git apply --3way --check exits 0 even when the real apply would write conflict markers and unmerged index stages, so the previous fix left the worktree dirty on conflicting patches while only flipping changesApplied to false.
Dropped --3way for patch-mode merge and used a --reverse --check probe instead: it succeeds only when the target state is already present (true no-op) and reads without touching the worktree. Conflicts fall through to the normal --check + apply path, which rejects them before writing anything.
Added regression coverage for the conflict scenario asserting the worktree stays clean, and for the fresh apply path.
- Removed the canonical model variant indexing, selection, and tracking logic from the model registry and resolver.
- Eliminated the `canonical` sub-command, tab view, search tokens, and equivalence configuration structures from the CLI and model selector components.
- Refined model identification, lookup, and provider fallback resolution to bind exclusively to standard, raw model IDs.
- Relocated the equivalence utility script within the catalog package to support script-only policy generation.
- Added a helper to build the Git process environment that explicitly clears common ambient Git environment variables.
- Applied the new environment builder to both async and synchronous Git commands to prevent environment bleeding from parent processes.
- Introduced `isProbablyBinary` utility to sniff file headers for NUL bytes or invalid UTF-8 sequences.
- Updated `ReadTool` to use the binary sniffer, preventing mojibake corruption in output when reading non-text files.
- Refined `file-mentions` auto-reads to skip binary files and mark them as `binary` in the message transcript.
- Added comprehensive unit tests for binary detection logic, covering NUL bytes, truncated multibyte characters, and path-based file sniffing.
Dirty isolated baselines can be accidentally committed by subagents that run git add -A. Fetching the raw isolation HEAD then cherry-picking the range would replay that baseline WIP into parent history.
Add a dirty-baseline replay path that rewrites each agent commit against the captured baseline tree, preserving the agent commit message and author while excluding staged, unstaged, and untracked changes that existed before isolation started. Clean baselines still use the raw git fetch path, and nested-only changes keep returning patches without creating an empty root branch.
Add a regression for baseline staged + untracked WIP committed by the agent, asserting the task branch contains only the agent file and parent WIP remains staged/untracked after merge.
Fixes#3842
Applied isolated branch patches with three-way fallback when unrelated parent dirt appears in patch context.
Surfaced branch preparation failures instead of reporting no changes.
Fixes#3841
- Added `git.repo.linkedWorktreeSync` to identify and resolve git worktree metadata without spawning subprocesses.
- Updated `StatusLineComponent` to detect linked worktrees and resolve project/worktree context names.
- Modified path segment rendering to collapse nested git worktree paths and display the worktree name when it diverges from the active branch.
- Introduced `icon.worktree` symbol across themes to visually distinguish git worktree paths.
Tracked received thinking content per interactive session so OpenAI-compatible providers that omit reasoning metadata can still reveal streamed reasoning blocks. Added a Ctrl+T regression covering the unlocked visibility path.
Fixes#3669
- Added `tiny` as a first-class model role to override online models for lightweight background tasks.
- Updated session title generation, auto-thinking difficulty classification, unexpected-stop detection, and mnemopi backend to resolve via the `tiny` role before falling back to `smol`.
- Updated configuration schema and documentation to reflect the new role precedence.
- Updated `normalizeGeneratedTitle` to reconcile model-generated titles against the user's input instead of forcing title-case.
- Added logic to restore distinctive proper-noun casing (e.g., `TinyVMM`) and flatten model-generated camelCase artifacts (e.g., `dAemon`) that do not appear in the user's message.
- Ensured model-cased proper nouns that are not in the source message (e.g., `GitHub`) are preserved.
- Added a fallback from `hashline` to `replace` mode for Kimi-family models to resolve compatibility issues.
- Introduced `PI_STRICT_EDIT_MODE` environment variable to bypass automatic model-specific edit-mode fallbacks.
- Updated `getEditVariantForModel` to perform case-insensitive matching for model variant configurations.
- Added comprehensive unit tests for edit mode resolution and settings configuration.
Reviewer caught (correctly) that the #3506 text fallback relied on
`clipboard.readText()`, which on Darwin shells out to `pbpaste(1)` —
pbpaste only surfaces plain text / RTF / EPS, so a Finder Cmd+C
pasteboard (`public.file-url` only, no plain text, no raw image bytes)
made readText() return empty and the new path-detection never ran.
Add a Darwin-only `readMacFileUrlsFromClipboard` helper that pipes a
small AppleScript through `osascript -` to coerce the pasteboard via
`«class furl»` and emit POSIX paths one per line. Wire it into
`InputController.handleImagePaste` between the readImage and readText
calls; the first image-shaped path routes through
`handleImagePathPaste`, non-image file URLs (e.g. a copied .pdf) fall
through to the existing text fallback. The clipboard interface field is
optional so existing test fixtures keep working without changes.
Tests: covers (a) Darwin file-URL pasteboard with empty pbpaste,
(b) non-image file URLs falling through to text, (c) the helper itself
on darwin/non-darwin and when osascript fails.
