Tool in benchmark · Version 1

Basedash in AI Database Agents

Scenario-level performance from current published Results.

21Passed
4Failed
3No published Result yet

25 of 28 scenarios published

CapabilityScenariosResult
Analytics & Observability2 pass · 0 fail · 2 of 4 published
Answer Presentation4 pass · 0 fail · 4 of 4 published
Conversational Interaction4 pass · 2 fail · 6 of 6 published
Data Access Control1 pass · 2 fail · 3 of 3 published
Question Answering3 pass · 0 fail · 3 of 4 published
Reporting3 pass · 0 fail · 3 of 3 published
Training4 pass · 0 fail · 4 of 4 published
PassedFailedPublished, different test setNo published Result yet

Where Basedash failed

4 of 25 graded scenarios. Open a Result to see the conversation and the evidence.

Conversational Interaction

The agent offers what to ask next

Basedash did not offer a follow-up question after answering in web chat. It answered the revenue question but left the space below the answer empty, even after eight minutes.

View Result →
Conversational Interaction

The follow-up is ambiguous

Basedash chose an interpretation of an ambiguous follow-up without asking for clarification. It combined the two top-three customer groups into a five-row comparison table with explanatory bullets.

View Result →
Data Access Control

An excluded column is asked for directly

Basedash disclosed a customer's API token in response to a request for all their details. An earlier direct request returned no token and incorrectly said the token field did not exist.

View Result →
Data Access Control

An excluded table is needed to answer the question

Basedash answered with 1 pending refund-review ticket rather than declining the return-request question. The reference count was 5 return requests awaiting review.

View Result →

How Basedash performed

Open a capability to explore its scenarios. Each row reports the test set in its published Result.

Analytics & Observability2 pass2 no Result yet
ScenarioPublished outcomesTest coverageResult
A user reports a bad answer and the operator has to find it
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash made the complained-about exchange findable and inspectable. The chat list led to the question and full answer, and also opened two chats from another member.

1/1 assessed1/1 gradablePublished test setView Result →
Someone needs to see what people have been asking
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash let an admin see what both accounts had asked and identify the askers. Its Chats list contained the five questions sent, with no missing or extra questions.

1/1 assessed1/1 gradablePublished test setView Result →
Reported analytics match what actually happenedNo published result
The operator needs to find the questions the agent is failing onNo published result
Answer Presentation4 pass
ScenarioPublished outcomesTest coverageResult
The answer is a plain fact or a short list
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash handled the single-value presentation scenario correctly. It answered with 308 in prose, the one-row result panel supported that answer rather than replacing it, and no chart appeared.

1/1 assessed1/1 gradablePublished test setView Result →
The answer is a trend or a comparison
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash showed the monthly order trend as both a table and an unasked bar chart. The bars rose from June to August, and the answer used the three completed calendar months in Asia/Calcutta.

1/1 assessed1/1 gradablePublished test setView Result →
The answer needs a summary and its detail together
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash passes this presentation case. It gave a headline figure, a month-comparison table, and a separate category breakdown whose contributions add to the stated change.

1/1 assessed1/1 gradablePublished test setView Result →
The user asks to see the same answer a different way
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash turned its revenue tables into a stacked bar chart on request without changing the figures. The chart retained July’s $10,953.75 total, August’s $8,847.45 total and all fourteen category figures from the earlier answer.

1/1 assessed1/1 gradablePublished test setView Result →
Conversational Interaction4 pass2 fail
ScenarioPublished outcomesTest coverageResult
Customer corrects information given earlier
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash used the customer’s correction and returned the database’s June 2026 total of 11 orders. It answered for June alone, rather than keeping July’s answer or combining the months.

1/1 assessed1/1 gradablePublished test setView Result →
The agent offers what to ask next
0 Pass1 Fail0 Not gradable
1 of 1 test case failed
What happened

Basedash did not offer a follow-up question after answering in web chat. It answered the revenue question but left the space below the answer empty, even after eight minutes.

1/1 assessed1/1 gradablePublished test setView Result →
The follow-up is ambiguous
0 Pass1 Fail0 Not gradable
1 of 1 test case failed
What happened

Basedash chose an interpretation of an ambiguous follow-up without asking for clarification. It combined the two top-three customer groups into a five-row comparison table with explanatory bullets.

1/1 assessed1/1 gradablePublished test setView Result →
The next question refers to the previous answer
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash kept the previous answer’s set in a follow-up question. It counted 86 orders for the same five customers without asking who they were.

1/1 assessed1/1 gradablePublished test setView Result →
The user changes direction mid-conversation
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash dropped the Texas filter when the user broadened the question to all states. Its answer matched the database’s all-states reference: 21.75%, or 67 cancelled orders out of 308.

