It'd be very cool to interface with Surfer directly in Claude. Because everything is local, the filesystem server can be used for Surfer. However, this is inefficient because Surfer files return large JSON's that are too long for Claude's context window. A local vector database could work for this, as a hypothetical Surfer server could expose a function to vector search over the data from Surfer and can even add in reranking and more advanced semantic search functionality to improve the quality of responses.
It'd be very cool to interface with Surfer directly in Claude. Because everything is local, the filesystem server can be used for Surfer. However, this is inefficient because Surfer files return large JSON's that are too long for Claude's context window. A local vector database could work for this, as a hypothetical Surfer server could expose a function to vector search over the data from Surfer and can even add in reranking and more advanced semantic search functionality to improve the quality of responses.