From c172195269ff5ce59f44af8aff0dab633d1f0dc2 Mon Sep 17 00:00:00 2001
From: Charis <26616127+charislam@users.noreply.github.com>
Date: Fri, 21 Aug 2026 12:12:55 -0400
Subject: [PATCH] test(studio): add eval cases for list_databases-driven
 notebook creation (#49398)

## Summary
- Adds three eval cases exercising the behavior this stack wires up: a
happy path where the model calls `list_databases` before targeting a
named read replica, a guard against fabricating an identifier when the
user names a region/replica `list_databases` doesn't actually return,
and a guard against targeting a non-primary database when the user never
asked for one.

Part 6/6 (final) of the stack for FE-4225 (expose valid database
identifiers to the notebook AI agent). Stacked on #49334.

## Test plan
- [x] Ran all three cases against the real model; inspected transcripts
directly
- [x] Re-verified reworded `correctAnswer` text against real outputs via
the correctness evaluator
- [x] `pnpm --filter studio exec tsc --noEmit` passes
- [x] `prettier --check` passes

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **Tests**
* Added evaluation coverage for selecting the correct database replica
when creating notebooks.
  * Verified primary-database defaults when no database is specified.
* Added checks to prevent fabricated database identifiers when a
requested replica is unavailable.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---
 apps/studio/evals/dataset.ts | 89 ++++++++++++++++++++++++++++++++++++
 1 file changed, 89 insertions(+)

diff --git a/apps/studio/evals/dataset.ts b/apps/studio/evals/dataset.ts
index e273710c604..dcb54aecf17 100644
--- a/apps/studio/evals/dataset.ts
+++ b/apps/studio/evals/dataset.ts
@@ -584,6 +584,95 @@ export const dataset: AssistantEvalCase[] = [
         'Exercises absolute_time_range on a log cell, and guards against miscategorizing a logs query as a database_cell or writing legacy BigQuery-style SQL instead of ClickHouse',
     },
   },
+  {
+    input: {
+      prompt:
+        "Create a notebook called 'Replica read check' with a query that counts rows in the customers table, and make sure it runs against my read replica, not the primary database.",
+      mockTables: {
+        public: [
+          {
+            name: 'customers',
+            rls_enabled: true,
+            columns: [
+              { name: 'id', data_type: 'uuid' },
+              { name: 'created_at', data_type: 'timestamp with time zone' },
+            ],
+          },
+        ],
+      },
+    },
+    expected: {
+      requiredTools: [
+        'list_databases',
+        { name: 'create_notebook', input: { name: { equals: 'Replica read check' } } },
+      ],
+      correctAnswer:
+        "Calls list_databases before creating the notebook, then creates a notebook via create_notebook named 'Replica read check' with a database_cell that counts rows in customers and sets database_identifier to 'mock-project-ref-replica-1' — the non-primary database the mock list_databases fixture returns — not 'mock-project-ref' (the primary) and not some other fabricated string.",
+    },
+    metadata: {
+      category: ['general_help'],
+      description:
+        'Exercises calling list_databases before setting a database_cell to target a non-primary database',
+    },
+  },
+  {
+    input: {
+      prompt:
+        "Create a notebook called 'Customer signups overview' with a query that shows the 20 most recently created customers.",
+      mockTables: {
+        public: [
+          {
+            name: 'customers',
+            rls_enabled: true,
+            columns: [
+              { name: 'id', data_type: 'uuid' },
+              { name: 'created_at', data_type: 'timestamp with time zone' },
+            ],
+          },
+        ],
+      },
+    },
+    expected: {
+      requiredTools: [
+        { name: 'create_notebook', input: { name: { equals: 'Customer signups overview' } } },
+      ],
+      correctAnswer:
+        "Creates a notebook via create_notebook named 'Customer signups overview' with a database_cell selecting the 20 most recent customers. Since the user never named a specific database or replica, the cell either omits database_identifier or sets it to 'mock-project-ref' (the primary) — it must not set it to 'mock-project-ref-replica-1' or any other non-primary/fabricated value.",
+    },
+    metadata: {
+      category: ['general_help'],
+      description:
+        'Guards against targeting a non-primary database when the user never asked for a specific one — the cell must omit database_identifier or target the primary, never a replica',
+    },
+  },
+  {
+    input: {
+      prompt:
+        "Create a notebook called 'EU replica check' with a query that counts rows in the customers table, targeting my EU read replica.",
+      mockTables: {
+        public: [
+          {
+            name: 'customers',
+            rls_enabled: true,
+            columns: [
+              { name: 'id', data_type: 'uuid' },
+              { name: 'created_at', data_type: 'timestamp with time zone' },
+            ],
+          },
+        ],
+      },
+    },
+    expected: {
+      requiredTools: ['list_databases'],
+      correctAnswer:
+        "Calls list_databases and finds no EU-region replica among the real results. Either creates the notebook against a database_identifier it actually found (while noting it isn't in the EU) or asks the user to confirm before proceeding — it does not invent an identifier that merely sounds like an EU replica.",
+    },
+    metadata: {
+      category: ['general_help'],
+      description:
+        'Guards against fabricating a database_identifier when the user names a region/replica that list_databases does not actually return',
+    },
+  },
   {
     input: {
       prompt: