Preferences Workflows and Baselines
This page is part of the Preferences (Memory) docs:
Common workflows
1) Audit what’s currently in effect
Prompts:
“Show me the effective preferences and explain the precedence (model/workspace/global).” “Also include a copy/paste-friendly JSON summary I can share with my team.”If you want the answer to include workspace/model scopes, ask:
“Confirm which scopes are available right now, and what I’m currently connected to (Desktop vs Service).”Tip: some clients can show MCP “resources”. If yours supports that, you may also see entries like “Effective Preferences (Memory)” that the assistant can open and summarize.
2) Establish team modeling conventions
Prompt:
“Create a minimal baseline of naming rules + guardrails for this model (measures, columns, folders, descriptions). Then show me what you saved.”3) Make outputs safe to paste into tickets/PRs
Prompt:
“Enable masking and set conservative row limits so anything we paste into chat is safe and small.”4) Share conventions across the team
Prompt:
“Export our preferences as JSON and also summarize them in plain English.”If your org wants change control:
“Export preferences; I’ll open a PR with the JSON; don’t import until I confirm.”Tip: export usually includes the scopes available in your current context. If you need to export workspace/model preferences too, connect to the relevant Service dataset/model first and then ask for export again.
5) Apply a baseline to a new model/workspace
If your org allows changes:
“Here is our approved preferences JSON. Apply it to this environment, then show me the effective preferences.”If your org is locked down:
“Try to apply this preferences baseline. If it’s blocked, explain who needs to do it and what mode/policy prevents it.”6) Clean up: delete or reset preferences
If your preferences have gotten messy, you can ask the assistant to remove specific items or wipe a category in a scope.
Prompts:
“List saved preferences with IDs, then delete the ones that are obsolete (ask me before deleting anything).” “Reset all rules for this model only (keep settings).” “Reset all aliases in global scope.”Export/import baselines (team-friendly)
Preferences support exporting and importing a versioned JSON bundle (intended for sharing and review).
From a user perspective:
- Export gives you a single JSON blob you can paste into a ticket or commit to a repo for review.
- Import applies an approved baseline back into the server’s preferences.
This is also how you move preferences between environments (new machine, new server instance, consultant handoff): export from one environment, import into the other.
Good prompts:
“Export our preferences JSON and summarize the changes since last week.” “We’re onboarding a new workspace. Apply this approved baseline and then show me what’s now in effect.”Things that commonly surprise people:
- Export/import can include global + workspace + model scopes, but only if those scopes are available in your current context (connection + tier).
- Workspace/model scoped import typically requires connecting to the relevant Service dataset/model first.
- In read-only or browse-only deployments, import will be blocked.
- Preferences stores have size limits. If you keep very large rules/alias maps, the server may reject additional items; consolidate or split baselines when needed.
- Import can be applied as a merge or as a full replace (depending on your environment). If you’re unsure, ask the assistant to explain what it will do before applying anything.
If you need change control:
“Export preferences JSON, but do not import anything until I confirm.”How changes take effect
When your environment allows edits, preference changes typically take effect immediately for the running MCP server session.
Prompts:
“I changed preferences; confirm they’re active now and show the effective preferences.”Troubleshooting
You’re likely in a locked-down deployment (read-only or browse-only), or a policy is blocking changes. Ask the LLM to explain what’s allowed in your environment and who can change it.
Workspace-scoped preferences require a Power BI Service connection context (and typically Pro tier). Ask the LLM to confirm which dataset/workspace you’re connected to, and whether workspace scope is available.
Some features (workspace/model scoping, masking settings) require a Pro license. Ask the LLM to propose a fallback approach that works in your tier (e.g., safer row limits still work).
Ask the LLM:
“Am I currently connected to Desktop or Service, and which scopes are available right now?”If you’re not connected to a model, model scope can’t apply; if you’re not connected to Power BI Service, workspace scope can’t apply.
Free tier has a small global-only quota. Ask the LLM to remove old items you don’t need, or consolidate multiple small rules into a single baseline rule.
Preferences item IDs must be unique. Ask the LLM to either (1) update the existing item instead of creating a new one, or (2) create a new item with a different ID.
Deployment references (optional)
If you need admin/deployment controls (environment variables, service-wide defaults, custom masking patterns), use: