CCAP-Kernel (v0.2.0) is an AI-native semantic compiler that compresses large codebases into high-entropy semantic maps using Spectral Graph Theory and Minimum Description Length (MDL), achieving over 95% token savings. It is developed by Reack and released under MIT license. Key differentiators include multi-model token budgeting for OpenAI/Claude/Gemini, path-agnostic linking for cross-platform isomorphism, ghost link detection for architectural debt, and calibrated auditing. The project is in early alpha stage with rapid development (moving from v0.1.x to v0.2.0 within months). It targets AI-assisted software engineering workflows by providing a progressive precision workflow from global navigation to surgical patching.
npm install ccap-kernelNo compatibility data collected yet for this library.
Verified import paths — ran on the pinned version, not inferred.
Shows how to analyze a codebase into a semantic map and apply a surgical patch, plus simulate token budgeting for a specific AI model.
Re-run analysis with v0.2.0 to regenerate maps.
Use the default path-agnostic linker; remove `--legacy` flag.
Ensure Node.js version is 18 or higher.
Use one of: 'gpt-4o', 'claude-3.5', 'gemini-1.5'.
Replace `--verbose` with `--log-level debug`.
Initialize a git repository or use `--no-ghost-link` to suppress.
Add `--format json` to commands expecting JSON output, or update tools to handle binary format.
Use `npx ccap-kernel` or install globally with `npm i -g ccap-kernel` and run as a command.
Upgrade Node.js to version 18 or later.
Create a ccap-config.json file with appropriate settings, or run `npx ccap-kernel init` to generate one.
Use one of the supported model names: gpt-4o, claude-3.5, gemini-1.5.
No dependency data recorded yet.