## Why
Command execution and plugin measurement events lack model and reasoning-effort labels. Attribution needs to reflect the step that invoked the command, even when model settings change before a background process finishes.
## What changed
- Add `model_slug` and `reasoning_effort` to command execution and plugin measurement analytics.
- Capture model context from resolved step settings and carry it through execution, approval, Guardian review, and plugin metrics collection.
- Preserve the first command-start model context when subsequent start notifications arrive.
- Keep the carried context out of serialized protocol items and generated schemas.
## Testing
Extend analytics tests to cover model switches before invocation and during background execution, default reasoning effort, Guardian-denied commands, and repeated start notifications retaining the original model context.
GitOrigin-RevId: af90e1c0d39bab625f2e89786085b61a9b96c0ce
## Why
Running code-mode cells can request Guardian review while the next response is in flight. Clearing the response ID before that response emits `response.created` leaves those reviews without a `parent_response_id`.
## What changed
Keep the latest response ID received in the turn until a later `response.created` replaces it, including across sampling retries.
## Testing
Add regression coverage for reviews before and after a response handoff and for a fresh turn that must not inherit the previous turn's ID. Update retry coverage to expect the last known parent when the retry supplies no response ID.
GitOrigin-RevId: af0a08de09edad59e1a7ade7904a42d616b05679
## Why
Idle Apps clients and cached MCP bindings can retain replaced tool definitions. Shared catalog updates can also invalidate prepared calls even when the current definitions match the captured catalog, including after tools are restored while a call awaits approval.
## What changed
- Share immutable tool arrays across equivalent live discovery contexts and reuse storage for equal results. Let unused providers and cached bindings expire.
- Include requested capabilities and initialization results in the sharing scope, and detach servers that disable catalog caching from live sharing.
- Capture catalog snapshots for prepared calls. Accept equivalent shared catalogs regardless of tool-list order, while rejecting changed definitions and calls captured before an explicit refresh on that client.
## Testing
Add regression coverage for shared storage and scope isolation, release of replaced tools and unused bindings, equivalent catalog restoration, explicit refresh invalidation, and an Apps call completing after catalog restoration while awaiting approval.
GitOrigin-RevId: a0516186e4286d0ff13405fafbab6fcf3f1f3773
## Why
MCP tool parsing eagerly cloned structured output schemas and built full call-result envelopes. Cloning tool definitions also copied their output-schema JSON, even before a consumer needed it.
## What changed
- Introduce `ToolOutputSchema` with immutable `Arc` storage so tool definitions share output schemas when cloned.
- Retain MCP structured output schemas and materialize the call-result envelope only when JSON is requested.
- Update code-mode consumers and schema mutation sites to materialize JSON explicitly, reusing uniquely owned storage when possible.
- Move structured content into the MCP envelope without an extra clone, preserving property order.
## Testing
Add tests for JSON preservation, mutation isolation, equality between lazy and materialized schemas, reuse of uniquely owned storage, and equivalent code-mode definitions.
GitOrigin-RevId: e98ba4c2f0efedc99f7cbc7bba206cc63a3bd8f4
Remove entries from the owned snapshot maps when building MCP server status
responses, moving server metadata, tools, resources, and auth statuses into
the response instead of cloning them. Preserve pagination and missing-entry
defaults.
GitOrigin-RevId: 012301f55ddb9c52c7934a638bdd0310f11077fa
## What changed
Move request execution from `LunaSampler` into `SamplingExecution` in
`sampler/execution.rs`, keeping request preparation and active-request tracking
in the sampler. Preserve the existing retry, authentication recovery,
cancellation, streaming, connection reuse, and token accounting behavior.
GitOrigin-RevId: 5e50116459fe9cc196c30c58e42773fc32be3d02
## What changed
Add `ConversationState` and `ConversationCheckpoint` to `codex-guardian-reviewer` and use them in core review sessions to track transcript cursors, completed review counts, and committed snapshots. Keep history and admitted evidence host-owned.
Preserve the separation between live review progress and committed checkpoints so forks inherit the history, cursor, and review count from the last committed snapshot.
