Add Pi structured native chat through its RPC mode by brennanb2025 · Pull Request #25851 · stablyai/orca

brennanb2025 · 2026-10-06T09:39:15Z

	Files	Added	Deleted	Net
Test	42	$\color{#1a7f37}{\Huge{\mathbf{+}}}$4121	$\color{#cf222e}{\Huge{\mathbf{−}}}$11	$\color{#1a7f37}{\Huge{\mathbf{+}}}$4110
Prod	73	$\color{#1a7f37}{\Huge{\mathbf{+}}}$3871	$\color{#cf222e}{\Huge{\mathbf{−}}}$309	$\color{#1a7f37}{\Huge{\mathbf{+}}}$3562

ELI5

Pi chats currently use a terminal. With “Use updated structured native chat” enabled in Settings > Chat UI, supported Pi installations now open native chat: answers stream into messages, tools show their results, extension questions can be answered, and Stop ends the work. Turning the setting off keeps the existing terminal chats, including the OpenCode and OMP transcript readers.

Supported: stable @earendil-works/pi-coding-agent from 0.84.0 (0.84 and later 0.x releases, and 1.x), the releases that report when a turn has fully settled; recordings and live QA used 1.0.4. Older Pi, including @mariozechner/pi-coding-agent 0.73.x, keeps the terminal chat. Orca asks the execution host's own Pi binary for its version before creating a structured chat and again right before starting it, through the same version check and Command setting OpenCode's structured chat uses, so your Command setting picks the Pi that is checked and run.

What Changed

Pi uses its own --mode rpc interface. A reusable JSON-lines transport owns one supervised process and builds the existing chat history. Input, queued writes, pending prompts and final history writes have finite limits. Prompt delivery continues while an extension asks a question; steering and follow-ups travel with the prompt. A turn finishes after Pi reports settlement and an idle check confirms that streaming, compaction and queued messages have finished.

Pi’s session file is the resume handle. A native conversation can continue in Pi’s terminal interface and then resume natively from the same file. Resuming in another folder forks that file through Pi. Pi entries that originated only in terminal history still resume in the terminal; native chat renders Orca’s existing history and does not import prior terminal messages. Model and thinking choices, image prompts, usage, retry failures and extension dialogs use the existing chat surfaces. Stop sends abort, requests a bounded checkpoint, then terminates through the shared supervisor.

An extension input dialog the user answers with nothing (Pi accepts an empty answer) now shows "Empty answer" on its resolved card. Before, the card read "Selected answer unavailable", the wording for an answer Orca cannot read, because the card treated an empty text as no answer. An answer that names neither text nor a known option still reads as unavailable.

Built on #25845 (merged). Pi's version check and binary lookup now use the shared ones #25845 added (agent-cli-version-probe.ts, structured-agent-command-resolution.ts); Pi's own copies (rpc-version.ts, rpc-create-support.ts) are deleted, and Pi's registration lives in src/main/pi/rpc-runtime-registration.ts.

The existing registered-agent routing and setting gate are reused. Clients that cannot answer Pi dialogs are not offered Pi structured sessions; cleanup stays available. The current mobile client does not advertise Pi. Windows launches require the same process identity proof as other supervised providers. Folder workspaces use the host’s existing path resolver.

Why

The native RPC interface follows the common pattern and avoids an extra protocol adapter. The transport and history lane can support another agent later; this integration adds Pi, with OMP left out. It builds on the Grok/ACP integration from #25225, which is now on main; this branch carries only Pi's change on top of it.

Differences from the common pattern:

Intended: Pi runs with full access only. No supervised permission mode or approval-policy extension is injected. Pi extension questions still use existing question and approval cards.
Intended: settlement plus an idle check protects against compaction, retries and queued input outliving an earlier end event.
Intended: Pi’s own session store enables terminal/native context resume; native history remains in Orca’s existing chat store. Existing cards offer editable editor prefill and explicit empty answers through optional metadata.
Intended: an oversized protocol record closes the connection, so a dropped reply or completion cannot silently strand the session. The shared supervisor owns group/tree termination and preserves uncertainty about descendants separately from observed root exit.
Intended: Pi and the Agent Client Protocol (Grok) now share one stdin write queue. A request that times out while still waiting in the queue is dropped and only that request fails, as before. A request that times out while its line is already being written now closes the connection (previously Grok's queue left it in place): the line cannot be taken back, and leaving it would hold every later write, Stop's cancel included, behind an agent that stopped reading.
Intended: current mobile and older paired clients are withheld Pi until they support its launch and dialog contract. Direct WSL structured launch remains unsupported; the existing terminal route remains available.
Intended: Pi older than 0.84.0 keeps the terminal chat. The common pattern also requires a minimum Pi release (0.84.0 is accepted by it), because only newer Pi reports when a turn has fully settled.

