The local first approach is definitely the right move for project memory. How do you handle context window limits when a project memory stats getting large?

Interesting approach! I've been working on something complementary — local-first LLM observability called opensmith. Different angle but same philosophy of keeping everything local.