Refs #3506
PR #3474 second review: previous revision ran `umask 077` only BEFORE
sourcing the rc, so a typical `.bashrc`/`.zshrc` that calls
`umask 022` reopened the world-read window between the spawned shell's
first `>|` and the JS post-spawn chmod. Snapshot file (with inlined
env-var values) lived at 0644 for the full body of the script.
Two-layer fix:
- JS caller now pre-creates the snapshot file at 0600 with
`fs.writeFileSync(path, "", { mode: 0o600 })` before spawning. The
shell's `>|` (truncate) and `>>` (append) preserve the existing
inode mode, so the file is 0600 from byte zero regardless of the
spawned shell's umask state.
- Script also re-applies `umask 077` after the rc source so any
other file the script might create (none today, defensive) stays
private even when the rc resets umask.
New e2e regression test seeds a `.bashrc` containing `umask 022` and
asserts the resulting snapshot mode `& 0o077 === 0`.
PR #3474 review: the new export pass writes referenced env-var values
into a snapshot file under `os.tmpdir()/omp-shell-snapshots`. On Linux
where `os.tmpdir()` is `/tmp` and the umask is the default 022, the
file ended up world-readable (0644) until postmortem cleanup. A user
rcfile defining `deploy(){ curl -H "Authorization: $GITHUB_TOKEN" ...; }`
would have its token written verbatim to that file.
Three-layer mitigation:
- `umask 077` at the top of the snapshot script so the file is 0600
from the first byte (the shell creates it via redirection, not JS).
- JS caller now passes `mode: 0o700` to `mkdirSync` and chmods the
dir + file defensively after the script exits, covering pre-existing
dirs and exotic shells where the umask call might not take.
- Helper denylist gained the common secret-shaped name patterns
(`*TOKEN*`, `*SECRET*`, `*API_KEY*`, `*PASSWORD*`, `*PASSWD*`,
`*PRIVATE_KEY*`, `*ACCESS_KEY*`, `*CREDENTIAL*`, `*SESSION_KEY*`)
so even when the file is locked down, we don't materialise tokens
onto disk in the first place.
Tests cover both: a new helper-level test asserts none of the secret
names (or their values) appear in the export stream, and the e2e test
now stats the snapshot file + dir and asserts `mode & 0o077 === 0`.
generateSnapshotScript captured the user's shell functions via declare -f /
typeset -f and dropped everything except PATH on the export floor. mise
activate installs a mise() function whose body expands $__MISE_EXE; the
replay shell then ran `command "" "$@"` and died with
`command: command not found:` (exit 127). The same shape breaks asdf
shims, direnv-style helpers, and any other activation idiom that pairs a
shell function with a sidecar env var.
The snapshot script now scans captured function bodies for $VAR /
${VAR…} references and re-emits `export NAME='value'` for each name
that is currently set and not on a shell-internal denylist (PATH, HOME,
BASH_*, LC_*, …). getShellConfigFile also honours env.HOME so callers
(and tests) can target a sandboxed home — os.homedir() is cached by Bun
and ignores later process.env.HOME mutations.
Fixes#3470
Stream fetched tool assets to disk under the existing download abort signal instead of passing the Response object to Bun.write. Remove partial files when a stalled body is aborted and cover completed plus stalled downloads with regression tests.
Fixes#3369
Snapshot and restore PI_CODING_AGENT_DIR, OMP_PROFILE, PI_PROFILE, and
XDG_CACHE_HOME instead of relying on setAgentDir(originalAgentDir), which
cannot restore previously-unset or profile-derived env state. Reset the
profile snapshot and rebuild dirs from env in cleanup to prevent
suite-order pollution.
Also reject empty cached content in parseCacheEntry() to harden the
cache contract against corrupted/empty entries.
The .txt extension match was winning over the basename check in
getLanguageFromPath, returning 'text' instead of 'cmake' for
CMakeLists.txt. Reordered the function to check basename special-cases
first (mirroring detectLanguageId's structure), so CMakeLists.txt,
Dockerfile, .env., .emacs, and justfile all fire before the extension
lookup.
Rewrote tests to defend observable contracts (special filenames, case
handling, lookup ordering) instead of re-stating the EXTENSION_LANG
table. Added contract test pinning that CMakeLists.txt resolves to
'cmake' not 'text'.
Updated CHANGELOG: moved from ### Added (test-only) to ### Fixed
(bug fix).
- Add random UUID suffix to cache temp filenames to avoid same-pid/same-ms collisions
- Export pruneMarkitConversionCache and cover orphaned .tmp sweeping with a regression test
- fold coding-agent package version into the cache key so releases that
change markit converter output auto-invalidate stale entries
- sweep orphaned `.tmp` files during prune (crash between write and
rename previously leaked, invisible to the size cap)
- make prune fire-and-forget after rename so a cache miss returns once
the entry is on disk instead of waiting on a readdir + N×stat sweep
- document the FIFO-by-mtime eviction policy on-record
- use Bun.file()/Bun.write() for payload I/O per repo conventions
Repeated reads of unchanged PDFs, Office documents, and EPUBs re-ran the
full markit conversion every time. Add a transparent, content-addressed
cache for successful conversions keyed by SHA-256(content) + normalized
extension, so repeat reads reuse converted markdown instead of
reconverting.
