inferable

An open-source platform that streamlines the development of reliable, scalable, and secure AI agents and automations for developers.

Inferable

Inferable:

Hey today we want to share our new data connector project. Data connector allows you to chat with your APIs and databases using natural language. The data is returned in real time and doesn't require any ingestion pipelines. Under the hood it's executing SQL queries and issuing HTTP requests. Unlike some of the

Existing chat with your database tools. It doesn't require exposing the database connection to the internet. Let's have a look at using data connector for managing our local development environment. We have a Docker compos based environment here that has the data connector image included. We've set up two.

Connectors one pointing to our local Postgres instance and one pointing to the Clerk API. And Clerk is using this project for managing users and authentication. 

Let's jump into the chat interface and issue a prompt. We can issue a prompt that touches both of these connectors getting a specific cluster which is.

An entity within our database and then using the information about that cluster to work out the organization ID and fetch that organization's members from clerk so we can see that. To start with, the Run has issued a query to the postest database to get the cluster information using that cluster information.

It's gone ahead and issued two requests to the clerk API. One to get the organization to get its details and one to list the organization's memberships. We see there collated that in the final response. Data connector also supports a set of additional controls when running in production environments.

Approval mode requires that each query is approved by a user before it's executed. Privacy mode means that query results are returned to the user directly and not pass through the model. If we issue a query with both of these enabled, first the rum will pause before the query is executed and after the.

Query has been executed. We will receive the results but the model won't. So we can see that it's paused for approval. We can have a look at the query that it's suggesting we run and that looks good to me. So let's approve that. 

So the result was returned as ajacent object directly and the model doesn't have any.

Context of this data we also have the opportunity to download this. So he said the model also understands that the results been returned directly to the user. Thanks.

Note, that there may be slight differences in text and video. Watch video for the original content.

Inferable Video