Looking for answers to FAQs and links to helpful resources? Get started here.
On this page:
- How can I get started with AutoQL?
- How much does AutoQL cost? How does pricing work?
- Do you store a copy of our database or retain any of the data in it?
- Are there any specific software architecture requirements that must be met before integration is possible?
- Can you support an instance where the customer prefers to host on their own premise/security environment?
- Is it possible to query across different data sources?
- Do queries ever fail to return data? What happens if a query doesn't work?
- How and when is data encrypted?
- Are there user permissions or role access permissions in place to control who is accessing the data?
- Can the SQL that is generated through user queries join multiple tables in the database?
- How does AutoQL handle similar column, table, or item labels in the database?
- Can you copy a session ID with log entries so that it shows the chain of queries and SQL results that were generated?
- Do you provide a metadata package that contains items like the SQL, which columns and tables are used, etc.?
How can I get started with AutoQL?
If you're ready to get started with AutoQL, reach out to us at [email protected] or visit our website to get in touch. We'll work diligently with you and your team throughout each stage of your AutoQL integration process: from customizing our AI for your unique database to planning your feature release and providing ongoing strategic support, so you can maximize your investment and realize value, fast. We look forward to working with you!
How much does AutoQL cost? How does pricing work?
AutoQL is priced on a per API call model. Book a call with our team to learn more about how this works and what options are available for your business.
Do you store a copy of our database or retain any of the data in it?
One question that we often encounter is, “Does Chata.ai store my data?” The answer is: we store only the minimal data required for AutoQL to run securely and efficiently in your software application. We refer to this as nominal data, or the unique data we identify in your database that users may want to ask questions about. This enables us to ensure that queries run faster and return the exact results the user is expecting.
An example of a question a user may wish to ask is “How many of each item did Jane Doe buy last month?” In the context of this user's query, we store the information that “Jane Doe” exists in the user’s database. In other words, we store "Jane Doe" (the nominal data) in Chata.ai’s cloud environment so an optimal database query can be generated dynamically.
We also store user queries, allowing us to provide the Integrator with insight into what their users are asking.
Are there any specific software architecture requirements that must be met before integration?
AutoQL is an extremely flexible, API-first solution that can be customized for any software that uses a relational database. It works exceptionally well with cloud-native software solutions that have a web-based interface.
Independent Software Vendors (ISVs) that perform continuous deployments can readily take advantage of our end-to-end system that includes flexible front end widgets. Each widget takes just 2-3 hours to implement.
We can also integrate with on-premise software systems.
Can you support an instance where the customer prefers to host on their own premise / security environment?
We’re not currently offering this option, but we’re building towards supporting this type of integration opportunity in the future.
Is it possible to query across different data sources?
At this time, users can only query one data source. AutoQL can be integrated with a data warehouse or a data lake that brings data from multiple sources together.
Do queries ever fail to return data? What happens if a query doesn’t work?
AutoQL’s expected query success rate is 87-95%. This means that almost all of the time, there’s a data will be returned in response to the query that the user asks.
There are a few reasons why data may not be successfully returned from natural language queries:
- Our system is designed to ensure that the most correct data response is returned. AutoQL will not return data if it’s not confident that an optimal SQL statement has been generated.
- AutoQL has built-in models that act as "safety nets" or "check points" when a natural language query is received that is not immediately understood. In these cases, the auxiliary models kick in to try to determine exactly what data the user is asking for. If this verification step is unsuccessful (whether that be because the meaning of the user's query is still unclear, or because the data the user wants is not actually present in the database), AutoQL will not return a data response.
How and when is the data encrypted?
Data is encrypted at rest and in flight and we use non-numeric IDs to ensure that data is secure at every level. We use JWT to secure who has access to our system, and for how long.
Are there user permissions or role access permissions in place to control who is accessing the data?
Integrators can set permission levels for users within our Integrator Portal. If you have in-house permissions structures, we’ll work with you to build a permissions framework that works for you.
Can the SQL that is generated through user queries join multiple tables in the database?
Yes, AutoQL automatically generates table-joining SQL when a user query requires it.
The system is built to dynamically translate natural language into accurate SQL statements that can include multi-table joins, filters, and aggregates as user queries require them.
How does AutoQL handle similar column, table, or item labels in the database?
For example, “invoice” might refer mean different things for users in the accounting department vs. users in the warehouse.
A key feature of AutoQL’s technology is the ability to disambiguate unique labels. This means the system is trained to determine what a user is referring to, even when that meaning is not explicit.
Unique labels are taken into account during the integration process and we have built in a machine learning model that disambiguates the value label based on the context of the question.
Can you copy a session ID with log entries so that it shows the chain of queries and SQL results that were generated?
Yes. We enable this functionality through the API and the Integrator Portal. In the Integrator Portal, you can see the natural language query, the user ID, and the generated SQL, as well as the success status of the query. Through the API, you can make a call to access this same information.
See how in the API documentation here.
Do you provide a metadata package that contains items like the SQL, which columns and tables are used, etc.?
Yes, there are several options for metadata return:
- We offer a model that returns only the SQL when users enter a question in natural language, so that Integrators can use the SQL however they need (e.g. customizing how the user receives their data response).
- We also return metadata through the standard query flow: Integrators can log the SQL response for any query alongside the data visualizations that were returned for that query.
Links to Resources
Updated 3 months ago