mirror of
https://github.com/stablyai/orca.git
synced 2026-10-07 00:02:29 +00:00
Brennan Benson
020cebeff6
* Add standalone ACP protocol client and session runtime * Protect ACP transport teardown from late stream errors * Retire incoming ACP request ids before publishing responses * Narrow ACP configuration requests and transport message types * Remove redundant ACP request handler return unions * Keep ACP waits caller-owned and preserve protocol extensions * Preserve open ACP decisions through prompt completion * Generate open ACP enums and check the generated schema offline A newer or vendor enum value (tool kind, tool status, option kind, stop reason) no longer fails the whole message: generated enums accept the known literals plus any other string, typed so callers can still narrow on the known ones. The generated header now records the pinned input digests, the generator digest and a body hash, so `verify:acp-protocol` catches a stale or hand-edited file without network access; it runs in lint and the PR workflow. * Land the ACP runtime contract the agent adapters use - Deliver notifications other than session/update through onExtensionNotification, in arrival order with session updates. - Accept _meta on prompt, setMode, setModel, setConfigOption and cancel. - cancel() always sends session/cancel once the session runs, since the agent can be in a turn it began itself; only a successful send is shared, so a failed write is retried. - Cancel aborts each open agent request's signal and lets its handler send its own answer; -32800 only when the handler rejects. - Permission requests validate only the session, tool call id and options; unreadable fields are dropped with a diagnostic, and any answer Orca cannot send is `cancelled` instead of a JSON-RPC error. Agent-started turns may ask; whether to show it is the caller's decision. - AcpAgentError marks the agent's own errors; AcpInvalidResponseError keeps the raw answer and validation issues for answers Orca could not read. - Lines over the size limit are classified by prefix (shared with the Codex reader): the owed request fails, an oversized agent request is answered with an error, and an unattributable response closes the connection. * Answer every agent request after an ACP cancel A cancel that lands before a permission handler starts now still runs the permission path, so the agent gets the `cancelled` outcome rather than a request-cancelled error. A handler that ignores the abort no longer leaves the agent waiting: once the abort has run through, any request still unanswered gets request-cancelled. Handlers that answer on abort keep their own reply. Also renames a lint-rejected helper parameter, replaces a Reflect.apply in a test, and stops the permission diagnostic from firing with an empty list. * Let each ACP request handler own its answer after a cancel Removes the next-event-loop-turn fallback that answered request-cancelled for any handler still silent after a cancel. It raced answers that were still being saved (an approval mid-journal-write reached the agent as an error) and made the outcome depend on event-loop timing. The handler that owns an agent request now always sends its answer, or throws for request-cancelled; a request it never answers ends when the connection closes. A permission whose handler had not started still answers `cancelled`. * Register the ACP schema verify step in the PR preflight phase test * feat(acp): a steer's cancel asks once and never ends the agent The runtime had one cancel: send session/cancel, wait at most 10 s for Orca's prompt to settle, then close the connection, which ends the agent. A steer used it too, so a slow agent lost its process just because the person added a message. requestSteerCancel() now sends session/cancel once per prompt, cancels the agent's open requests and answers later permissions cancelled, and never bounds or closes: the prompt's own reply ends it and the steer's prompt follows. cancel() stays the Stop: bounded, then close. A Stop after a steer still bounds and closes. Both cancel paths move into acp-prompt-cancel.ts over one cancel channel. * fix(acp): a repeated steer shares the cancel in flight; say what the caller owns Per review: a second steer before the first write lands returns that write instead of resolving early. The steer's JSDoc says the wait for the prompt's reply is unbounded and that a prompt that fails instead must not take the steer until the caller rebuilds the session; the Stop's says a prompt that settles in time leaves the agent for the Stop's owner to end. The steer test now gives the runtime a handler that would allow: the open permission's signal aborts and the late one never reaches it. * test(ratchet): require src/main/acp now that this PR lands it