Linked Issue

Internal coordinated native-chat integration; no new issue opened.

Visual Proof

Live QA in a hidden Orca build of this branch on macOS, with a private Pi 1.0.4 (@earendil-works/pi-coding-agent) signed in through a Codex account. Every launch ran with its own home, profile and agent folders, and the real ~/.codex, ~/.claude, ~/.orca, ~/.grok and ~/.pi files were hashed before and after every check: no change came from the test app.

Check	Result	Screenshot
Setting off: Pi opens the existing terminal chat	Pass
Setting on: structured Pi chat, streamed reply, the shell command's row (folded into the finished turn)	Pass
Stop while a command runs: interrupted, the command's file never written, the next prompt continues the same Pi session	Pass
A second message sent while Pi works is queued and delivered once	Pass
Extension dialogs: confirm Yes / No, select, an empty input (now reads "Empty answer"), and an edited editor text that Pi received as edited	Pass
Restart Orca, reopen the chat: Pi still remembers the conversation and keeps writing to the same Pi session file	Pass

Not verified live: an older Pi (0.73.x) keeping the terminal chat (no older Pi was installed on the test machine; the version check is unit-tested), Windows, Linux, SSH and mobile. Checks 1-4 ran at e54c521, checks 5-6 at 4c25ad0; the commits between them are a main merge and the empty-answer fix. Not re-run live after the merge with #25845 and the move to the shared version check; Pi 0.84.x was not run live (unit-tested).

Testing

I manually tested these changes locally.
Automated tests added/updated.

Scrubbed recordings of the exact supported Pi 1.0.4 package were replayed through the existing chat history implementation: prompts/tools, delayed prompt replies with and without request IDs, steering/follow-ups, Stop, provider errors, signed-out state, extension dialogs and compaction. All assigned Pi capture cells are complete.

From live QA: an explicitly empty (or whitespace-only) input answer now reads "Empty answer" (NativeChatResolutionReceipt.empty-answer.test.tsx, fails with the old receipt); the editor dialog submits the text the user edited rather than its prefill (NativeChatStructuredQuestionCard.test.tsx); a chat reopened after an Orca restart, with a fresh record store reading the journal back, launches Pi with --session on the stored session file (rpc-launch-resolution.test.ts).

Focused checks passed: 62 tests in four version/launch files, 20 tests in three location/adapter/registry files, 120 tests in twelve files covering the earlier main merge and last messages under history-write pressure, and 68 tests in eight files covering the latest reasoning/settlement merge, Pi capture replay and unchanged provider contracts. Earlier full integration and protocol review snapshots passed 230 tests in 22 files and 104 tests in twelve files respectively; these are separate runs. Scoped lint, formatting and changed-code/design-system checks passed.

Final head f3be80a4f9c (origin/main with #25845 merged in): both typechecks pass locally; 74 test files (516 tests) covering the changed and dependent code pass, plus the Pi version tests for the new 0.84.0 minimum; lint and the changed-code quality gate pass. Not run locally: the 4 mobile test files (mobile config; no mobile code changed in the merge). Earlier head 4c25ad02e4c passed all 19 applicable CI checks.

Review

Fresh protocol and integration reviewers, an architecture challenger and narrow follow-up reviewers checked the change. Findings were fixed and tracked in a private ledger. Final readiness was reviewed against the complete diff without the counsel step, as requested.

An existing shared shutdown path waits for chat-history writes before releasing ownership. If a write stalls, bookkeeping can delay opening the chat again. That pre-existing cross-provider issue is deferred to a separate fix; this PR adds no new ownership-release wait.

Agent skill upstream boundary

Not applicable; no skill-installer source, tests, fixtures, registry or path tables copied.

Notes

The execution host owns the session file, account environment, process and status. No OMP integration, new transcript store, global extension install, custom tool injection, permission hook, rewind or thread-goal feature is included. The optional extension-specific OpenRouter token-budget hook is absent because this lane injects no extension.

Author (X): @BrennanKB5.

Checklist

This PR is focused on Pi structured chat and its required shared transport/lifecycle support.
I explained the user-facing before/after, mechanism and approach.
Before/after visual evidence attached (see Visual Proof).
Self-reviewed correctness, security and performance.
Cross-platform, SSH/remote, mobile compatibility and path impact considered.
Final CI typechecks, tests and builds pass; CI is the authority.

…embler' into brennanb2025/acp-c5-tool-interrupted

A call still running when its turn ends takes the state of that turn's row: a row another writer settled first (a person's Stop) stands, so its calls read interrupted whatever the provider's later end reports. The no-ending path that settled calls from the Stop row is gone, since a Stop now leaves running calls to the provider. Adds the two Spanish strings.

