Azure / Azure/data-api-builder

[Enhancement]: Let aggregate_records (MCP) compute multiple aggregations in one call (single GROUP BY pass)

Open
#3,725 1 comment 1 reaction 0 assignees View on GitHub
cri enhancement mcp-server
Dominant language
C#
Stars
1.5k
Forks
370
Avg merge
3d 17h
Merged PRs (30d)
8

Description

### Is your feature request related to a problem?

The MCP `aggregate_records` tool accepts only **one** aggregate per call (`function` + `field`). To retrieve several metrics over the same grouping, a client must issue **N separate calls**, each generating its own query with the same `GROUP BY`:

```sql
SELECT region, SUM(amount) FROM dbo.Sales WHERE ... GROUP BY region
SELECT region, AVG(amount) FROM dbo.Sales WHERE ... GROUP BY region
SELECT region, SUM(cost) FROM dbo.Sales WHERE ... GROUP BY region
```

This scans the same rows and recomputes the same `GROUP BY` **N times**, plus N× query-planning / result-shaping / round-trip overhead. On large tables/views this is a real, avoidable database cost. Issuing the N calls concurrently on the client side only reduces latency — it does not reduce total DB work and can increase peak load/contention.

### Describe the solution you'd like

Allow `aggregate_records` to accept **multiple aggregations in a single call**, producing one `SELECT` with one `GROUP BY` pass, e.g.:

```json
{
"entity": "Sales",
"aggregations": [
{ "function": "sum", "field": "amount", "alias": "sum_amount" },
{ "function": "avg", "field": "amount", "alias": "avg_amount" },
{ "function": "sum", "field": "cost", "alias": "sum_cost" }
],
"groupby": ["region"],
"orderby": ["sum_amount desc"]
}
```

which maps to a single query:

```sql
SELECT region, SUM(amount) AS sum_amount, AVG(amount) AS avg_amount, SUM(cost) AS sum_cost
FROM dbo.Sales
WHERE ...
GROUP BY region
```

`orderby`/`having` should be able to reference the aggregate **aliases** (e.g. `sum_amount`), consistent with how grouped results are projected.

### Why this looks feasible

DAB already builds multi-aggregation queries internally for GraphQL: the shared `SqlQueryStructure.GroupByMetadata.Aggregations` is a `List` and `BaseTSqlQueryBuilder.BuildAggregationColumns(...)` emits all requested aggregate expressions into a single `SELECT`. The MCP `aggregate_records` tool (`AggregateRecordsTool`) currently parses and loads only a single `AggregationOperation`. Extending its argument parsing to accept an array and pushing multiple `AggregationOperation`s onto the existing structure would reuse the machinery GraphQL already relies on.

### Backward compatibility

Keep the current single-metric form (`function` + `field`) working as-is; add the new optional `aggregations` array alongside it. Single-metric callers are unaffected; multi-metric callers opt in.

### Additional context

- Related: #3695 (a separate correctness bug in the aggregate `ORDER BY` for `dwsql`). This enhancement is independent of that fix.
- Benefit is largest for wide analytical tables/views where several measures are aggregated over the same grouping in one logical request.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.