Skip to main content

Codebase intelligence

EditVolt understands your repo through a local semantic index — built, stored, and queried entirely on your machine.

How indexing works​

  • On first open, files are embedded with a bundled on-device embedding model — the model ships inside the app and never touches the network
  • Vectors are stored in an on-disk vector database, per workspace
  • Retrieval is hybrid: semantic similarity plus fast text search (ripgrep)
  • The index powers @Workspace answers, the agent's semantic_search tool, and code discovery

What stays out of the index​

Paths matched by your .editvoltignore file are never indexed and never enter AI context. Sensible defaults (.env*, private keys, SSH and cloud credentials) are on by default.

Project memory (optional)​

EditVolt can keep a small local memory of project facts it learns while working with you — stored in a local database inside the workspace data folder. It's off by default; enable it under Settings → Memory & Context.

Where the data lives​

Everything sits in the workspace's data folder (index, memory, chat history). Delete the folder and it's gone — nothing is synced anywhere.