A failed Grok turn ended with no reason on screen: the translator dropped every copy of Grok's message. The failed turn now gets one status row in Orca's existing "provider did not accept this message" words with Grok's reason, read from whichever copy arrives first (the given-up retry, the turn's end, the prompt's completion notice, or the prompt's error answer); later copies only fill a reason the row still lacks. A running background command no longer reads "Background task <id> started": a task's summary is mapped only once it has settled. A monitor stays a monitor when the agent reads its output: a frame that names no kind keeps the known one, and a "[monitor" command is a monitor. A prompt's turn is marked started, so a late frame for an ended prompt neither reopens it nor becomes the active turn. A tool's turn is held in one place at a time.

A turn that started and then failed was told "The provider did not accept this message", Orca's sentence for a message refused before its turn. The row now reads as a Codex turn-ending error does: an error status row with the provider's own words. With no words, the dialect names the failure ("Grok ended this turn with an error." / "Grok usage limit reached."), else the agent's display name does.

…start too The restart sweep ended every running call by the death evidence alone, so after a person's Stop with no proof the child died the call read failed under a turn that read interrupted. The sweep and the live dead-generation settlement now ask the same rule the assembler does: a call in a turn already settled ends as that row ended; only a turn still running leaves its calls to the evidence.

…mit what they claim A StructuredAgentRegistry, built once from the {definition, adapter} registrations, is now a required host dependency and the router routes with it. The adapter interface loses its optional router-only capabilities?() and definition?(); every reader (options at rest, thread goal, rewind, the /compact handover) asks the registry. A live session still narrows rewind through the adapter, and the host combines declared and narrowed in one helper. The registry refuses a registration that declares compact, a thread goal or rewind without the adapter method behind it. /compact is admitted by the declared capability, so an agent that declares compact:false gets the commandRefused fact instead of a thrown error. Also: the cut-turn notice names the agent from the catalog, create-support builds the account home through agentSessionAccountHome, and the turn status text older clients read names the session's agent instead of defaulting to Codex (Claude/Codex text unchanged).

…to brennanb2025/acp-d2-rebase2

…otocol client's own types The translator now reads a permission request with the client's lenient reader, a session update with its session-event reader, and takes only the agent's own error answer as a failed prompt's reason, so an Orca-side error never reads as the provider's words. Tests cover protocol values newer than this build.

The router holds the registry, so its rewindSupport answers the owner's declared rewind narrowed by the adapter, in one place no reader can bypass. The host-side combining helper is gone; readers ask the adapter they hold.

…ry' into brennanb2025/acp-a3-wire # Conflicts: # src/main/native-chat/agent-session-wire/structured-agent-session-adapter-router.ts # src/main/native-chat/agent-session-wire/structured-agent-session-adapter.ts # src/main/native-chat/agent-session-wire/structured-agent-session-host-types.ts # src/main/runtime/orca-runtime-get-structured-agent-session-create-support.ts # src/main/runtime/structured-agent-session-runtime.ts

The adoption replay test builds its host over this build's registered agents instead of an adapter definition method, and the stored-form test passes the stored agents the record guard now requires.

…nly empty capability record

…gent-registry

…ry' into brennanb2025/acp-a3-wire

…he desktop The desktop renderer calls the restart methods as a paired runtime client without the registered-agents capability, so it was handed a Claude/Codex audience and its "dismiss all" took the scoped path: no dismissal fence and no clearing of unwritten teardown witnesses, so a late teardown write could bring a dismissed offer back. The audience is now derived against this host's registry: a client that can show every agent the host registered gets none, and dismisses exactly as before (one fence for every offer). A client that truly cannot show some agent still gets a final dismissal for what it sees: each cleared session is fenced on its own (bounded, superseded by a dismiss-all fence) and only the witnesses of agents it sees are dropped. One predicate now answers which agents a client renders, for both tab projection and restart offers; the Claude capability rule is folded in.

…'s chats A record was readable only if every handle used the transport its agent's definition declares today, and its account variable was declared by some registered agent. A later build that drives the same agent over another protocol, or under a renamed variable, would have set every existing chat of that agent aside: no tab, no history, no error. Readability now asks only that the record agree with itself: its agent is registered, every handle shares one namespace owned by that agent, and the account variable is a well-formed name. Whether this build can drive it (the chain's transport is the one the agent speaks, and the account variable is that agent's own, since it becomes the child's environment) is decided in performAttach, the one admission every start of every agent passes through, and refused there as hostUnsupported. A record pinning another agent's variable is refused the same way rather than passing because some agent declares it.

… and which account it pins createSupport, create and the model catalog's account read still chose a location rule and an account resolver by name (Claude, Codex, else none), so a new agent would have needed a third branch beside the registration list that already decides storage, routing and publication. Each runtime registration now carries supportsLocation(location) and resolveAccountHomePath({ launchEnv, location, purpose, workspacePath }), and Claude's and Codex's rules move into their entries unchanged: a read still syncs no home and starts no bridge, a Codex launch still trusts the workspace first, and WSL locations resolve as before. The list exists before the host is built, so none of these reads installs the host, and an agent the list does not hold is answered no without opening the journal.

