Can camelAI answer questions using connected database data?
camelAI answered three of the four database questions as the reference queries did: 10 more orders in August than July, an August total of $14,403.87 across 28 orders, and a tie at 6 units for the top product. Asked for orders in the last 30 days as of 2 September, it counted from 4 August instead of 3 August and gave 24, against the reference 26.
3 of 4 test cases passed · 1 failed
Every test case in this scenario has a result.
The 4 test cases
Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on these test cases.
Order count in a defined periodFailEvidence
“As of September 2, 2026, how many orders were placed in the last 30 days?”
24 orders were placed in the 30 calendar days from August 4 through September 2, 2026 (inclusive), based on public.orders.placed_at in the connected database. The latest matching order was placed on September 1 UTC.
Transcribed from Proof 1 · Screenshot
Every message the tool sent, in order (2)
- Query camelAI ran · not part of the answer
const r = await tools.camel__connections_query({connection:"customer_support_db", query:"SELECT COUNT(*) AS orders_last_30_calendar_days, MIN(placed_at) AS earliest_in_window, MAX(placed_at) AS latest_in_window, COUNT(*) FILTER (WHERE placed_at IS NULL) AS null_placed_at FROM public.orders WHERE placed_at >= TIMESTAMPTZ '2026-08-04 00:00:00+00' AND placed_at < TIMESTAMPTZ '2026-09-03 00:00:00+00'"}); return r; - Answer
The final response, printed above.
| ✗ | Expected: 26 orders from 3 August through 2 September 2026: the 30 days before the as-of date plus the as-of date itself.Found: camelAI answered 24 orders. The reference query returned 26 for 3 August through 2 September, and 24 for camelAI's own window of 4 August through 2 September. |
| ✗ | Expected: The answer covers 3 August through 2 September 2026 inclusive (30 days before the as-of date, plus that date), stated or evident.Found: camelAI stated its window as 4 August through 2 September 2026, inclusive, and the query it ran used that window, which starts a day after 3 August. |
The test database was refreshed after the test was written, so each answer is compared with a reference query run at test time. The window is the 30 days before 2 September plus that day, so it starts 3 August; camelAI counted from 4 August and returned 24, not 26.
Date window stated in answer. camelAI's answer named its date window and the latest matching order, and its query also checked for orders with no order date, so the one-day difference in the window is visible from the answer itself.
Order-count difference between two monthsPassEvidence
“As of September 2, 2026, how many more orders were placed in August 2026 than in July 2026?”
10 more orders were placed in August 2026: 28 in August versus 18 in July, based on the orders.placed_at dates in the connected database.
Transcribed from Proof 1 · Screenshot
Every message the tool sent, in order (2)
- Query camelAI ran · not part of the answer
return await tools.camel__connections_query({connection:"customer_support_db", query:"SELECT date_trunc('month', placed_at)::date AS month, COUNT(*) AS orders_placed, MIN(placed_at) AS first_order, MAX(placed_at) AS last_order FROM orders WHERE placed_at >= TIMESTAMPTZ '2026-07-01 00:00:00+00' AND placed_at < TIMESTAMPTZ '2026-09-01 00:00:00+00' GROUP BY 1 ORDER BY 1"}); - Answer
The final response, printed above.
| ✓ | Expected: August had 10 more orders than July (28 against 18), as a reference query on the database returned at test time.Found: camelAI answered 10 more orders in August. The reference query returned 28 August orders, 18 July orders and a difference of 10. |
| ✓ | Expected: August had more orders than July (28 against 18 at test time).Found: camelAI named August as the larger month and gave both counts in its answer: 28 in August and 18 in July. |
Both the difference and its direction had to agree with a reference query run at test time. camelAI's answer did, and the query it ran used the same calendar-month boundaries as the reference.
Total order value in a calendar monthPassEvidence
“As of September 2, 2026, what was the combined total amount of all orders placed in August 2026?”
Orders placed in August 2026 totaled $14,403.87 USD across 28 orders. I summed orders.total_amount for orders placed from August 1 through August 31, 2026.
