Browse/Search Model
list_model is the main way the assistant reads your semantic model: list objects, search by name/metadata (and, when permitted, by expressions), and do lightweight analysis like “describe” and “preview”.
It’s the best starting point for almost every task because it builds accurate context before querying or editing.
You describe the goal in plain language (“find measures related to margin”), and the LLM uses list_model behind the scenes.
What to ask the LLM (quick prompts)
“Summarize the model: main tables, relationships, measures, calc groups, and roles.” “List tables and highlight likely fact tables and date tables.” “Search for ‘margin’ across names and descriptions; then expand to expressions if needed.” “Show relationships involving ‘Sales’ and flag inactive or bi-directional ones.” “Describe table ‘Sales’: key columns, data types, hidden fields, and relationships.”Typical use cases
- Find “where is X defined?” (measure/column/table).
- Find “what mentions X?” (names/descriptions; expressions when permitted).
- Document a model quickly (inventory + summaries).
- Spot common issues (missing date table, suspicious relationships, unused objects).
What the assistant can look at with this tool
Depending on your deployment mode, the assistant can browse most semantic model object types, including:
- tables, columns, calculated columns, measures
- relationships, hierarchies
- calculation groups/items, KPIs
- partitions and named expressions
- roles, perspectives, cultures/translations, calendars
- model properties (high-level metadata)
How to work with it effectively
Search recipes (copy/paste prompts)
“Find the measure behind a report visual”
“Search for measure named like ‘Gross Margin’ and show exact table + name. If multiple, list candidates with short summaries.”“Find objects by business term”
“Search names and descriptions for ‘churn’. Group by type. Then tell me which 3 measures are most likely relevant.”“Find where a concept is implemented in DAX”
“Search expressions for ‘SAMEPERIODLASTYEAR’ and list matching measures/calc items. Summarize patterns you see.”“Locate Power Query logic”
“Search M expressions for ‘SharePoint’/‘SQL’/‘Snowflake’ and list affected tables/partitions.”“Find relationship issues”
“List relationships and flag: inactive, many-to-many, bi-directional, and suspicious key columns.”Locked-down environments (mode behavior)
| Mode | Availability |
|---|---|
| Full mode | AvailableAvailable |
| Read-only mode | AvailableAvailable |
| Browse-only mode | LimitedLimited |
Notes: Browse-only: list/search only; expressions/details suppressed.
Common workflows
Model “tour” (fast onboarding)
“Give me a short guided tour of this model: what it’s for, main tables, key measures, relationships, and security.”Find the source of a number
“I’m seeing ‘Gross Margin %’ in a report. Find the measure, explain it, and list what it depends on.”Identify risky changes before editing
“Before renaming ‘Customers[Region]’, find everything that references it and summarize impact.”
Troubleshooting
Ask: “Narrow to measures only” / “Limit to one table” / “Use name-only search first”.
You may be in browse-only mode, or the assistant is intentionally minimizing exposure. Ask: “Explain what’s suppressed and why.”
The environment may cap results. Ask the assistant to raise/lower row limits (if allowed) or to return aggregates instead.