OKF is designed to consolidate (compile) scattered knowledge into Markdown files that can be managed with Git and then maintained. The advantage is that instead of having to search for the information every time a question index.mdis asked, agents and people read the source code and translate it into OKF bundles, so you can just follow along from there.
Furthermore, OKF v0.1 is not a finished standard, but rather an initial specification that will evolve through future implementations and use cases. In the age of AI, one key issue will be organising internal documents and metadata in a way that is easily accessible to AI agents. OKF can be referenced as a candidate format for this purpose
In short, OKF is responsible for “knowledge that is written in advance,” while MCP and RAG are responsible for “mechanisms for retrieving and connecting data on the spot .” For example, the “definition of weekly active users” can be written in OKF, and the actual query to BigQuery can be done via MCP. RAG is not mutually exclusive, and it is possible to use OKF bundles as search targets in RAG as well.
No comments:
Post a Comment