…f the capsule state A spread input already carries the registry, so the explicit key was overwritten (TS2783). The session-fence helper now returns only the fence, not the capsule state it was given.

The per-session dismissal fence added a forever-persisted capsule field with an arbitrary cap that served no caller: only the local desktop calls the restart methods, and on a host whose agents it all shows it gets no audience and takes the unscoped, fenced dismissal. A scoped dismiss-all (a caller that cannot show some registered agent) keeps that agent's offers, clears only its audience's unwritten witnesses, and writes no fence; a late write from this process is serialized behind it. The field never left this unreleased branch.

… own The start check judges the record's agent while the router starts the adapter of params.agent, and the attach wire let the two differ. Admission now refuses agent !== provider as requestMalformed, so the agent checked is the agent started. Every host-built attach already sets them equal.

…ccept it The restart offer, a failure's retryable flag and the pre-send check asked only whether the adapter runs the chat's location, while the start also refuses a record this build cannot drive; such a chat was offered Resume and its Retry failed forever. One host predicate, hostCanStartRecord (location support and agentDrivesSession), now answers all four; adapterSupportsRecord stays the reader gate, and an undrivable chat's offer is kept, not retired.

…nly if this build drives it A record-scoped catalog read started the agent's lister under the record's account path whatever variable the record pinned. It now uses that path only when agentDrivesSession holds, and otherwise resolves the account as for a read with no record.

…say where transport is checked The store only ever asked whether an agent is registered, so its admission list is now the registered ids; the definition is the one home of an agent's transport and account variable. Comments that said the record store checks transport, and one that cited a nonexistent function, now point at agentDrivesSession and the attach admission. The isPersistedAgentSessionRecord(value, agents) call shape and the test fixture the cross-version probe imports are kept.

…) into brennanb2025/acp-d3-grok

…hat-pi-rpc # Conflicts: # src/main/native-chat/agent-session-wire/structured-agent-session-dead-generation-settlement.ts # src/main/runtime/orca-runtime-get-worktree-ps.ts # src/main/runtime/structured-agent-session-runtime.ts

brennanb2025 · 2026-10-06T14:50:21Z

Pi chats previously ran through the terminal. With the updated structured native-chat setting on, supported Pi 1.x now streams messages and tools into native chat, answers extension dialogs through existing cards, and stops through the shared process supervisor. The setting off and older/unknown Pi versions keep terminal chat; Pi runs with full access only.

Fresh reviews and an architecture challenge led to fixes in bounded writes, message-queue ordering and retries, crash reporting, dialogs, cancellation and last history writes under pressure. Pi audience checks cover content, mutations, status and completion events; older clients and current mobile are withheld Pi. Native-origin Pi files can continue in the terminal and then resume natively with that context. Terminal-only history entries retain terminal resume; arbitrary history adoption and OMP are outside this work. The branch includes the approved Grok/ACP dependency from #25225.

The floor is stable @earendil-works/pi-coding-agent 1.x, recorded at 1.0.4. Temporary: older Pi line (agent_end-only) support awaits a captured follow-up with a guarded fallback. An existing shared shutdown path can let stalled chat-history bookkeeping delay opening the chat again; that pre-existing P2 is deferred to a separate cross-provider fix. This change adds no ownership-release wait.

Verification: recorded supported-version flows replay through the existing history implementation. 62 version/launch tests, 20 support/lifecycle tests, 120 earlier-main overlap/admission/provider-contract tests and 68 newest-main reasoning/settlement/Pi overlap tests passed in separate focused runs. Fresh version review found no proven defect; final readiness was completed without counsel as requested. Current head 4c25ad02e4ca487b99648abc2f35743dadf23a8c (empty input answers read "Empty answer"; origin/main merged) passed all 19 applicable checks (16 skipped), including typechecks, builds, all five unit-test shards, cross-version wire checks, mobile checks and macOS/Linux/Windows ownership contracts: CI run, first attempt. Earlier c410 CI required retries after an unchanged terminal-history timer assertion failed and a subsequent runner shut down; no unrelated test or production code was changed.

Live app QA is BLOCKED: no authorized UI driver is available, and the coordinator has held further launches. The earlier hidden M4Air app startup wrote managed Codex hooks into real ~/.codex outside its test profile; the driver stopped only its owned process group and restored nothing. No Pi user flow or screenshot was completed. Rendered desktop/mobile and real Pi on Windows/WSL/SSH remain unverified. All assigned Pi captures are delivered, with no capture blocker. This remains a draft pending that evidence.

