With ChatAible you can query and interact with unstructured (e.g., document) as well as perform augmented analytics on structured (tabular) data or indeed query the LLM directly.
When you first log into ChatAible you land on the New Chat screen:
ChatAible works with unstructured data in the form of Document Sets - which are simply containers with one or more text documents.
To create your first document set, navigate to the Datasets page and click Create a Document Set:
In the dialog give your Document Set a name related to the document(s) you are planning to use.
For example, if you wish to query an employee handbook you could simply call the document set. “The Employee Handbook”. Or, if you’re planning to upload a set of contractual documents relating to your engagement with suppliers , you could call the document set “Supplier contracts”.
The (empty) document set shows up as a new dataset card - here it’s called Privacy Policy (Notice that you can filter the datasets to show only structured or only unstructured datasets):
Now click Upload Document to add documents to your set. ChatAible currently supports PDF document but other types will be added over time. It takes a little while to load the documents as they are split into chunks, converted to embeddings, stored and indexed ready for querying.
Once it’s ready (the spinner has stopped) you can jump into Ask ChatAible and query the document.
At a high level, querying the document works by searching all of the document(s) for a small number of chunks that best match the question then sending the question to the LLM with this small subset of text snippets along with the instruction to use only this text as the source for its response.