bash output compression: 12 per-tool filters, failures-first, ~80-97% smaller#3
Merged
Merged
Conversation
… smaller PreToolUse hook on Bash detects compressible commands and rewrites them to flow through `token-goat compress`, which runs the original through the system shell, captures stdout + stderr, applies the per-tool filter, and emits a failures-first compressed view that strips progress bars, dedupes warnings, groups linter issues by rule, and keeps every error block verbatim. Built-in filters cover pytest, jest/vitest, cargo, npm/pnpm/yarn, docker/buildah/podman, kubectl/helm, aws, ruff/eslint/mypy/pyright/tsc, git, make/ninja/gradle/mvn/go, terraform, and pip. Per-stream capture is capped at 32 MiB, total output at 1000 lines / 64 KiB; the wrapper preserves the original exit code and kills the entire process group on timeout. Configurable via `[bash_compress]` in config.toml or disabled entirely with `TOKEN_GOAT_BASH_COMPRESS=0`. Also fixes three pre-existing bugs: - paths.open_log_file claimed to return FileHandler but returned a bare StreamHandler on POSIX, breaking the worker isinstance check. - Two test_canonicalize_* tests asserted Windows-shell normalisation invariants that only hold on Windows; now skipped on POSIX.
…ashes Windows CI was failing on os.killpg / os.getpgid / signal.SIGKILL attribute checks because mypy treats them as POSIX-only. Extracted the kill body to _posix_kill_tree which looks the attributes up with getattr so the symbol lookups don't fire on Windows. Also went through new files and replaced inline em-dashes with regular punctuation, and tightened a few of the more marketing-flavored phrases in the changelog and docstrings.
Zelys-DFKH
added a commit
that referenced
this pull request
May 25, 2026
…xecutor - #2: _is_git_repo cheap stat check (accepts .git dir OR worktree .git file) before each call to _get_uncommitted_changes / _get_git_diff_stat_summary. Process-local cache. Saves ~60-100 ms per non-git-cwd manifest build; zero cost on git repos. - #3: ThreadPoolExecutor removed — two git ops now sequential. Saves ~3-8 ms of executor overhead per manifest. Existing per-call TTL caches still warm on second invocation. New TestIsGitRepo (4) + TestNonGitShortCircuit (4) tests; existing git-stat tests patched with _is_git_repo mocks.
Zelys-DFKH
added a commit
that referenced
this pull request
May 25, 2026
…ts/cold sections - Item #3: _format_grep_entry emits bare (N) instead of (N results); ~1 token saved per grep entry, semantics unchanged. - Item #4: _format_glob_entry uses the ASCII g: marker instead of the multibyte 📂 emoji and drops the scope path when it equals the session cwd. - Item #5: _get_session_commits lines no longer prepend - ; the section is already inside a bulleted block, so the prefix was pure overhead. - Item #10: small (<1KB) untruncated bash outputs render (e=N) without the byte count; truncated/large outputs keep the full metadata. - Item #11: the Cold Outputs header collapses from a full H3 + parenthesised recall hint into the bold-label **Cold:** evict, recall via … form.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PreToolUse hook on Bash detects compressible commands and rewrites them to
flow through
token-goat compress, which runs the original through thesystem shell, captures stdout + stderr, applies the per-tool filter, and
emits a failures-first compressed view that strips progress bars, dedupes
warnings, groups linter issues by rule, and keeps every error block
verbatim. Built-in filters cover pytest, jest/vitest, cargo, npm/pnpm/yarn,
docker/buildah/podman, kubectl/helm, aws, ruff/eslint/mypy/pyright/tsc,
git, make/ninja/gradle/mvn/go, terraform, and pip. Per-stream capture is
capped at 32 MiB, total output at 1000 lines / 64 KiB; the wrapper
preserves the original exit code and kills the entire process group on
timeout. Configurable via
[bash_compress]in config.toml or disabledentirely with
TOKEN_GOAT_BASH_COMPRESS=0.Also fixes three pre-existing bugs:
StreamHandler on POSIX, breaking the worker isinstance check.
invariants that only hold on Windows; now skipped on POSIX.