GitHub radar
Amber — a portable verified vector store for offline RAG
A Python library for producing a single, portable file containing all the information needed to verify the integrity of an embedding corpus via cryptographic means.
Embedding corpora can be large and time consuming to create, but for many applications they are embedded repeatedly. This python library allows you to serialize the embeddings into a file (.amber) that contains cryptographic proofs of the integrity of the embedding. This means that users can verify your embedding via spot checks (i.e., re-embedding some portion of the corpus and checking that the results match). This also means that you can distribute your embedding file without the need to share your model. The repository contains formal proofs of correctness.
Why a vibe-coder should care
If you’re sick of having to re-embed your corpus for your RAG applications, then give Amber a try!
▌ More finds