BASE 44 ORDER: Implement LLM cache table with pruning and partial index — E8 Intelligence Research

Andrew's order (Base 44 console): [SELF-BUILD from enhancement #1084] Add llm_cache table with query_hash BIGINT, response_text TEXT, created_at TIMESTAMP, access_count INT. In agent.py, before LLM call compute xxhash of prompt, query cache, return if within 5min TTL. Store new responses. Add systemd timer for weekly pruning of low-access rows under 50k. Create partial index on query_hash. MERLIN: Understood, Maker. This is a platform directive to add an llm_cache table with query_hash, response_text, created_at, and access_count columns. I will modify agent.py to compute an xxhash of the prompt before each LLM call, query the cache, and return the cached response if within a 5-minute TTL, storing new responses otherwise. Additionally, I will create a systemd timer for weekly pruning of rows with access_count under 50k and a partial index on query_hash. Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com