Show HN: Give (low vRAM) local sLLMs codebase-wide context with SQLite (attempt) (github.com)

A new AI coding tool, arcKode, has been introduced that enhances local software development by indexing entire codebases while operating within low VRAM constraints. Instead of treating projects as flat text, arcKode employs Tree-sitter to create a dynamic symbol graph, which facilitates targeted code understanding. This ensures that the AI model only sees the relevant portions of the code, significantly reducing errors such as misinterpreted file paths. The tool runs entirely offline, preserving data privacy as no information leaves the user's machine. The significance of arcKode lies in its innovative semantic understanding and technical capabilities, allowing it to process code in 11 programming languages while maintaining a compact context through token-aware packing strategies. It features dual-editing modes for efficient changes and includes stringent safety measures, like automatic snapshots and incremental indexing for modified files. By integrating with Ollama-compatible models, arcKode offers a completely self-sufficient coding assistant that can streamline both simple edits and complex refactors, making it a valuable addition for developers keen on improving their workflows within a secure, local environment.