## Testing
Add a unit test verifying that forks retain committed history and progress after an uncommitted review, then advance when the next snapshot is committed.
GitOrigin-RevId: 9f92410b11beec6b8f413c4c922fabba65852399
## What changed
Move full/delta transcript selection into the shared `TranscriptMode::select` API and use it when building guardian prompts. Export `TranscriptCursor`, `TranscriptMode`, and `TranscriptSelection` from `codex-guardian-context`.
Preserve full-transcript fallback when the history version changes or the saved cursor exceeds the collected entry count. Select entries before profile retention, preserving their numbering and returning a proposed cursor that counts all collected entries. Hosts remain responsible for committing and invalidating cursors.
## Testing
Add a regression test verifying that sliding-window retention preserves the collected-entry cursor and that an appended entry is selected and numbered correctly in the next delta.
GitOrigin-RevId: a2192c08e23c18302b0105ba47aeb2780eb4a015
## Why
A summary-free parent context reset could reuse a Guardian review session and carry forward rationale from before the reset.
## What changed
Track destructive history replacements with `reset_version` and include it in the Guardian session reuse key. Ordinary input and compaction preserve this version.
## Testing
Update the review-session reuse test to require a new Guardian thread after a summary-free reset, with no prior review context or previous rationale.
GitOrigin-RevId: bd05c149774839f68b75294751f7619ba7ce0bfe
## What changed
Include `sessionId` and the originating `windowId` alongside `threadId` and optional `itemId` in MCP request metadata. Retain the originating item and window for code-mode cells across waits and compaction, including the window when no matching history item is found.
## Testing
Extend metadata assertions for direct and nested MCP calls. Add a regression test verifying that a code-mode cell retains its original item and window IDs when it resumes after compaction.
GitOrigin-RevId: 5e3cf761f2bed4c16565334c94654496367325fd
## Why
Each `setTimeout` spawned a sleeping thread that remained alive until its delay elapsed, even after `clearTimeout` or cell completion.
## What changed
Replace per-timer threads with Tokio sleep tasks held by `AbortOnDropHandle`, so removing a timeout or dropping the isolate cancels its task. Enter the caller's Tokio runtime on the code mode runtime thread to support scheduling these timers.
## Testing
Add a regression test using virtual time to verify that cleared timers release their tasks, an awaited timer completes, and cell completion cancels remaining timers without emitting their output.
GitOrigin-RevId: 59c7e57ed226cb32633e2d4c25b778cfab2af423
## What changed
- When `CODEX_PROVISIONED_MACOS_CANDIDATE` is `true`, require the provisioned macOS job to succeed and upload its verified packages as release assets. Allow releases to proceed when the job is disabled and skipped.
- Include provisioned archives in `codex-package_SHA256SUMS` and remove per-architecture `SHA256SUMS` files before upload.
- Publish a `codex-provisioned` DotSlash manifest for macOS ARM64 and x86_64 packages, pointing to `bin/codex`.
GitOrigin-RevId: 7e36df484a3b7da6b66180fa31418a6958d1fe77
## What changed
Expose `prepare_windows_sandbox_config` and `PreparedWindowsSandboxConfig`
and use them during config loading. Preserve requirement enforcement and the
separation between the configured mode and the effective sandbox level.
## Testing
Add a unit test covering explicit mode precedence over feature fallback and
ensuring a feature-only fallback leaves the configured mode unset.
GitOrigin-RevId: 98e7f9fbc473b6a841257eca80d9a51950986055
## What changed
- Add a configurable `new_worktree` action, bound to `w`, for local sessions with worktree support enabled.
- Create the worktree from the cached project default branch, preferring remote `HEAD` and falling back to conventional `main` or `master` refs without fetching. Preserve the selected subdirectory and source checkout edits.
- Start and bind a blank session in the new worktree, preserving the selected approval reviewer without sending an initial turn.
- Show creation progress, pause conflicting actions, clean up unclaimed clean worktrees, and report retained checkout paths when session startup fails. Clear pending creation state on reconnect.