… RPC Main's squash of the Grok ACP change is authoritative; this branch keeps only Pi's own change on top of it, including the provider-neutral start tracking.

Main added prompt alerts to the turn-completion stream; the merge extended Pi's audience filter to them.

… being written A line still waiting in the queue is dropped and only its request fails, as before. A line already handed to the stream cannot be taken back and would hold every later write, a cancel included, so that timeout closes the connection; tests now cover both for the shared queue and for ACP.

…hat-pi-rpc

Its host id is a branded type; the literal only checks under the wire type, as the completion fixture beside it does.

It opens a real SQLite record store, which main's new boundary check now requires to run on Node.

…/native-chat-pi-rpc

…n file A fresh record store reading the journal back from disk must hand the launch the session file the chat proved before the restart, and the spawn must carry it as --session.

…able Live QA: answering a Pi input dialog with nothing (which Pi accepts) left the resolved card reading "Selected answer unavailable", the wording for an answer Orca cannot read. The receipt joined the answer's labels and text and turned an empty result into "no answer". It now keeps an answer whose text is present but empty (or whitespace) as an empty answer, shown as "Empty answer"; an answer naming neither text nor a known option still reads as unavailable. Also pins that the editor dialog submits the text the user edited, not its prefill.

…hat-pi-rpc

Pi's runtime registration moves to its own module in main's registration shape (resolveAccountHome, supportsLaunch), so createSupport's Pi version check runs through main's create-support composition instead of a Pi branch in the runtime class. agentSession.agents reads main's static registrations and still hides Pi from clients without the Pi dialog capability. Pi's question card forwards main's answer-input ref.

…n probe and command resolver Pi's own `--version` runner and PATH lookup are gone. Create support and every launch now resolve the binary through resolveStructuredAgentCommand, so the Pi Command setting picks the binary both check and spawn, and an unrunnable Command is refused at launch with agentCommandNotRunnable as for the other agents. The version check runs through probeAgentCliVersion with the same rule as before: stable 1.x of the current package; the older 0.x package keeps the terminal chat. Pi's launch reads the shared per-agent environment overlay and Command settings instead of Pi-only runtime deps.

coderabbitai · 2026-10-08T00:17:12Z

📝 Walkthrough

Walkthrough

This change adds Pi structured-session support through a new JSONL-RPC transport, Pi launch and session adapters, runtime registration, and client capability gating. It also adds shared provider-process queue and acquisition helpers, updates ACP to use them, and adds timeline final-tail handling. In parallel, it extends question schemas with freeTextInput metadata and uses it in shared helpers, mobile UI, renderer UI, and receipt rendering so prompts can preserve empty or whitespace answers when allowed and can carry initial text, placeholders, and multiline settings.

Priority: ➖ Normal

Severity of issue fixed: Medium

Merge Risk: 🟡 Moderate · up to f3be8

Closing a Pi native chat session normally can fail with an error because the close result is misread. This can break chat cleanup and reopening. The fix is a one-line condition change plus a test, and it should be made before merging.

🚥 Pre-merge checks | ✅ 4 | ❓ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Docstring Coverage	❓ Inconclusive	Docstring coverage is 19.05% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 42 functions across 50 files. (65 skipped…	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the primary change: adding Pi structured native chat through RPC mode.
Description check	✅ Passed	The description is detailed and covers the required sections, including user impact, implementation, rationale, visual proof, testing, compatibility, review notes, and checklist status. It states that…
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

Full details: Docstring Coverage

Explanation

Docstring coverage is 19.05% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 42 functions across 50 files. (65 skipped: 14 unsupported, 51 over the file limit.)

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

Commit to this branch
Create a new PR

Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

coderabbitai

Actionable comments posted: 1

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI
Review profile: CHILL
Plan: Advanced
Run ID: fdb6dc11-0ea3-4878-a95f-e01f5a1283ff

📥 Commits

Reviewing files that changed from the base of the PR and between a551aa5 and f3be80a.

📒 Files selected for processing (115)

