Expand description
LLM Response Cache — avoid burning tokens on repeated prompts.
Stores LLM responses in a separate SQLite table keyed by a SHA-256 hash of
(model, system_prompt_hash, user_prompt). Entries expire after a
configurable TTL. The cache is optional and disabled by default — users
opt in via TANDEM_RESPONSE_CACHE_ENABLED=true.
Lives alongside memory.sqlite as response_cache.db so it can be
independently wiped without touching memory chunks.
Structs§
- Response
Cache - Response cache backed by a dedicated SQLite database.