## Testing
Add coverage for default-branch selection, source checkout preservation, session ownership and approval settings, cleanup, startup failures, reconnect state, and busy-state rendering. Extend Git environment isolation coverage and test unrelated non-UTF-8 refs.
GitOrigin-RevId: 771ddc12d04dbf0d12e77e9c96c8386cf31e67c5
## Why
Scrolling with `CSI S` can discard history rows in QTermWidget and xterm.js when the viewport grows.
## What changed
Use newlines at the bottom of the history scroll region for `ScrollbackStrategy::Standard`, preserving the composer and restoring the cursor and full-screen scroll region afterward. Fall back to full-screen scrolling when only one history row exists, since `DECSTBM` requires two distinct rows. Skip scrolling when there are no history rows or no rows to scroll.
## Testing
Add regression tests for composer and cursor preservation, scroll-region restoration, emitted newline sequences, single-row scrollback preservation, and no-op growth. Snapshot visible geometry for growth by one, three, and four rows.
GitOrigin-RevId: 1dbfc30bb6997dcc2738ec6b2f42d35c786124de
## Why
Pasting during `Ctrl+R` history search previously went through normal composer paste handling instead of updating the search query.
## What changed
- Append sanitized pasted text to the active query and restart matching from the newest history entry, including for large pastes and image paths.
- Ignore empty pastes so they preserve the selected match.
- Display newlines and tabs as `↵` and `⇥` in the footer while matching the original query, and clamp the cursor safely for very large queries.
## Testing
Add regression tests and snapshots covering pasted query acceptance, empty pastes, draft restoration on misses and cancellation, sanitization, large pastes, image paths, separator rendering, and cursor placement.
GitOrigin-RevId: 2ec13fdddcb7503a72a21fc03b2a27ed36313a8a
## What changed
- Replace the inline task composer with a session list: `n` opens a blank session in the selected checkout without sending an initial turn or interrupting running agents.
- Use single-letter action shortcuts and `Enter` to open a session. `Esc` cancels search or rename without closing the list; `Ctrl+C` quits from the list.
- Load destination settings for new sessions while carrying over explicit permission choices. Retain blank sessions and their drafts across navigation until the first turn.
## Testing
Update tests and snapshots for command center navigation, shortcuts, blank draft restoration, destination defaults, permission carryover, and directory trust.
GitOrigin-RevId: bddd04fdbe6ac2f5b7bd70293fd04f673b97ba02
## Why
Model and reasoning effort updates during a turn can leave request metadata and tool hooks reporting the turn's initial settings. Metadata should describe the step that issued the request or tool call.
## What changed
- Share captured execution metadata across Responses, MCP, and extension tool calls, including model, reasoning effort, and automatic review and Node REPL flags.
- Build Responses tool inventory metadata from the issuing step's finalized tool router. Attach the finalized inventory separately for remote compaction.
- Use captured step settings for pre- and post-tool hooks, and captured review settings for permission-request hooks.
## Testing
Add and extend regression tests for model and effort changes during a turn, MCP metadata, pre-tool hook model attribution, captured review flags, and tool inventory matching the issuing request.
GitOrigin-RevId: 9dee46a8b4839de8b434cdad4ed441cc0bb6ff9b
## Why
Desktop uninstall cleanup needs an installation owner even when the user has not signed in or configured the Windows sandbox. Recording ownership only during provisioning leaves those installations unregistered.
## What changed
- Add an authenticated installation registration request and attempt it during Windows desktop stdio initialization, with a five-second timeout before the initialization response.
- Persist ownership independently of provisioning, preserve existing desktop ownership, and prevent another user or home from replacing the registered owner.
- Validate write authority and retain directory handles and guards to protect the registered home against junction conversion through privileged cleanup.
- Preserve existing CLI homes during desktop uninstall while removing their `CodexSandboxUsers` ACL entries. Report ACL revocation errors and avoid propagating unchanged ACLs.
- Grant the owner `WRITE_DAC` on `.sandbox-bin` and allow elevated-helper fallback when older permissions need repair or the service cannot establish an uninstall watcher.
