Tyran

A Vector Search as a Service, Set up in Minutes.

Tyran is a service designed to efficiently retrieve relevant context for large language models (LLMs). Tyran organizes documents alongside their associated metadata, such as topic, author, and project or team, to enhance the search process.

When a query is made, Tyran follows the following:

• It uses the OpenAI embedding API to convert the query into a vector representation.
• The system then performs a semantic search using this vector, applying any specified filters to narrow the search to relevant documents.
• Upon identifying similar documents, Tyran fetches their content from the local database.
• Finally, it returns the relevant documents along with their similarity scores, providing a clear indication of how closely each document matches the query.

View the Project