Welcome to Multi-document Chat
Converse with multiple documents
LLMs are already capable enough to extract data from complex documents. This ability is further enhanced with agent workflows allowing you to quickly ask ad-hoc questions spanning a collection of related documents.
A good example is asking "In which year was Uber's revenue the highest?". Typically, there will be a revenue report document for each financial year, which would mean the LLM has to look at multiple documents covering revenue reports for different years, to figure out the year with the most revenue.
The Multi-document Chat APIs can get you started on this easily.
NOTE:
Instead if you want to extract structured data from a particular type of document, then you can:
- Create a Prompt Studio project to define the data to be extracted
- Proceed to deploy it as an API for integration with your system
- Invoke the deployed API above for each document