## Testing
Add tests for installation registration without sandbox settings and for preserving a child's null DACL when revoking an absent SID from its parent.
GitOrigin-RevId: fb48923e9d76758d1bf5b50c7305aa60f91629db
## Why
Direct tool-call records need to stay associated with the invocation that produced each output, including when call IDs are reused. Completeness must describe the recorded call inventory, independently of tool success.
## What changed
- Attach direct-call records to outputs before they enter history, and set `tool_calls_complete` when the invocation's arguments are fully recorded.
- Bound pending recordings and retained metadata, release reservations on completion or cancellation, and invalidate pending records when capture is disabled.
- Apply request budgets to direct metadata and strip it from inference and compaction inputs when capture is disabled.
- Remove executed-call metadata from app-server raw response notifications and exclude its size from Guardian history retention budgets.
- Track call IDs that bypass dispatch so their reuse cannot incorrectly establish Code Mode completeness.
## Testing
Add regression coverage for direct-call attribution, malformed calls, metadata budgets, cancellation, configuration changes, compaction, notification filtering, and Guardian context isolation.
GitOrigin-RevId: 2ebd39c7f141d04788736491495109841656b4c0
## Why
Logon SID lookup previously walked token group entries and SID pointers without checking that they fit within the returned buffer.
## What changed
Add a shared `token_groups` helper with a caller-supplied size limit. Validate the group layout, SID bounds, revision, and length before copying SIDs into owned `TokenGroup` values, preserving group order, duplicates, and attributes. Use this helper for logon SID lookup.
## Testing
Add Windows tests for owned SID storage, preserved order and attributes, truncated layouts, malformed and out-of-buffer SIDs, query size limits, and current-token logon SID lookup.
GitOrigin-RevId: f51ff6aa5266b6c73fe07968015e462c888a95e4
## What changed
- Introduce `PreparedNetworkConfig` to separate proxy preparation from applying managed network requirements, preserving preparation before permission fallback in local configuration loading.
- Add helpers to build portable environment policies, retain selected or managed policies even when the proxy is disabled, and validate policies against the final permission profile using the execution resolver.
- Strip listener addresses from environment network configuration and reject unsupported settings, malformed domain patterns, and invalid Unix socket paths.
- Expose supporting managed-feature and permission-profile configuration helpers.
## Testing
Add coverage for policy retention, listener removal, unsupported and malformed policies, and validation against permission profiles. Strengthen configuration tests to compare rebuilt proxy specifications and verify proxy configuration survives permission fallback.
GitOrigin-RevId: 444bf02d8543eaebdbd2ebc14b56d92ce6d2ad2f
## What changed
Expose `prepare_packaged_windows_sandbox_cleanup` to disable sandbox accounts and stop their processes before returning a `PreparedWindowsSandboxCleanup` guard that retains the setup lock. Its `finish` method removes resources and protections; dropping it only releases the lock, leaving accounts disabled and protections intact.
Keep `clean_up_packaged_windows_sandbox` as a wrapper around both phases, and extract the service's existing package cleanup logic into a dedicated module.
GitOrigin-RevId: a32663171bce027f771f738d71b0c46404edbedc
## What changed
- Add explicit MXC backend selection and carry its identity through exec-server process reporting and sandbox violation classification.
- Launch MXC through the Codex executable with the effective permission profile and command environment.
- Reject exec-server MXC requests when native MXC is unavailable or when they request a TTY, an `arg0` override, or managed networking. Reject private desktop isolation during MXC preparation.
- Allow an explicitly empty child environment and avoid exposing request payload values in launcher decode errors.
## Testing
Add coverage for sandbox selection and unsupported-request rejection, plus Windows RPC tests for stdin writes and temporary-directory permissions derived from the command environment. Native MXC tests skip when MXC is unavailable.
GitOrigin-RevId: 3626ff0f9ad7f9b812ce09b68c31ea9a5a9c72b1
## What changed
- Move the setup helper implementation and its existing tests into `codex-windows-sandbox`, with the binary delegating to `setup_helper_main`.
