current subprocessor page identifies OpenAI and Anthropic AI services in the United States.
Provider retention follows the applicable provider agreement and service settings. DbVis backend content non-retention does not mean providers retain nothing.
AI can be disabled for an installation. There is no per-connection AI disable switch. Customers can use separately governed installations for environments with different policies.
SQL execution permissions control what statements may be run. They should not be treated as a prohibition on sending AI requests from a particular connection.
The user guide describes archived conversations that users can reopen and continue. This history is separate from DbVis Software AB backend content non-retention.
The AI does not autonomously execute SQL. A human user always initiates execution, either through an available execution action or after copying or inserting the SQL into an editor.
User-initiated execution can change data or schemas if the statement and applicable permissions allow it. The user guide describes separate Assistant execution permissions combined with connection permissions, with the stricter setting applying.
Christer Wilsby is the DbVis Software AB compliance contact. The published subprocessor page directs privacy and subprocessor questions to privacy@dbvis.com.
For suspected disclosure or misuse, retain the time, account identifier, selected model and a description of the issue. Avoid sending additional sensitive content until the appropriate reporting channel is established.
AI requests depend on service availability, connectivity, an active eligible licence and sufficient credits. Administrators can disable the Assistant using the installation setting. The applicable agreement governs service availability and suspension. DbVisualizer, the core application, can be used irrespective of AI availability.
The client needs an Internet connection with access to the DbVisualizer authentication and AI backend services. In the described architecture, the DbVis backend contacts the external AI provider and returns the response to the client.
AI content is transmitted over HTTPS from the client to the backend; enabling AI involves server-side processing.
The customer retains ownership of prompts and data submitted to the service. Rights in generated outputs and restrictions on their use follow the applicable agreements.