PN — An optional inference cache stores (model, prompt_hash, temperature, seed)…
An optional inference cache stores (model, prompt_hash, temperature, seed) → response in a local SQLite file; on a cache hit the Ollama call is skipped entirely, keeping repeated runs deterministic when the user opts in.
Source: PUMA project documentation · Traceability: corpus unit
PUMA-009· Confidence: repo-backed