- Expose installation record types and storage operations from the library, preserving the registry key, size limit, and serialization format. Reuse them in the service and remove its direct `serde` dependencies.
- Extract service provisioning from IPC handling into a dedicated module, preserving authentication and machine-policy checks before provisioning and retaining directory handles through helper execution.
GitOrigin-RevId: dd1ea015aff901d5dfe39b395de1971ee0f2d33f
## Why
The latest `openssl-src` 300.x crate still bundles OpenSSL 3.6.3. Build the 3.6.4 security release directly while preserving the existing 3.x ABI.
## What changed
- Build static OpenSSL libraries with the musl compiler for `x86_64` and `aarch64`, verifying the source archive's SHA-256 checksum.
- Set target-specific OpenSSL overrides to bypass vendored builds without affecting host build dependencies.
- Add `perl` and `make` to the musl build prerequisites and reuse completed OpenSSL installations.
GitOrigin-RevId: 4276c7d6cf30c31c8554246cea2f406dfbf7568b
## What changed
Remove the animated stars shown when selecting Astra, along with their input, terminal-focus, and model-selection hooks.
## Testing
Retain draft text and live voice control snapshots in a standalone composer snapshot test, and remove sparkle-specific tests and snapshots.
GitOrigin-RevId: cbe4396e8700cabfb1e0db9fbe8f997b94316c0f
## Why
Unterminated prose stayed hidden until a newline arrived or the stream completed, leaving long single-line responses invisible while they streamed.
## What changed
- Show live prose previews for agent messages and proposed plans, requesting redraws when the preview changes.
- Keep previews out of the scrollback queue and render the full source on newline or completion without duplicating text.
- Bound previews to the most recent 8 KiB at a Unicode character boundary, with an ellipsis for omitted text, and reflow them when the width changes.
- Preserve holdback for incomplete table and code structures, retaining the last safe prose preview when a pipe arrives.
## Testing
Add controller tests and TUI snapshots covering visibility before completion, resize reflow, finalization without duplication, long Unicode previews, table and code holdback, and inline visualization context.
GitOrigin-RevId: 26c83c218b986625820b20779a46d1b800a7e0ff
## What changed
Add the disabled-by-default `send_message_to_user_async` feature flag so root agents can use the tool without model catalog support. Preserve catalog-based opt-in and keep the tool unavailable to subagents. Register the flag as under development and expose it in the configuration schema.
## Testing
Extend integration tests to cover feature and catalog opt-ins, deduplication when both are enabled, subagent exclusion, and the retired `send_async_message` flag. Check that either opt-in allows messages to be emitted without ending the turn.
GitOrigin-RevId: 3f53181ab074b0432c795550de73a9b1da7caf15
## Why
Unbracketed multiline pastes could submit the report form at the first newline, before the full note was entered.
## What changed
Use `PasteBurst` in the report note editor to insert newlines during rapid paste input and allow `Enter` to submit after the burst ends. Extend the burst window across newlines and tabs, and clear burst state after explicit pastes or other editing input.
## Testing
Add regression tests for multiline paste bursts with short first lines, blank lines, and tabs, verifying the complete submitted note with and without logs. Add a multiline rendering snapshot and a test that bracketed paste stays in the report form, leaves the chat composer empty, and allows immediate submission.
GitOrigin-RevId: a76e680a01767c619a4fcc150d1dcad537e67194
Update the mock user transcript, assistant response, and expected delegation
transcript in `websocket_v2_background_agent_returns_function_output` to use
`blueberry` instead of `strawberry`.
GitOrigin-RevId: a2c6209d2565529d63c7a45a65d1162ee1f60425
## Why
Saving a manual thread name leaves pending title generation running and its progress indicator visible until the request finishes.
## What changed
- Cancel pending title requests after a successful manual rename, including renames from the agents overview, and clear the progress indicator immediately.
- Interrupt canceled title-generation turns and unsubscribe their temporary threads. Ignore late start and completion events so canceled requests cannot affect newer requests.
- Cancel pending title requests when resetting connection state.
- Show only a spinner when generating a title without an existing name, removing the `renaming...` placeholder.
