Integrate LLMS into any editor that supports LSP.
Read more about how & why I built sage.
Sage started off as a project to add features like LLM support to Helix, and enable features like workspace symbol search in large codebases I was working with that crippled language servers like pyright.
It's since grown into an AI multitool for any editor that supports language servers.
Dynamically includes the most relevant symbols in the context window, using a combination of tree-sitter and LSP clients. No need to manually @mention functions or types to include them in context!
Uses ollama to power on-device code completion and LLM integrations.
Supports using LLMs hosted on Cursor's cloud (via cursor-rpc.
Sage can index a repository (using llmcat) to power global symbol search for language servers that don't support it (like pylsp), or for repos that are too large.
It can index a repository with >10M symbols (~20M LOC) in a few minutes, and searches through even gigabytes of symbols in milliseconds:
