llm-cacherCache LLM responses

Exact and semantic matching for OpenAI, Anthropic, and more. Cut costs, reduce latency.

Get Started

API Reference

View on GitHub

⚡️

Exact match

SHA-256 hashes every request. Identical calls are served instantly from cache with zero API calls.

🧠

Semantic match

Embed prompts and find similar cached responses using cosine similarity or HNSW. "What is 2+2?" and "What does 2 plus 2 equal?" share the same cache entry.

🗄️

Five storage backends

Memory, file, Redis, SQLite, DynamoDB — pick what fits your infrastructure. All are optional peer dependencies.

🔌

Framework integrations

Drop-in middleware for Express and Hono. NestJS module with dependency injection. Works out of the box.

📦

Transparent proxy

Wrap your existing client with one line of code. No interface changes, full TypeScript types preserved.

🌊

Streaming support

Accumulates stream chunks, stores them, and replays as AsyncGenerator. Your streaming code doesn't change.