config/scripts/check-runtime-electron-ratchet.mjs
config/scripts/check-runtime-electron-ratchet.test.mjs
config/scripts/vitest-sqlite-runtime-files.mjs
mobile/src/session/MobileNativeChatQuestion.test.tsx
mobile/src/session/MobileNativeChatQuestion.tsx
mobile/src/session/mobile-native-chat-question.test.ts
mobile/src/session/mobile-native-chat-question.ts
mobile/src/session/mobile-structured-agent-prompts-free-text.test.ts
mobile/src/session/mobile-structured-agent-prompts.ts
mobile/src/session/mobile-structured-grouped-question.test.ts
mobile/src/session/mobile-structured-grouped-question.ts
src/main/acp/acp-json-rpc-peer.test.ts
src/main/acp/acp-stdio-error-boundary.ts
src/main/acp/acp-structured-session-adapter.ts
src/main/acp/acp-write-queue.ts
src/main/agent-launch/agent-launch-executor.test.ts
src/main/ipc/desktop-renderer-runtime-capabilities.ts
src/main/jsonl-rpc/agent-connection.test.ts
src/main/jsonl-rpc/agent-connection.ts
src/main/jsonl-rpc/peer-limits.ts
src/main/jsonl-rpc/peer.test.ts
src/main/jsonl-rpc/peer.ts
src/main/jsonl-rpc/timeline-final-tail.test.ts
src/main/jsonl-rpc/timeline-lane.test.ts
src/main/jsonl-rpc/timeline-lane.ts
src/main/native-chat/agent-session-journal/journal-prompt-body-bounds.ts
src/main/native-chat/agent-session-timeline/provider-timeline-assembler.ts
src/main/native-chat/agent-session-wire/structured-agent-session-event-sink-queue.ts
src/main/native-chat/agent-session-wire/structured-agent-session-final-tail-reservation.test.ts
src/main/native-chat/agent-session-wire/structured-agent-session-final-tail-reservation.ts
src/main/native-chat/agent-session-wire/structured-agent-session-transition.ts
src/main/pi/__fixtures__/command-only.jsonl
src/main/pi/__fixtures__/compaction.jsonl
src/main/pi/__fixtures__/dialogs.jsonl
src/main/pi/__fixtures__/provider-fetch-error.jsonl
src/main/pi/__fixtures__/provider-socket-error.jsonl
src/main/pi/__fixtures__/signed-out.jsonl
src/main/pi/__fixtures__/steer-followup.jsonl
src/main/pi/__fixtures__/stop-mid-tool.jsonl
src/main/pi/rpc-agent-definition.ts
src/main/pi/rpc-capture-replay.test.ts
src/main/pi/rpc-compaction.ts
src/main/pi/rpc-context-usage.ts
src/main/pi/rpc-dialog-callbacks.ts
src/main/pi/rpc-extension-dialogs.ts
src/main/pi/rpc-idle-check.ts
src/main/pi/rpc-launch-resolution.test.ts
src/main/pi/rpc-launch-resolution.ts
src/main/pi/rpc-launch.test.ts
src/main/pi/rpc-launch.ts
src/main/pi/rpc-location.test.ts
src/main/pi/rpc-message-dialog-shapes.test.ts
src/main/pi/rpc-messages.ts
src/main/pi/rpc-options.test.ts
src/main/pi/rpc-options.ts
src/main/pi/rpc-prompt-delivery.test.ts
src/main/pi/rpc-prompt-delivery.ts
src/main/pi/rpc-prompt.test.ts
src/main/pi/rpc-prompt.ts
src/main/pi/rpc-protocol.ts
src/main/pi/rpc-retry-activity.ts
src/main/pi/rpc-runtime-registration.test.ts
src/main/pi/rpc-runtime-registration.ts
src/main/pi/rpc-session-adapter.test.ts
src/main/pi/rpc-session-adapter.ts
src/main/pi/rpc-session.ts
src/main/pi/rpc-turn-races.test.ts
src/main/pi/rpc-turns.ts
src/main/provider-process/provider-acquisition-starts.test.ts
src/main/provider-process/provider-acquisition-starts.ts
src/main/provider-process/provider-stdio-error-boundary.ts
src/main/provider-process/provider-stdio-write-queue.test.ts
src/main/provider-process/provider-stdio-write-queue.ts
src/main/runtime/rpc/methods/session-tab-registered-agent-projection.test.ts
src/main/runtime/rpc/methods/structured-agent-session-agents.ts
src/main/runtime/rpc/methods/structured-agent-session-conversation-outline.ts
src/main/runtime/rpc/methods/structured-agent-session-gate.ts
src/main/runtime/rpc/methods/structured-agent-session-hold.ts
src/main/runtime/rpc/methods/structured-agent-session-options-read.ts
src/main/runtime/rpc/methods/structured-agent-session-pi-dialog-capability.test.ts
src/main/runtime/rpc/methods/structured-agent-session-policy.test.ts
src/main/runtime/rpc/methods/structured-agent-session-policy.ts
src/main/runtime/rpc/methods/structured-agent-session-queued-methods.ts
src/main/runtime/rpc/methods/structured-agent-session-reveal.ts
src/main/runtime/rpc/methods/structured-agent-session-send-compatibility.ts
src/main/runtime/rpc/methods/structured-agent-session-status-stream.ts
src/main/runtime/rpc/methods/structured-agent-session-thread-goal.ts
src/main/runtime/rpc/methods/structured-agent-session-turn-completion-stream.ts
src/main/runtime/rpc/methods/structured-agent-session.ts
src/main/runtime/structured-agent-runtime-registrations-acp.test.ts
src/main/runtime/structured-agent-runtime-registrations.ts
src/main/runtime/structured-agent-session-runtime.ts
src/renderer/src/components/native-chat/NativeChatQuestionCard.tsx
src/renderer/src/components/native-chat/NativeChatResolutionReceipt.empty-answer.test.tsx
src/renderer/src/components/native-chat/NativeChatResolutionReceipt.tsx
src/renderer/src/components/native-chat/NativeChatStructuredQuestionCard.test.tsx
src/renderer/src/components/native-chat/NativeChatStructuredQuestionCard.tsx
src/renderer/src/components/native-chat/NativeChatStructuredSession.newer-approval.test.tsx
src/renderer/src/components/native-chat/NativeChatStructuredSession.tsx
src/renderer/src/components/native-chat/native-chat-resolution-receipt.ts
src/renderer/src/components/native-chat/use-structured-agent-session.ts
src/renderer/src/i18n/locales/en.json
src/renderer/src/i18n/locales/es.json
src/renderer/src/i18n/locales/fr.json
src/renderer/src/i18n/locales/ja.json
src/renderer/src/i18n/locales/ko.json
src/renderer/src/i18n/locales/zh.json
src/shared/agent-session-journal-schemas.ts
src/shared/agent-session-journal-types.ts
src/shared/agent-session-question-answer.ts
src/shared/agent-session-question-free-text.test.ts
src/shared/electron-remote-runtime-client-capabilities.ts
src/shared/protocol-version.ts
src/shared/structured-agent-session-surface-capabilities.ts
src/shared/structured-native-chat-launch-route.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 8 remain after this review.