## Testing
Add regression tests for manual and overview renames that verify turn interruption, cleanup, and preservation of the saved name. Cover late events after cancellation and update footer and terminal-title snapshots for the spinner-only display.
GitOrigin-RevId: f9490ea7be9968ea6c35d80bfbac06fee57d0135
## Why
Serialized response items include message IDs, metadata, and JSON escaping that inflate token estimates without adding model-visible content.
## What changed
- Estimate each response item from its content, retaining JSON syntax for structured tool payloads.
- Apply image estimates to all image inputs, including non-base64 URLs, and count audio and encrypted content through their modality-specific estimates.
- Exclude plaintext reasoning and bookkeeping-only items from replay accounting.
## Testing
Update unit expectations for text, images, audio, and encrypted content. Add a remote compaction regression test showing that equal-length text produces identical token usage estimates despite different message IDs, metadata, and JSON escaping, while preserving the submitted messages.
GitOrigin-RevId: 1c43d5abbcc1668b4ab413901a2b4b6511ca5892
## Why
The 900-byte recap prompt limit leaves little room for completed progress, unresolved caveats, and recent corrections. Recaps need this context to distinguish completed work from pending requests.
## What changed
- Introduce a shared `RecapPrompt` with a 32 KiB ceiling for instructions and history, using an 8,192-token estimate. Instruct summaries to retain the broader goal, completed outcomes, and unresolved availability or validation caveats.
- Select up to eight answered exchanges plus a pending request, preserving adjacent steering and progress messages. Drop older whole exchanges before excerpting both ends of oversized messages, while retaining the newest answer and pending correction.
- Require a `summary` and nullable `next_action`, bounded to 700 and 200 characters respectively. Reject malformed or oversized responses and display an optional, separately styled `Next:` line.
## Testing
Add coverage for UTF-8 prompt bounds, exchange selection, pending corrections, excerpt boundaries, strict response parsing, and next-action rendering. Extend the recap generation integration test to verify bounded history and the structured response schema.
GitOrigin-RevId: 6c6a84bc602a168418c1952feb1d286ec0cb9e28
## What changed
- Increase the automatic recap delay from 3 to 30 minutes.
- Replace the recap heading and divider with an italic `↳ Recap:` layout, hanging indentation, and right padding. Preserve explicit line breaks and Unicode, splitting URLs only when they exceed the available text column.
- Use Tokio's clock for automatic recap eligibility and create the timer before spawning its task.
- Remove the `token_budget.use_history_notes_extension` override from temporary structured requests.
## Testing
Add coverage for the 30-minute deadline, recap styling, narrow terminals, Unicode, and URL wrapping. Update recap generation and reconnect tests to advance Tokio's clock instead of relying on host uptime.
GitOrigin-RevId: 79a9bca779020b59a409207163b02944b30c1a20
Download target artifacts, supplemental release assets, and staged npm
packages in a single `actions/download-artifact` step in
`.github/workflows/rust-release.yml` using a combined artifact pattern.
GitOrigin-RevId: 64c63cc7a2e263418b2a4064c1fd46e1c514cc29
## Why
Avoid rebuilding the upstream Docker image and downloading its apt dependencies for each release by using the Python and `gh` already available on Ubuntu runners.
## What changed
Add a shared composite action that checks out the pinned upstream publisher and uses pinned `uv` to install hash-verified Python dependencies from wheels in a virtual environment. Accept multiple configuration paths and publish each manifest for the requested tag.
Update the Rust and zsh release workflows to use the action, combining the Codex and argument-comment-lint manifests into one invocation.
GitOrigin-RevId: 2a7ffef9a076a1d426349bcea587b10599cb8175
## What changed
- Use the same text rendering for model instructions in request settings and Responses Lite developer content.
- Make `rewrite_known_segments` emit plain tags for recognized guidance, including collaboration and multi-agent instructions. Check message roles and complete tag boundaries so lookalike user or tool content stays visible, and retain generated compaction summaries.
- Normalize working-directory, workspace-root, and temporary paths in permission guidance while preserving policy differences.
