Module semantic_cache 

Semantic Cache

LRU-evicting similarity cache for prompt/response pairs.

Prompts are embedded into a 64-dimensional TF-IDF-style vector using FNV-1a token hashing, then L2-normalised. Cache lookup returns a stored response if the cosine similarity between the query embedding and any stored embedding is at or above the configured threshold.

Structs

CacheEntry

A single entry stored in the SemanticCache.

CacheStats

Aggregate statistics for a SemanticCache.

SemanticCache

LRU-evicting semantic similarity cache.

Functions

cosine_similarity

Cosine similarity between two equal-length vectors.

embed_prompt

Produce a deterministic TF-IDF-style 64-dim embedding for prompt.