CodeNib speeds coding agents 8.7x with multi-view repo context
New system cuts tokens by 87% and latency by 4.7x for AI coders.
Deep Dive
CodeNib is a multi-view data system that serves repository context to coding agents. It maintains lexical, dense, and structural views per commit, delivering ranked search, symbol navigation, and bounded context through one runtime. Across 100 snapshots, graph and vector updates were 8.7× and 25.4× faster (median) than independent rebuilds. Selected context policies preserve localization with 50–87% fewer trajectory tokens than paired grep/read.
Key Points
- Graph and vector updates were 8.7× and 25.4× faster at the median than independent rebuilds across 100 snapshots.
- Selected context policies reduced trajectory tokens by 50–87% while preserving localization accuracy across five AI models.
- Static navigation latency was 4.7× faster than live servers for 63% of 1,000 requests.
Why It Matters
Cuts token costs and latency for AI coding agents, enabling faster, cheaper repository-level code understanding.