coderabbitai · 2026-10-08T00:17:15Z

+    if (result.root !== 'exited') {
+      return false
+    }
+    if (result.tree !== 'exited') {
+      throw new AgentSessionAcquisitionRootExitObservedError(new Error('Pi root exit observed'))
+    }


🩺 Stability & Availability | 🟠 Major | ⚡ Quick win

🔎 Supported by static analysis

🏁 Script executed:

#!/bin/bash # How does the ACP connection derive processTreeUnproven from a close result's tree verdict? rg -nP -C6 '\bprocessTreeUnproven\b' src/main/acp src/main/provider-process # How does the managed process produce tree: null on a graceful root exit? rg -nP -C4 'tree:\s*null' src/main/provider-process

Repository: stablyai/orca

Length of output: 13385

Do not treat tree: null as an observed descendant exit.

ManagedProviderProcess returns { root: 'exited', tree: null } when the root exits after stdin closes. The current closeSession check throws for that result, so a normal Pi close can reject. The ACP adapter treats null as no tree observation and only reports live or unverifiable as unproven.

🐛 Suggested fix

--- "a/src/main/pi/rpc-session-adapter.ts" +++ "b/src/main/pi/rpc-session-adapter.ts" @@ -240,12 +240,12 @@ return true } const result = await session.close(requested) if (result.root !== 'exited') { return false } - if (result.tree !== 'exited') { + if (result.tree !== null && result.tree !== 'exited') { throw new AgentSessionAcquisitionRootExitObservedError(new Error('Pi root exit observed')) } return true } releaseAcquisition(input: { sessionId: string }): Promise<boolean> {

Add an adapter test with closeResult = { root: 'exited', tree: null }.

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

if (result.root !== 'exited') {

return false

}

if (result.tree !== 'exited') {

throw new AgentSessionAcquisitionRootExitObservedError(new Error('Pi root exit observed'))

}

if (result.root !== 'exited') {

return false

}

if (result.tree !== null && result.tree !== 'exited') {

throw new AgentSessionAcquisitionRootExitObservedError(new Error('Pi root exit observed'))

}

brennanb2025 added 30 commits October 5, 2026 01:47

Merge remote-tracking branch 'origin/brennanb2025/acp-c1-timeline-ass…

6d8eefc

…embler' into brennanb2025/acp-c5-tool-interrupted

Read a monitor from Grok's exact output prefix

e88bbf0

Keep the dead-generation settlement under the line cap

70fd4ba

Register the ACP schema verify step in the PR preflight phase test

7aa40a9

Merge remote-tracking branch 'origin/brennanb2025/acp-d1-protocol' in…

cc38107

…to brennanb2025/acp-d2-rebase2

Merge remote-tracking branch 'origin/main' into brennanb2025/acp-a3-wire

9e7b78a

test(agent-session): register the agents the merged-in tests now need

b494ed6

The adoption replay test builds its host over this build's registered agents instead of an adapter definition method, and the stored-form test passes the stored agents the record guard now requires.

