## What changed
- Use cancellation guards to tie reusable reviewers and temporary forks to their lifetimes, including when a review future is dropped. Replace cancelled reusable reviewers before reuse.
- Share `ReviewerTasks` between the pool and `ThreadManager`, and wait for tracked cleanup during pool shutdown. Remove the separate session shutdown protocol.
- Move denial cleanup from core task handling into Guardian's turn start, stop, and abort hooks. Expose `install_reviewer` to register both thread and turn lifecycle contributors together.
## Testing
Add an integration test that exercises stale denial cleanup at turn start, completion, and interruption, including starting another turn after interruption.
GitOrigin-RevId: 85badb235309c1e1576547d1746f7ee9fc66d545
## What changed
Extract target validation, runtime reloading, and message delivery from the V2 tool handler into `AgentControl::deliver_message`. Keep target resolution, analytics, and tool-facing error mapping in the handler.
Represent plaintext and encrypted payloads with `AgentMessage`, sharing communication construction with agent spawning. Preserve queue-only and follow-up turn semantics, target checks before reload, and turn metadata propagation.
GitOrigin-RevId: a88bcede015a17f6c11c7cad95ee58add5dc2820
## What changed
Move shared spawn and resume configuration helpers into `agent::child_config` and route both multi-agent versions through `prepare_agent_spawn_config`.
Preserve each version's role and model precedence, full-history inheritance, service-tier selection, runtime policy, and default-role metadata for reloads. Return configuration errors as strings and convert them to `FunctionCallError::RespondToModel` at tool-handler boundaries.
GitOrigin-RevId: 9d539050aa44f94043a6487f1f6f1037de1d15ca
## What changed
Add an optional `elicitation_type` to app-used and MCP tool-call events, with `auth_or_link` and `approval` classifications and `null` for unclassified calls. Existing core app-use tracking passes `None`.
Expose an API to queue MCP classifications by thread, turn, and item before completion. Preserve them across turn completion, consume them on item completion, and clear them on thread closure. Bound pending classifications to 256 entries, evicting the oldest when full.
## Testing
Add coverage for serialization, first-classification retention during app-use deduplication, per-call classification across turn completion, pending-state eviction and cleanup, and exclusion of tool arguments from emitted events.
GitOrigin-RevId: 185a2ccb38153e123fc731231edff0259a26bce3
## Why
The TUI renders supported inline math as Unicode but leaves display equations to ordinary Markdown rendering. Display fractions need spatial layout, and streamed equations must remain mutable while their closing delimiter is pending.
## What changed
- Render supported standalone `$$...$$` and `\[...\]` equations as Unicode, with stacked fractions and aligned baselines.
- Fall back to source for unsupported expressions, nested fractions, and spatial layouts that exceed the available width or occur inside lists or block quotes.
- Preview unfinished display math as wrapped source and hold it out of scrollback within the conversion budget, then render it when closed.
- Preserve display delimiter boundaries across incremental rendering and bound lookahead for rejected openers so shell `echo $$` does not retain the entire response.
## Testing
Add snapshots and regression tests for display layouts, narrow widths, literal fallback, streaming closure, resize reflow, raw source preservation, shell dollars, and bounded incremental rendering.
GitOrigin-RevId: ab54675f2e27a243f5d7165c76d9d7355f9fabd1
## Why
`slow_down` errors were treated as terminal server overloads, while exhausted credit balances and spend limits fell through to retryable stream errors.
## What changed
- Classify `slow_down` as a retryable rate limit in HTTP 503 responses and SSE failures, preserving the server message and parsing retry delays from SSE error messages.
- Map `credit_balance_exhausted`, `organization_spend_limit_exceeded`, and `project_spend_limit_exceeded` SSE errors to quota exhaustion so they terminate without retries.
## Testing
Extend tests to cover HTTP error classification and retryability, `slow_down` stream retry exhaustion and message-provided delays, and a single `UsageLimitExceeded` error with no retries for each quota error code.
