About the Pryon Retrieval Engine
The Pryon Retrieval Engine securely provides accurate, instant, and verifiable answers, understands natural language queries, extracts matched content, provides precise attribution, and maintains access level controls.
The retrieval APIs power enterprise RAG applications as retrievers, query engines, and conversational or chat interfaces.
Retrieval API categories
There are three major categories of Retrieval APIs: Exchange, Content, and Feedback.
Content
URI: /api/knowledge/v1/content/
A content object is a single content source (e.g., a PDF document, a Google Doc, or the HTML file associated with a single web page) and is an element of a content group (such as a Sharepoint folder). A set of content groups make up a knowledge domain (or index) which can be ingested into the platform and queried against.
Exchange
URI: /api/conversation/v1/exchange
The Exchange API allows you to make natural language inquiries about your ingested information, retrieving the top matched answers and surrounding context to power RAG applications.
Responses to the inquiries may include:
- Short answers
- Surrounding context chunks
- Links for attribution
- Bounding boxes for answer highlighting
- Content metadata
For conversational use cases, the Exchange API initiates and continues a conversation. Each call performs one synchronous round-trip by sending a text request to the service and receiving the text response, which is then stored to keep a record of each verbal/textual interaction.
Feedback
URI: /api/conversation/v1/ratings
The Feedback API is used to provide and manage user feedback for a given exchange and response answer, rating the accuracy of the question/answer pair as a basis for becoming a verified answer.