how does knowledge sequestration work?

Knowledge Sequestration is EcoRouter’s approach to reducing unnecessary computation by preserving and reusing useful information.

It works alongside our routing system. While routing ensures that when we do need to generate an answer, we use the lightest model capable of the task, Knowledge Sequestration aims to avoid the need for new generation entirely when a high-quality answer already exists.

Currently, this is handled through Knowledge Reuse V1:

• Matching: Before generating an answer, EcoRouter checks our Public Knowledge database. It looks for strong matches using exact and conservative lexical matching.
• Limitations: This system does not use semantic matching or embeddings, and it is not applied to questions involving volatile or current information that requires up-to-the-minute data.
• Choice: If a strong match is found, the system recommends it to you. You can then choose to use that existing knowledge or proceed with generating a new answer.

By building a shared library of reliable question-and-answer pairs, we can reduce the frequency with which the network needs to perform new computations. It is part of our broader goal to help AI become more efficient and useful as it grows.