Transcribed from Proof 1 · Screenshot
Every message the tool sent, in order (2)
- Query camelAI ran · not part of the answer
const r=await tools.camel__connections_query({connection:"customer_support_db",query:"SELECT currency, COUNT(*) AS order_count, SUM(total_amount) AS combined_total, MIN(placed_at) AS first_placed_at, MAX(placed_at) AS last_placed_at FROM public.orders WHERE placed_at >= TIMESTAMPTZ '2026-08-01 00:00:00+00' AND placed_at < TIMESTAMPTZ '2026-09-01 00:00:00+00' GROUP BY currency ORDER BY currency"}); return r; - Answer
The final response, printed above.
| ✓ | Expected: The August 2026 total is $14,403.87 across 28 orders, the result of a reference query on the database at test time.Found: camelAI answered $14,403.87 across 28 orders; the reference query returned 28 orders totalling 14,403.87 US dollars. |
| ✓ | Expected: The answer covers August 2026 and all orders, with none left out because of their status.Found: camelAI stated that it summed the order totals for 1 through 31 August 2026, and its query had no status filter, like the reference query. |
The total had to agree with a reference query run at test time, over every August order. camelAI's $14,403.87 across 28 orders equals the reference, and its query used the same window with no status filter.
Total grouped by currency. camelAI split its August total by currency instead of assuming one; its single US dollar row shows no mixed-currency sum is hidden in the $14,403.87.
Top product by units ordered in a monthPassEvidence
“As of September 2, 2026, which product had the most units ordered in August 2026, and how many units?”
There's a tie for the most units ordered in August 2026: Brass picture frame, 8×10 and Round wall mirror, 24in, with 6 units each. I summed order-item quantities for orders placed from August 1 through August 31 (UTC).
Transcribed from Proof 1 · Screenshot
Every message the tool sent, in order (2)
- Query camelAI ran · not part of the answer
const r = await tools.camel__connections_query({connection:"customer_support_db", query:"SELECT p.id AS product_id, p.name AS product, SUM(oi.quantity)::bigint AS units_ordered, COUNT(DISTINCT o.id)::bigint AS orders FROM orders o JOIN order_items oi ON oi.order_id=o.id JOIN products p ON p.id=oi.product_id WHERE o.placed_at >= TIMESTAMPTZ '2026-08-01 00:00:00+00' AND o.placed_at < TIMESTAMPTZ '2026-09-01 00:00:00+00' GROUP BY p.id, p.name ORDER BY units_ordered DESC, p.name ASC LIMIT 10"}); return r; - Answer
The final response, printed above.
| ✓ | Expected: The top product is the tie between Brass picture frame, 8x10 and Round wall mirror, 24in, as a reference query returned at test time.Found: camelAI reported a tie between Brass picture frame, 8x10 and Round wall mirror, 24in. The reference query returned both at 6 units and the next products at 5. |
| ✓ | Expected: 6 units for each tied product, as the reference query returned at test time.Found: camelAI answered 6 units each; its own query returned 6 units for both products and 5 for the next two. |
Any product tied for the most units counts as the top product. The reference query run at test time showed a tie at 6 units, and camelAI named both tied products with 6 units each, using the same join and date window.
Tie kept visible in query. camelAI said plainly there was a tie instead of picking one product, and its query listed the top ten products by units and then by name, so both tied products show in its work trace.
Configuration and setup
How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.
How this scenario is graded
How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.
Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.
- Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
- Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
- Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.
Where this sits in the benchmark
This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.
| Level | Name | Scope |
|---|---|---|
| Benchmark | AI Database Agents → | 28 scenarios · 11 tools |
| Capability | Question Answering → | |
| Scenario | The answer is available in the connected data → | |
| Tool | camelAI → |
Global scenario definition → · Global capability definition → · camelAI product page →
History of this result
What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.
Act on this result
Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.
You run the same kind of test against your own setup and get the same behaviour.
Agree →Yours behaves differently. Tell us what you got, with a screenshot if you have one.
Disagree →Something here is wrong — a reference value, a transcription, a grade.
Report an issue →Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.
Vendor notice →The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.
Verify the proof files
These files support this result. Open a file to inspect the original evidence.
File fingerprints (SHA-256)
A fingerprint identifies the exact file used for this result.