GitOrigin-RevId: 7985eba71ba4a6fe897a45957e448fadd42f3963
## What changed
Add `codex app-server daemon update --from-cli` to copy and pin the invoking CLI's complete package, including downgrades and local builds. Require interactive confirmation or `--yes`, and warn that restarting a running daemon may interrupt work.
Validate and stage the package before stopping the daemon, recheck the selection and running state after confirmation, and restart only a previously running daemon. Migrate legacy daemon installations to the dedicated package directory while preserving the standalone CLI selection. Explicit replacements disable automatic package updates.
## Testing
Add coverage for cancellation, packages changed during confirmation, version pinning, and legacy migration. Extend CLI integration tests to verify noninteractive confirmation requirements, rejection of incomplete packages without stopping the daemon, and successful replacement and restart.
GitOrigin-RevId: 2d971d6f0259aa576a15a8b115ca9d0e89060fdf
## What changed
Copy the source thread's current attachments when creating a non-ephemeral fork, including forks at an earlier turn. Copies receive new attachment IDs and creation timestamps while preserving resource identities and payloads. Attachment membership can then change independently on either thread; referenced resources are not copied.
Await the atomic copy before publishing the fork. If copying fails, log the error and allow the conversation fork to succeed without attachments. Resuming a fork does not copy attachments again.
Document that clients should use `forkedFromId` on `thread/started` and fetch the fork's attachments with `thread/attachment/list`; copying does not emit per-attachment updates.
## Testing
Add coverage for atomic rollback, independent attachment membership, inheritance across history cutoffs, successful conversation forks after copy failures, and resuming without restoring removed attachments.
GitOrigin-RevId: 939d2c6a3073ccfcd922380c6beda45d3deac606
## Why
Registering the service-bearing package can restart the provisioning service once for each managed sandbox account. A restart can interrupt the response after runtime readiness has been revoked, preventing the refresh from completing.
## What changed
- Retry registered sandbox refresh requests after response pipe disconnections, allowing up to two service restarts within the original deadline. Require a different service process ID and authenticate it against the running service before resending.
- Allow refresh to resume without `ready_package` when account ownership is complete, teardown is not underway, and the existing owner, package family, live account SIDs, and settings checks pass. Runtime execution still requires readiness for the current package.
- Keep authentication failures, failed writes, explicit replies, and protocol errors outside the retry path.
## Testing
Add regression tests for disconnect classification, the two-restart budget, deadline expiry, and authentication failure before writing. Extend ownership tests to distinguish resumable registration from runtime readiness and reject incomplete accounts or retirement state.
GitOrigin-RevId: d932f179b31752fe918297b7b12e5077c17b19e5
## Why
Daemon lifecycle commands previously required a standalone managed installation. A complete CLI package can supply the daemon executable and helpers without requiring a separate installer run.
## What changed
- Let `codex app-server daemon start`, `restart`, and `bootstrap` copy the invoking package into `CODEX_HOME/packages/app-server-daemon` when no daemon installation exists.
- Validate the package's platform, required helpers, executable identity, and copied contents before selecting the staged release.
- Preserve existing dedicated and legacy daemon selections, reject broken selections instead of replacing them, and leave the CLI package and selection unchanged.
- Preserve standalone release pins and latest-channel eligibility, and select dedicated releases using Unix symlinks or Windows junctions.
## Testing
Add package preparation tests for complete copies, incomplete packages, broken selections, legacy preservation, and update-channel handling. Add Windows junction creation and retargeting coverage, plus CLI integration tests that launch the copied package through `start`, `restart`, and `bootstrap`.
GitOrigin-RevId: abd2f4f82eae0885434ea30603c2c1c0ca760bef
## What changed
Replace `AttachmentStore::persist` with `upload` and `resolve`. Uploads return inline bytes or a file ID; resolution returns optional file metadata and a download URL only when a minimum URL lifetime is requested. Add image dimensions, digest, size, MIME type, and categorized errors to the API.