- Separate text clipping from normalization and retain eight lines at each end of long sections, with fingerprints for omitted content. Update affected snapshots.
## Testing
Add regression coverage for instruction rendering across transports, literal and rewritten guidance, malformed and lookalike tags, compaction summary preservation, and permission path normalization that keeps policy changes detectable.
GitOrigin-RevId: 33b052911323c73b0387d650478f9cc3370fd001
## What changed
- Display input/output token counts and estimated credits and USD cost in task details. Prefer live token totals, falling back to complete totals from usage breakdowns.
- Fetch estimates for the selected task on a one-minute cadence for supported Business and Enterprise plans. Cache results, preserve prior nonzero estimates when a response reports zero, and stop fetching when the capability is unavailable.
- Clear cached usage and discard pending results on account changes, reconnects, or missed server events. Prioritize activity and usage over the original prompt when space is limited.
## Testing
Add a rendering snapshot and tests covering caching, stale results, unavailable usage, token resets, and incomplete usage breakdowns.
GitOrigin-RevId: b72c96053557389fdc723a70fab0dab841a52e49
## Why
The command center previously refused to open tasks managed by another app server, preventing users from viewing their saved history.
## What changed
- Fall back to a frozen, read-only history snapshot when attaching fails because another server owns the task. Preserve drafts and skip buffered request replay and paused-goal prompts while viewing.
- Keep the snapshot when reopening the displayed task, and allow explicit retry to attach once the other server releases it.
- Allow pasting into the command center while a read-only conversation is displayed, while keeping the conversation composer protected.
- Box the TUI startup future to keep it out of the CLI caller's frame.
## Testing
Add coverage for read-only history, draft and selection preservation, retry before and after ownership release, and history-read failure without leaving the current conversation. Assert that opening a read-only task does not start a turn and that the TUI startup future remains below 64 KiB.
GitOrigin-RevId: f1bdb5b2d9182079de4e53346a56742a4a02a1c7
## What changed
- Cycle task grouping through project, status, and model with `Ctrl+S`, and show the active grouping in the footer.
- Group tasks by model with the most recently updated tasks first within each group. Use `Unknown` for missing or empty model names.
- Display the model in task details and refresh it when thread settings change.
- In model grouping, dispatch new tasks without inheriting the selected task's working directory, matching status grouping.
## Testing
Add coverage for model grouping, selection preservation, navigation, and new-task dispatch across grouping modes. Update snapshots for model details and grouping hints.
GitOrigin-RevId: 429cf61c1e7438ba3b2f989d3ed14e12545856f7
## Why
Completed voice captions could disappear before their queued history insertion ran. Interleaved user and assistant updates could also hide the other speaker's caption and restart its animation.
## What changed
- Preserve separate live caption cells for both speakers, rendering the user above the assistant without restarting settled text.
- Render completed captions while history insertion is pending, then move them into history in one app event. Keep them visible when agent streaming delays insertion.
- Invalidate the transcript overlay when an interleaved caption is cleared.
## Testing
Add regression tests for caption visibility through deferred history insertion, settled animations with either speaker arriving first, and empty caption completion during active and stopping sessions. Update replay coverage to exercise the app's history handoff.
GitOrigin-RevId: 31de6fc74b6dc7ff148c554e3176bf3ee92dced9
## What changed
- Add a multi-turn scenario covering `request_user_input_async`, continued work while awaiting an answer, and delivery of the answer into the active turn.
- Add a scenario covering plugin configuration reload in an existing thread, including discovery and use of newly installed skills and MCP tools across turns.
- Render explicit tool output names and namespaces in context snapshots, with a regression test ensuring outputs do not inherit metadata from their calls.
GitOrigin-RevId: 64fca8cb51a3a697fa84bf444ce55b8408961ab5
## What changed
- Use literal model instruction templates and the standard fallback prompt, ignoring legacy personality variables. Retain catalog decoding compatibility and report `supports_personality` as `false` for generated model presets.
- Stop emitting `<personality_spec>` developer messages and assigning an implicit Pragmatic configuration default.
