ContextZip started as a fork of RTK (Reduce Toolkit). Three weeks later, it was a different tool with 1,056 tests and coverage for 102 CLI command patterns.
Here's what building it looked like.
Week 1: The Core Filters
The first week was ANSI stripping and duplicate detection. ANSI codes seem simple until you hit edge cases — partial sequences, nested attributes, malformed escapes from buggy tools. 247 tests just for ANSI handling.
Duplicate detection was harder. "Duplicate" doesn't mean "identical." These lines are duplicates in meaning:
npm warn deprecated inflight@1.0.6: This module is not supported...
npm warn deprecated glob@7.2.3: Glob versions prior to v9...
Both are "npm deprecated" warnings. The package names differ but the pattern is the same. ContextZip groups by pattern, not exact match.
Week 2: Stack Trace Intelligence
Each language has different stack trace formats. Node.js uses at Function (file:line:col). Python uses File "path", line N, in function. Rust uses numbered frames with at path:line. Go dumps entire goroutine stacks.
For each language, ContextZip needs to know which frames are "framework" and which are "yours." The heuristic: paths containing node_modules, standard library paths, runtime internals → framework. Everything else → application code.
412 tests for stack trace filtering across 6 languages.
Week 3: Command-Specific Patterns
Package managers, build tools, Docker, git — each has its own noise patterns. Progress bars in pip, layer hashes in Docker, tree-shaking stats in webpack. 102 command patterns tested individually.
The Result
1,056 tests. Zero false positives in the benchmark suite (no useful information stripped). 60-90% average noise reduction across all tested commands. Written in Rust for zero overhead.
cargo install contextzip
eval "$(contextzip init)"
GitHub: github.com/contextzip/contextzip
Part of the ContextZip Daily series. Follow for daily tips on optimizing your AI coding workflow.
Install: npx contextzip | GitHub: jee599/contextzip
Top comments (0)