scry#
Self-hosted semantic code search. One server holds a hybrid dense + BM25 index of your repos; any device with a checkout of the same remote searches and syncs the same index. Embeddings come from any OpenAI-compatible endpoint, so everything can run on your own hardware.
scry "how are chunks defined" # semantic search in the current repo
scry "auth middleware" src/server # scoped to a subdirectory
scry --web --answer "sqlite-vec quantization options"
scry watch # keep the index in sync while you work
scry serve # run the index server
built with cargo soothfast docs build
source