chore(agent-session): the registry is the one lookup; drop the test-o…

7591167

…nly empty capability record

Merge remote-tracking branch 'origin/main' into brennanb2025/acp-a2-a…

288f72c

…gent-registry

Merge remote-tracking branch 'origin/brennanb2025/acp-a2-agent-regist…

e260b93

…ry' into brennanb2025/acp-a3-wire

Merge remote-tracking branch 'origin/main' into brennanb2025/acp-a3-wire

65a967b

test(agent-session): drop a duplicate registry key; keep fences out o…

a83313b

…f the capsule state A spread input already carries the registry, so the explicit key was overwritten (TS2783). The session-fence helper now returns only the fence, not the capsule state it was given.

docs(agent-session): the record store admits the registered agents' ids

364f11f

Merge remote-tracking branch 'origin/brennanb2025/acp-a3-wire' (364f11f…

ad2b25a

…) into brennanb2025/acp-d3-grok

brennanb2025 added 4 commits October 6, 2026 06:37

Complete the Pi location-test adapter fixture

c410a3c

Retain the settled-work extraction in the main merge

4bb3881

Use the settled-work module in the new reasoning sweep test

78d37e5

brennanb2025 added 7 commits October 6, 2026 22:17

Merge origin/main (Grok structured chat over ACP landed) into Pi over…

717be71

… RPC Main's squash of the Grok ACP change is authoritative; this branch keeps only Pi's own change on top of it, including the provider-neutral start tracking.

Test that Pi prompt alerts skip clients that cannot read Pi

1b10d1d

Main added prompt alerts to the turn-completion stream; the merge extended Pi's audience filter to them.

Merge origin/main (main's lint fix #26079) into brennanb2025/native-c…

3d12804

…hat-pi-rpc

Type the prompt-alert fixture as the wire's prompt attention

f6d5438

Its host id is a branded type; the literal only checks under the wire type, as the completion fixture beside it does.

Run the Pi launch-resolution test in the Node runtime project

e54c521

It opens a real SQLite record store, which main's new boundary check now requires to run on Node.

Merge origin/main (main's web typecheck fix #26097) into brennanb2025…

a38aa64

…/native-chat-pi-rpc

brennanb2025 mentioned this pull request Oct 7, 2026

OpenCode structured chat over the Agent Client Protocol #25845

Merged

3 tasks done

brennanb2025 added 4 commits October 7, 2026 02:00

Merge origin/main into brennanb2025/native-chat-pi-rpc

57ee8e9

Pin that a Pi chat reopened after a restart resumes its stored sessio…

55f8b26

…n file A fresh record store reading the journal back from disk must hand the launch the session file the chat proved before the restart, and the spawn must carry it as --session.

Merge remote-tracking branch 'origin/main' into brennanb2025/native-c…

4c25ad0

…hat-pi-rpc

brennanb2025 mentioned this pull request Oct 7, 2026

fix(native-chat): a local chat this computer can't run opens the agent in a terminal #25947

Merged

7 of 9 tasks

brennanb2025 added 3 commits October 7, 2026 16:46

feat(pi): run stable Pi from 0.84.0 in the structured chat, not only 1.x

f3be80a

brennanb2025 marked this pull request as ready for review October 8, 2026 00:03

coderabbitai Bot reviewed Oct 8, 2026

View reviewed changes

brennanb2025 merged commit c71601f into main Oct 8, 2026
52 of 64 checks passed

This was referenced Oct 8, 2026

feat(omp): open OMP in the structured chat over ACP #26401

Merged

Save a new chat's first message during create #26332

Merged

This was referenced Oct 8, 2026

chore: sync main into phase35 (sync 1) #26419

Merged

Native chat for third-party agents: pi (pi-coding-agent) adapter / public session-host contract #21558

Closed

nwparker mentioned this pull request Oct 18, 2026

[Feature]: Desktop: support the pi-coding-agent in native Chat UI #13185

Closed

Navigation Menu

Repository navigation

Add Pi structured native chat through its RPC mode - #25851

Add Pi structured native chat through its RPC mode#25851
brennanb2025 merged 433 commits into
mainfrom
brennanb2025/native-chat-pi-rpc

brennanb2025 commented Oct 6, 2026 •

edited

Loading

Uh oh!

brennanb2025 commented Oct 6, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Oct 8, 2026 •

edited

Loading

Walkthrough

❌ Failed checks (1 inconclusive)

Uh oh!

coderabbitai Bot left a comment

Uh oh!

coderabbitai Bot Oct 8, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

brennanb2025 commented Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

ELI5

What Changed

Why

Linked Issue

Visual Proof

Testing

Review

Agent skill upstream boundary

Notes

Checklist

Uh oh!

brennanb2025 commented Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

coderabbitai Bot commented Oct 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

❌ Failed checks (1 inconclusive)

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Oct 8, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

brennanb2025 commented Oct 6, 2026 •

edited

Loading

brennanb2025 commented Oct 6, 2026 •

edited

Loading

coderabbitai Bot commented Oct 8, 2026 •

edited

Loading