Expand description
Per-row inference result cache, keyed (content_hash, model_id, params_version).
The composite key ensures different models and parameter sets never collide on
the same input. Backed by quick_cache::sync::Cache with S3-FIFO eviction.
Structs§
- AiCache
Key - Cache key. All fields are
Copy; lookups need no allocation. - AiResult
Cache quick_cache-backed in-memory cache of per-row inference results.- AiResult
Cache Config - Configuration for
AiResultCache.
Enums§
- Cached
Output - One row’s cached inference output.
Functions§
- content_
hash - xxh3-128 of the input content.
- params_
version - Hash of the parameters that affect model output (currently the label set).