Back to home
Shell compression
11 strategies. 645+ classifiers. 93% average compression.
Raw shell output is the largest hidden token bottleneck for AI coding agents. unerr classifies every command, routes it to a strategy tuned for that output shape, and delivers the compressed result to the agent. The raw bytes stay on disk — the agent can recover them on demand.
93%avg overall
11strategies
645+classifiers
| Strategy | Targets | Compression |
|---|---|---|
| diff | git diff, patch output | 99% |
| structured | JSON APIs, docker inspect | 97% |
| progress | npm install, pip install | 95% |
| log_text | build logs, server logs, make, cargo build | 89% |
| test_results | vitest, pytest, cargo test, playwright | 80% |
| tabular | ps aux, docker ps, kubectl get | 77% |
| error_diagnostic | tsc, eslint, rustc, shellcheck | 72% |
| key_value | env, kubectl describe, systemctl status | 48% |
| tree_paths | find, tree, ls -R | 42% |
| yaml | YAML configs, kubectl get -o yaml, Helm output | adaptive |
| omni | Fallback for unrecognized output | adaptive |
Benchmarked across 40 real-world test cases: 2 MB → 138 KB. Raw output is preserved on disk; the agent can recover it on demand via file_read against the captured tee.