Skip to Content
AgentQuerying Data

Querying Data

The Agent can execute read-only SQL queries against your BigQuery datasets and display results in interactive, sortable tables.

How It Works

When you ask a data question, the Agent:

  1. Identifies the right tables — Browses your datasets and schemas to find relevant data
  2. Writes SQL — Generates a SELECT query based on your question
  3. Executes the query — Runs it against BigQuery with a 30-second timeout
  4. Displays results — Shows data in a sortable table on the right panel

What You Can Ask

Question TypeExample
Counts”How many users signed up last month?”
Aggregations”What’s the average order value by month?”
Top N”Show me the top 10 events by count”
Trends”How has revenue changed week over week?”
Comparisons”Compare event counts between iOS and Android”
Exploration”What columns are in the user_data table?”

Results Table

Query results appear in the right panel as an interactive table:

  • Sortable columns — Click any column header to sort ascending, descending, or reset
  • Formatted values — Numbers show locale formatting (1,234,567), nulls are styled distinctly
  • Row count — Badge showing total rows returned
  • View SQL — Collapsible section showing the exact query that was executed

Safety

  • Only SELECT and WITH (CTE) queries are allowed
  • Mutation statements (INSERT, UPDATE, DELETE, DROP, CREATE, ALTER, TRUNCATE) are blocked
  • Queries are scoped to your account’s datasets — you cannot access other accounts’ data
  • Results are limited to 500 rows
  • Queries time out after 30 seconds

Dataset Naming

Vendo BigQuery datasets follow a naming convention:

  • {account_slug}_{account_id_prefix} (e.g. abode_money_c23a2914a767) — Your account’s unified dataset. Raw imports, modeled tables, and transformation outputs all live here.
  • Some older accounts still use a legacy split: v_{name}_source_* (raw imported data) and v_{name}_prod_* (processed/transformed data).

The Agent knows these conventions and will pick the right datasets for queries automatically.

Charts & Visualization

After running a query, the Agent can visualize the results as a chart. Simply ask:

“Show that as a bar chart” “Visualize revenue by month as a line chart” “Create a pie chart of events by category”

The Agent supports eight chart types:

TypeBest For
BarCategorical comparisons (top 10 events, revenue by channel)
Stacked BarPart-to-whole comparisons by category
Horizontal BarRanked lists
LineTrends over time (daily active users, weekly revenue)
AreaCumulative trends (running totals, stacked metrics)
PieProportional breakdowns (fewer than 8 categories)
ScatterRelationships between numeric pairs
FunnelStage-by-stage dropoff

The Agent can also render several charts at once as a multi-chart dashboard — just ask for “a dashboard”.

Charts display in the right panel with hover tooltips, legends, and a collapsible “View SQL” section. The Agent automatically picks the best chart type based on your data shape, but you can specify one explicitly.

Tips

  • Be specific about time ranges — “last 30 days” or “since January 2026” helps the Agent write better queries
  • Ask for column names first — If you’re not sure what data is available, ask “What columns are in the events table?”
  • Follow up — After seeing results, ask follow-up questions like “now break that down by country” — the Agent remembers the conversation context
  • Request specific formats — “Show me a monthly summary” or “group by week” to control granularity
Last updated on