ronantakizawa/cacheaugmentedgeneration

A Demo of Cache-Augmented Generation (CAG) in an LLM

31
/ 100
Emerging

Leverages Hugging Face Transformers' `DynamicCache` API to preload domain knowledge into Mistral-7B's context window, eliminating runtime document retrieval overhead. The implementation achieves 76% token reduction compared to traditional RAG by persisting cached context to disk for reuse across multiple chat sessions. Demonstrated through a Jupyter notebook that shows knowledge preloading, query answering against cached context, and cache serialization workflows.

123 stars. No commits in the last 6 months.

No License Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 10 / 25
Maturity 1 / 25
Community 18 / 25

How are scores calculated?

Stars

123

Forks

21

Language

Jupyter Notebook

License

Category

rag-applications

Last pushed

Jun 10, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/ronantakizawa/cacheaugmentedgeneration"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.