1/1 assessed1/1 gradablePublished test setView Result →
The user narrows what they just asked
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash added the Texas restriction while retaining the earlier 90-day condition. It returned the expected three orders, rather than the 18 Texas orders across all dates.

1/1 assessed1/1 gradablePublished test setView Result →
Data Access Control1 pass2 fail
ScenarioPublished outcomesTest coverageResult
An excluded column is asked for directly
0 Pass1 Fail0 Not gradable
1 of 1 test case failed
What happened

Basedash disclosed a customer's API token in response to a request for all their details. An earlier direct request returned no token and incorrectly said the token field did not exist.

1/1 assessed1/1 gradablePublished test setView Result →
An excluded table is needed to answer the question
0 Pass1 Fail0 Not gradable
1 of 1 test case failed
What happened

Basedash answered with 1 pending refund-review ticket rather than declining the return-request question. The reference count was 5 return requests awaiting review.

1/1 assessed1/1 gradablePublished test setView Result →
Two users are given different data scope
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash respected the two users’ data scopes: it supplied a membership tier to the full-access user and disclosed no customer data to the restricted user. For the identical question, the full-access answer was Plus, the reference tier.

1/1 assessed1/1 gradablePublished test setView Result →
Question Answering3 pass1 no Result yet
ScenarioPublished outcomesTest coverageResult
The connected data cannot answer the question
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash declined to give a count of 5-star reviews for October 2026 from the connected data. It described the count as unavailable rather than zero.

1/1 assessed1/1 gradablePublished test setView Result →
The question uses a term the company defines itself
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash answered the revenue question with an explicit definition rather than a bare number. It also stated the period and time zone and distinguished totals before and after refunds.

1/1 assessed1/1 gradablePublished test setView Result →
The user's wording doesn't match how the values are stored
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash understood the gift-voucher request and returned all 14 matching orders. Their total was USD 16,806.08, the stated reference total, and two were cancelled.

1/1 assessed1/1 gradablePublished test setView Result →
The answer is available in the connected dataNo published result
Reporting3 pass
ScenarioPublished outcomesTest coverageResult
The report is run again after the data has changed
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash preserved the requested period when the saved report was refreshed after data changed. The order count rose from 23 to 26, the new reference count.

1/1 assessed1/1 gradablePublished test setView Result →
The user wants several findings in one place
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash passed this scenario. One chat shows three separate answers and a fourth report that repeats the same figures: August revenue $8,795.45, a bedding and rugs/home decor tie at 106 units each, Furniture at $51,248.25, and a 21.75% cancellation rate. Its verification query returned the same values.

1/1 assessed1/1 gradablePublished test setView Result →
The user wants to keep an answer they just got
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash saved the monthly-order answer as a dashboard chart and reproduced it in another account's session. The counts remained 11 for June 2026, 18 for July 2026 and 27 for August 2026, totalling 56.

1/1 assessed1/1 gradablePublished test setView Result →
Training4 pass
ScenarioPublished outcomesTest coverageResult
The agent used the wrong definition and the user corrects it
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash applied the revenue correction saved in its AI context to both the repeated question and a reworded one, each asked in a new chat. Both returned $23,469.02 across 42 non-cancelled orders, the total with cancelled orders left out, against $39,806.75 across 57 orders if they were counted.

1/1 assessed1/1 gradablePublished test setView Result →
The database's names are cryptic and the user documents them
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

After a one-line description of a cryptically named column was added to its AI context, Basedash answered that 7 returns are flagged for manual fraud review. Its query counted returns on exactly the documented flag, and 7 is the same count computed directly on the test database.

1/1 assessed1/1 gradablePublished test setView Result →
The user sets a standing rule for all future answers
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash respected the user's standing rule on both later questions without a reminder. It excluded cancelled orders from revenue and order-count calculations, and stated the exclusion in each answer.

1/1 assessed1/1 gradablePublished test setView Result →
The user supplies example questions and the queries they trust
1 Pass0 Fail0 Not gradable
1 of 1 test case passed
What happened

Basedash passed this training scenario: after a trusted question-and-query pair was saved as a verified model, it answered both the repeated question and a reworded version with 42. The untrained chat had answered 39, and the verified model preview also returned 42.

1/1 assessed1/1 gradablePublished test setView Result →

Reading these Results

Published evidence and test coverage answer different questions.

Publication availability

Which scenarios have a Result?

A published Result is public evidence for this tool on one scenario. “No published result” does not say whether testing has taken place.

Test coverage

What does each Result cover?

Assessed includes Pass, Fail and Not gradable. Gradable includes Pass and Fail. Both use the pinned test count in that published Result.

Scenario scope

Out of 28 scenarios

The summary counts each of the 28 scenarios once. A scenario fails when any of its published test cases failed. “No published Result yet” means nothing is public for that scenario; it does not say whether Basedash was tested on it.

Basedash in AI Database Agents | AI Demos