- packages/utils: XDG-aware getDocumentConversionCacheDir() helper
- coding-agent: markit-cache module (bounded 256 MiB, oldest-first prune,
best-effort writes that never fail conversion) layered over the central
convertFileWithMarkit/convertBufferWithMarkit wrappers
- imageDir conversions stay uncached (cache:"skipped") to preserve PDF
image extraction side effects; failed/empty/aborted conversions are
never cached
- abort-safe: file byte reads run under untilAborted; cache I/O rechecks
the signal
- Transitioned the eval tool from batch multi-cell execution to a single-step input structure with flat parameters.
- Updated core agent logic, UI components, and documentation to support state persistence across incremental eval calls.
- Restricted bash tool capabilities by requiring explicit use of `read` or `find` instead of `ls` or `find`.
- Added support for Ruby and Julia language runtimes to the eval tool and associated web renderers.
- Added `minDimension` option to ensure images meet minimum size requirements for vision backends.
- Implemented logic to scale up undersized input images while respecting maximum constraints.
- Clamped minimum dimension floor to avoid resolution conflicts with defined maximum bounds.
brush-core's alias expander resolves aliases via
`value.split_ascii_whitespace()` (`crates/brush-core-vendored/src/interp.rs:1500`,
upstream brush issue reubeno/brush#57): each whitespace piece is dropped
into argv as-is, completely bypassing the shell parser. Any alias body
containing `(`, `)`, `|`, `&`, `;`, `<`, `>`, or `\`` therefore
turns the first piece into the command name, so Fedora's default
`alias which='(alias; declare -f) | /usr/bin/which …'` produces
`error: command not found: (alias;` for every `which` invocation.
The user's shell snapshot is generated by sourcing their real rc-file
under `/bin/bash` or `/bin/zsh` (so we can capture functions, options,
PATH) and then sourced by brush per-session. `sanitizeSnapshotForBrush`
now scans the emitted `alias -- NAME='VALUE'` lines after generation,
drops any whose decoded body contains those metacharacters, and rewrites
the file in place before caching. Compatible aliases (`ll='ls -l'`,
`gc='git --color=auto commit'`, embedded-quote `say='echo '\\''hi'\\'''`)
are preserved untouched; dropped names are logged at debug. brush then
falls through to whatever lives on `PATH`, which is what the user
expected when they ran `which` in the first place.
Covered by unit tests for the sanitizer (Fedora-which case, every
incompatible-metachar shape, every preserve case) and an integration
test that loads a poisoned snapshot and verifies `which sh` now exits
`0` with a real path.
Fixes#3234
- Removed the `readHashLines` setting to consolidate hashline display logic.
- Simplified `resolveFileDisplayMode` to derive hashline visibility solely from the active edit mode.
- Added automatic cleanup of the `readHashLines` key from existing configuration files.
- Updated model detection to exclude WebP format for Codex-based providers, which do not support it.
- Forced the image resize pipeline to encode to PNG or JPEG for incompatible models to resolve transmission errors.
- Added test coverage to verify that WebP images are re-encoded when using the Codex Responses backend.
When the user set `providers.tinyModel` to a local key, `generateSessionTitle`
still raced local against the online `smol` path with a 10 s timeout and
silently fired the online request whenever the local worker returned `null`
(unknown key, model not downloaded, transformers.js failure). The online path
resolves the `smol` role through `priority.json` (haiku → flash → mini → …);
with an `OPENROUTER_API_KEY` picked up from env, that silently billed
OpenRouter without consent.
Drop the race entirely for local choices: honor the user's setting, log a
warning on local failure, leave the session untitled. The `raceFirstNonNull`
helper and `TITLE_LOCAL_FALLBACK_DELAY_MS` had no other consumer and are
removed; the obsolete \"silently bills online when local fails\" tests are
flipped into regressions that lock the no-fallback contract, including the
unknown-key path (e.g. \"ollama:gpt-oss\") which previously also leaked
straight through to the online billing path.
Fixes#3187
- Update `shouldRetry` to treat `EISDIR` and `ENOTDIR` as terminal errors, preventing unnecessary retries when encountering Git reference directory conflicts.
- Add a test suite to verify graceful resolution of branches in scenarios where a packed ref conflicts with a directory path in the filesystem.
- Added a `proseOnlyThinking` configuration setting to suppress raw code blocks in AI thinking traces.
- Implemented `formatThinkingForDisplay` utility to replace code blocks with ellipses in the UI.
- Integrated runtime toggling and live refreshing of message components via streaming reveal controllers.
- Added a live tokens-per-second indicator to the assistant thinking pulse.
- Verified logic with new unit and integration tests for thinking block presentation.
- Added `safeSend` helper wrapping `Subprocess.send()` so sync throws and async EPIPE rejections cannot escape.
- Replaced inline try/catch send wrappers in STT, TTS, and tiny-title clients with shared `safeSend`.
- Added `isIpcSendEpipe` predicate and made matching rejections non-fatal in the `unhandledRejection` handler.
- Added contract tests for `safeSend` and `isIpcSendEpipe` covering sync throws, async rejections, and edge cases.
roboomp