Dataverse People Lookup

Dataverse People Lookup

Answer "What is this person's title, location, and manager?" by querying RoverPeople through the Dataverse MCP.

Prerequisites

1. Dataverse MCP configured and authenticated (user-Dataverse).
2. If STATUS.md indicates auth is required, call mcp_auth for user-Dataverse first.
3. Optional: read ME.md at project root for context on the requester.

Workflow

Run the 4-step Dataverse pipeline. Do not skip steps.

Step 1 — Identify data product

identify_dataproducts
  user_query: "employee profile title location manager for [PERSON]"

Expected result: RoverPeople (ai_ready).

Step 2 — Shortlist tables

shortlist_tables
  data_product: roverpeople
  user_query: "title, location, manager for [PERSON]"

Expected table: ROVER_PEOPLE_CURR in schema RHAI_MARTS.

Step 3 — Generate SQL

get_sql
  data_product: roverpeople
  tables_list: [{ name: ROVER_PEOPLE_CURR, schema: RHAI_MARTS, description: ... }]
  user_query: <see Query patterns below>

get_sql prepends the authenticated user's identity from JWT. Use that for "who am I?" queries.

Step 4 — Execute and present

execute_sql
  sql: <generated SQL>

Present results using the output template. See reference.md for fields, matching rules, and edge cases.

Query patterns

Pass these as user_query to get_sql:

Output template

## [Name]

| Field | Value |
|---|---|
| **Title** | {TITLE} |
| **Location** | {GEO}{, OFFICE if available} |
| **Manager** | {MANAGER_NAME} — {MANAGER_TITLE} |
| **Email** | {MAIL} |

**Manager contact:** {MANAGER_EMAIL}

Multiple matches

If more than one person matches, list each in a compact table:

Ask the user to clarify only if matches are genuinely ambiguous (same name, different people).

No match

Say clearly that RoverPeople returned no current employee for the query. Suggest trying email or full name.

Rules

• Current employees only — ROVER_PEOPLE_CURR filters to active staff (term_date null, former_employee false).
• Do not guess — if Dataverse returns nothing, do not fabricate title or manager from memory or LinkedIn.
• Location — prefer GEO (region). Include OFFICE or city fields if present in the query result.
• Manager chain — return direct manager only unless the user asks for the full chain.
• Privacy — this mart is approved for AI use; still present only fields returned by the query.

Examples

User: What is Gunnar's title, location, and manager?

1. identify_dataproducts → RoverPeople
2. shortlist_tables → ROVER_PEOPLE_CURR
3. get_sql → lookup by name "Gunnar"
4. execute_sql → format response

User: Who am I?

Same pipeline; get_sql uses JWT identity automatically.

Additional resources

• reference.md — table schema, MCP tool details, troubleshooting
• dataaccess.md — full Dataverse catalog (project root)