- Filter Friendly/Pragmatic overrides from TUI requests while preserving explicit `personality = "none"`. With `features.personality` enabled, this opt-out still strips the model's personality section.
- Refresh inherited model instructions for agent roles only when the personality opt-out changes, preserving custom instructions.
## Testing
Update coverage for literal legacy templates, deprecated overrides on turn start and resume, role instruction refresh, and forwarding the explicit opt-out through thread start, resume, and fork requests.
GitOrigin-RevId: 1391db4e565e010569f0885f80013d1f86b45825
## What changed
Use `windowsSandbox/setupStart` for elevated and unelevated setup, and handle completion notifications while retaining the pending approval preset and permission profile selection. Verify the effective sandbox mode before enabling Agent mode.
Keep input locked and setup pending when the start request times out or its response is lost. Block thread replacement during setup, ignore completion notifications for a different mode, and clear interrupted setup state with a restart message after reconnection. Preserve the fallback prompt when elevated setup fails and report unelevated setup failures.
Remove the TUI's direct sandbox setup helpers and sandbox-mode configuration writes.
## Testing
Add regression coverage for uncertain setup responses, matching completion modes, blocked thread replacement with retained input, and interrupted setup during reconnection.
GitOrigin-RevId: 5ec1101a1a69b3f96f01fc2ba6a03112705f4918
## Why
Existing threads retain their model provider configuration, which can stop matching managed requirements after those requirements change.
## What changed
- Check retained providers against current managed `model_provider` and `model_providers` requirements before turn start/steer, review, compaction, manual queue start, and active goal updates. Reject requests when requirements cannot be loaded or no longer match; provider mismatches prompt users to restart Codex.
- Load managed requirements independently of user, project, system defaults, and thread configuration, and resolve Bedrock provider overrides before comparison.
- Keep interrupt, realtime stop, and goal pause/clear available. Realtime connections use separate routing and are outside these checks.
- Make detached reviews inherit their parent thread's configuration.
## Testing
Add unit and integration coverage for provider selection and definition changes, Bedrock overrides, requirement load failures, unchanged queue and goal state after rejection, and continued operation after local configuration changes. Extend detached review coverage to verify inheritance of the parent's provider route.
GitOrigin-RevId: 21abf08ad2bcc2ea4698209fa33a17485eed4740
Identify the bundled runtime as `bin/vcruntime140.dll`, relative to the
notice in release packages. Add a dedicated section with links to Microsoft's
license terms, redistributable downloads and support, and Visual Studio 2026
redistribution information.
GitOrigin-RevId: 9ef14ebd9b2d5b2ff0b449d50717f40e7823c727
## Why
The Windows TUI cannot configure a remote executor's sandbox. Required elevated sandbox setup also needs to reflect the local app server's readiness rather than the TUI's local setup files.
## What changed
- Restrict setup prompts and actions to local connections with local executors, including local daemon connections. Let remote servers own Agent permission selection, and warn when local and remote executors are configured together.
- Query `WindowsSandboxReadiness` at startup when elevated sandboxing is required locally, and track successful setup across chat widget replacement and reconnects.
- Restore pending initial input to the composer when mixed executors prevent required setup or a local connection goes offline.
## Testing
Add regression coverage for local, remote, and mixed host selection, remote Agent permission selection, pending input restoration with mixed executors, and setup state preservation during daemon reconnects.
GitOrigin-RevId: bd45b308ac3ae4c2489b0bd2ac5d7e12a7fc35d6
## Why
`connector_auth_failure_from_tool_result` requires an install URL to return an auth failure, even when the tool result contains valid auth failure metadata.
## What changed
Export `is_connector_auth_failure_from_tool_result` from `codex-mcp` to detect auth failures independently of an install URL. Share metadata validation with the existing parser, requiring an error result, an explicit auth failure flag, and a nonempty connector ID that does not conflict with the metadata.
## Testing
Add unit tests for detection without an install URL, rejection of missing or mismatched connector identities and non-auth errors or successful results, and detection for all three supported auth reasons.
GitOrigin-RevId: 6e4e177f26d9a34c56874b0360bae63f6c89107e
