AI assistant
The Boldo AI assistant helps you work with your knowledge base without leaving the current context. Use it to ask questions about your architecture, prepare changes, enrich asset fields, and explore diagrams, views, charts, nested maps, or process flows.
AI access depends on your organization plan and the AI access switch that administrators set for your user. What the assistant can do then depends on your organization's AI access level and your own Boldo rights.
The assistant follows your Boldo rights. It can only use data that your account is allowed to access.
Access
Open the assistant with the AI Assistant button (the sparkles icon) in the top bar. It opens as a side panel over the page you are on, so you keep your current context. From the panel header, use Open to expand the conversation into the full AI workspace.
When you open the assistant from a diagram, view, chart, nested map, process flow, or asset page, Boldo can pass that screen context to the assistant. This helps with questions such as "explain this diagram" or "suggest improvements for this view".
Ask questions
Use the assistant to inspect and explain your Boldo data.
For example, you can ask it to:
- explain the current diagram or process flow
- find assets that match a business question
- summarize the relationships around an application
- compare what appears in a view or chart
- help you understand the metamodel
The assistant can use Boldo tools to read data and documentation. It can also display a specific view, chart, diagram, nested map, process flow, or asset inline in the conversation when you ask to see it. When a response depends on the current screen, keep the relevant item open before asking.
AI can make mistakes. Verify important answers before using them for decisions, communication, or changes to your repository.
Prepare proposals
The assistant can prepare proposals for assets, relationships, views, charts, dashboards, diagrams, nested maps, or process flows. It does not silently change your knowledge base.
Review the proposal before applying it. Boldo validates the final action with the same rules as the normal interface, so your permissions, organization limits, and required editor rights still apply. Some proposals are only offered in context: updating a view, chart, dashboard, diagram, nested map, or process flow is available when you have that item open, and process flow proposals require a plan that includes process flows.
Enrich asset fields
When AI is enabled, asset forms show an AI bar at the bottom. Use it to fill, refine, or connect fields without leaving the form, in both create and edit modes.
Type a request in the bar and send it, for example "fill in the missing fields", "set criticality to high and rewrite the description", or "link this application to the CRM business capability". The assistant updates the form fields directly, and can search for and attach related assets. Mention another asset with @ to have the assistant link it.
Changes are applied to the form, not saved. Review them before validating the asset. Use Undo AI changes in the bar to revert the last request, and hover the history icon next to a changed field to see its previous value.
Enriching fields uses your organization data, so it is available only when the AI access level is Level 2 - Business data.
Use projects and attachments
AI projects group conversations and documents around a specific work context. Use them when you want the assistant to answer with project-specific material, such as a migration brief, audit notes, or architecture documentation.
From the prompt composer, you can select or create a project. You can also attach a file to a message.
When you save a document to a project, Boldo splits its text into small chunks and turns them into embeddings, a searchable index of what each passage means. When you chat inside the project, the assistant searches that index, by meaning and by keyword, and pulls in only the most relevant passages to ground its answer instead of reading every document in full. This is why project documents must be text-based and why long documents are capped.
Project limits are intentionally small so the assistant stays focused:
- up to 5 AI projects per user
- up to 6 saved documents per project
- up to 1 MB per uploaded document
- project document text is limited to 120,000 characters before being indexed for retrieval
Saved project documents support text-based files such as:
- text or markdown
- CSV
- JSON
- XML
- HTML
- YAML
PDF and Word documents (.doc and .docx) are not supported as project documents yet. Images are not stored as project documents either. If you need to use that content, export or convert it to a supported text format first.
For one-off message attachments, text files up to 1 MB are accepted and long content is truncated to 40,000 characters before being sent to the model. Image files may also be accepted when the selected model supports vision. If the selected model does not support vision, use a text attachment or choose a vision-capable model.
Choose a model
When several models are available, the prompt composer lets you choose the model before sending a message. Models are grouped by provider, with Boldo-hosted models first and the recommended default marked accordingly.
Not every model can read images. Models that support vision are marked with a small image icon; the others are text-only. All provider (BYOK) models (Claude, GPT, Gemini, and Mistral) accept images, while only some Boldo-hosted models do, and the recommended default is text-only. If you attach an image to a model without vision, the image is left out of the request and the assistant tells you, so pick a model marked with the image icon when you need to work from a screenshot, photo, or exported diagram.
Boldo-hosted models use the organization Boldo-hosted budget. BYOK models use the corresponding provider key configured by an administrator, and appear only when that key is present and readable. If a BYOK model returns a provider or key error, ask an administrator to check in AI settings that the provider key is valid and not expired or out of quota.
Good prompts
Good prompts give the assistant enough context and define the expected output.
| Weak prompt | Better prompt |
|---|---|
| "Check this" | "Explain the main dependencies in this diagram and list missing relations to verify" |
| "Create apps" | "Prepare assets for these five applications, with owner, criticality, and lifecycle status" |
| "Fix the view" | "Update this view to show only active applications grouped by business domain" |
For changes, ask for a proposal first and review it before applying.