94 lines
4.8 KiB
Plaintext
94 lines
4.8 KiB
Plaintext
# A shebang plus CRLF makes vite's SSR transform emit a literal `#!` mid-module,
|
|
# so any suite importing the script dies at load with a SyntaxError. Pin the whole
|
|
# directory rather than the scripts that happen to have a test today.
|
|
/config/scripts/**/*.mjs text eol=lf
|
|
/skill-guides/*.md text eol=lf
|
|
/skill-stubs/*.md text eol=lf
|
|
/skill-stubs/_shared/*.md text eol=lf
|
|
/skills/*/SKILL.md text eol=lf
|
|
/src/cli/bundled-skill-guides.ts text eol=lf
|
|
# Bundled plugin trees are byte-hashed; CRLF checkout would break the pinned hash.
|
|
/resources/plugins/** text eol=lf
|
|
# Relay assets are copied verbatim into the bundle and hashed byte-for-byte into
|
|
# .version, which names the immutable remote install dir. A CRLF checkout makes a
|
|
# Windows-built client disagree with a mac/Linux-built one on the same release,
|
|
# so one host ends up with two relay trees (#17886 review).
|
|
/config/relay-assets/** text eol=lf
|
|
# Pin the bytes so a patch reads and diffs identically on every host. It is NOT
|
|
# what makes the hash right: pnpm hashes a patch LF-normalized, so a CRLF checkout
|
|
# cannot change it. Believing otherwise put a hand-computed raw digest in the
|
|
# lockfile twice and broke every install (#17886).
|
|
# These files are stored LF, which is not always the encoding they were written
|
|
# against -- @vscode/windows-process-tree ships CRLF sources -- so any code that
|
|
# runs `git apply` on one must force `-c core.autocrlf=input` rather than trust
|
|
# the host's setting. See config/scripts/windows-process-tree-gyp-rebuild.mjs.
|
|
/config/patches/*.patch -text
|
|
# Same reason, and pnpm parses these too: a CRLF checkout makes the mobile
|
|
# patches unparseable, so Windows packaging dies on ERR_PNPM_INVALID_PATCH.
|
|
/mobile/patches/*.patch -text
|
|
# The xterm bundle hunks also make a diff nobody can read; review the hand-written
|
|
# source patch under xterm-src/ instead. The sibling patches stay diffable.
|
|
/config/patches/@xterm__xterm@*.patch -diff
|
|
/config/patches/xterm-src/*.patch text eol=lf
|
|
# pnpm parses these unified diffs during Windows installs; keep checkout bytes stable.
|
|
/mobile/patches/*.patch -text
|
|
# Generated wrapper fixtures: collapse them in the PR diff so they stop drowning
|
|
# the reviewable change, and pin LF because they are compared byte-for-byte.
|
|
# Not -diff: the shell diff is the review surface when a wrapper does change.
|
|
/src/main/__fixtures__/shell-wrapper-snapshots/*.txt linguist-generated=true text eol=lf
|
|
# Captured agent PTY transcripts. -text, not `text eol=lf` like the wrapper snapshots above:
|
|
# these carry real CR and CRLF bytes as the terminal emitted them, and line-ending
|
|
# normalisation on a Windows checkout would rewrite the evidence the fixture exists to be.
|
|
/src/main/runtime/__fixtures__/*.txt -text
|
|
/src/main/daemon/__fixtures__/pty-transcripts/*.txt -text
|
|
# Generated runtime English subset: compared byte-for-byte by
|
|
# verify:localization-runtime-catalog, so a CRLF checkout would fail the gate.
|
|
/src/renderer/src/i18n/en-runtime-required.json linguist-generated=true text eol=lf
|
|
# Generated method->params catalog: compared byte-for-byte by
|
|
# verify:rpc-params-catalog, so a CRLF checkout would fail the gate.
|
|
/src/shared/rpc-contract/rpc-params-catalog.generated.ts linguist-generated=true text eol=lf
|
|
# Mobile RPC recording goldens: replay requires the file text to equal what rpc:record writes
|
|
# (LF), so a CRLF checkout would fail every golden.
|
|
/mobile/rpc-foundation/goldens/*.json text eol=lf
|
|
# Mobile web page source. Every text byte here is hashed into an asset digest and from
|
|
# there into buildId, so a CRLF Windows checkout would ship a different bundle id for
|
|
# identical source (91af2897 vs 9d78435e, measured on the bundle this replaced).
|
|
/mobile/src/** text eol=lf
|
|
/mobile/app/** text eol=lf
|
|
/mobile/web-entry/** text eol=lf
|
|
# The blanket pin above would mark a future binary as text; exempt the asset types an
|
|
# RN page actually carries.
|
|
/mobile/src/**/*.png -text
|
|
/mobile/src/**/*.jpg -text
|
|
/mobile/src/**/*.jpeg -text
|
|
/mobile/src/**/*.gif -text
|
|
/mobile/src/**/*.ico -text
|
|
/mobile/src/**/*.webp -text
|
|
/mobile/src/**/*.ttf -text
|
|
/mobile/src/**/*.otf -text
|
|
/mobile/src/**/*.woff -text
|
|
/mobile/src/**/*.woff2 -text
|
|
/mobile/app/**/*.png -text
|
|
/mobile/app/**/*.jpg -text
|
|
/mobile/app/**/*.jpeg -text
|
|
/mobile/app/**/*.gif -text
|
|
/mobile/app/**/*.ico -text
|
|
/mobile/app/**/*.webp -text
|
|
/mobile/app/**/*.ttf -text
|
|
/mobile/app/**/*.otf -text
|
|
/mobile/app/**/*.woff -text
|
|
/mobile/app/**/*.woff2 -text
|
|
/mobile/web-entry/**/*.png -text
|
|
/mobile/web-entry/**/*.jpg -text
|
|
/mobile/web-entry/**/*.jpeg -text
|
|
/mobile/web-entry/**/*.gif -text
|
|
/mobile/web-entry/**/*.ico -text
|
|
/mobile/web-entry/**/*.webp -text
|
|
/mobile/web-entry/**/*.ttf -text
|
|
/mobile/web-entry/**/*.otf -text
|
|
/mobile/web-entry/**/*.woff -text
|
|
/mobile/web-entry/**/*.woff2 -text
|
|
|
|
# Generated ACP schemas are checked byte-for-byte against formatted output.
|
|
/src/main/acp/generated/*.generated.ts linguist-generated=true text eol=lf
|