Make `InlineAttachmentStore` return the original bytes and report `NotFound` when resolving file IDs. Redact attachment bytes and file URLs in debug output.
Pass the configured image store from the thread manager into sessions and inherit it in delegated sessions. Allow `TestCodexBuilder` to accept a custom image store.
## Testing
Update unit coverage for preserving PNG and JPEG bytes, redacting bytes and URLs in debug output, and rejecting file resolution in the inline store.
GitOrigin-RevId: f313e0048925998d84394c7e5a278e0d39d86a45
## What changed
Switch the SDK workflow to `prepare-bazel-ci` with the `sdk` cache scope to restore the Bazel repository cache using a dedicated namespace. Use the action's cache key and hit outputs when saving, and save only after a successful Bazel build when no exact cache hit was found.
GitOrigin-RevId: 15bad8ccf1a45685e0ab0191d1aa514b4ccb06bf
## What changed
- Capture `CODEX_WINDOWS_REGISTERED_CORE=1` at startup and propagate the selected runtime to sandbox wrappers.
- Launch registered runners through service-recorded execution aliases, validating ownership, OS package identity, and the staged runner image before sending commands. Require service provisioning without falling back to copied helpers or `PATH` lookup.
- Preserve package context for sandboxed child processes and descendants so they can launch executables from the protected package directory.
- Refresh stale package registrations during readiness checks without blocking unrelated RPCs, and reconcile effective proxy settings through the service.
- Resolve setup ownership from the process token and avoid treating the server directory as a writable workspace when setup omits `cwd`.
- Record startup and command outcomes by runtime.
## Testing
Add coverage for runtime selection, package query validation, helper resolution, setup roots, proxy reconciliation, and command result reporting. Add an ignored integration test for environment forwarding, output, and exit status that requires an installed test MSIX and service provisioning in a Windows VM.
GitOrigin-RevId: 976d64039d611be4406c3d0e354820775f8eb6e4
## Why
Interrupted or failed turns can discard buffered answer or plan source, leaving incomplete math content missing from the transcript or unable to reflow correctly.
## What changed
Flush and consolidate both answer and plan streams before `finalize_turn` clears streaming state. Reuse the same helper for normal completion, preserving received source for terminal resizing and raw output mode changes.
## Testing
Add regression coverage for interrupted answers and plans, failed plans, and policy errors with unfinished math blocks, with and without a trailing newline. Verify source preservation and stable content across width and raw output mode changes, with narrow transcript snapshots. Update the usage activity test to verify that history insertion waits for stream consolidation.
GitOrigin-RevId: 1ae0c192e2fee00dc0b5ee2ac07f3d5f2313ad1b
## Why
Seatbelt ignored Unix socket permissions in `ManagedNetworkSandboxContext`, which could omit allowed sockets or inherit a live proxy's broader permissions.
## What changed
Use the prepared context's `allow_unix_sockets` and `dangerously_allow_all_unix_sockets` settings when present, falling back to the live proxy only when no prepared context exists. Normalize allowlisted paths and preserve explicit extra socket allowances.
## Testing
Add regression coverage for prepared-policy precedence, empty allowlists, invalid relative paths, and explicit extra allowances. Extend macOS exec-server tests to cover prepared and executor-local proxy socket permissions, including explicit allow-all behavior without unrestricted network access.
GitOrigin-RevId: 3372e1f2f1b161e60bde585db8d70b4e3a0b1854
## Why
The daemon shares the standalone CLI package, coupling daemon updates to the visible CLI installation. A dedicated package lets daemon updates preserve the CLI selection and shell profile.
## What changed
- Add dedicated daemon package selection and separate daemon and updater PID files, with diagnostics for the new files.
- Make `codex app-server daemon update` migrate eligible legacy stable latest-channel installations to a dedicated package. Prepare and validate the release and its package-ownership support before switching; preserve settings and whether the daemon was running. Scheduled updates keep legacy installations in place.
- Add daemon-only and deferred-selection modes to the shell and PowerShell installers, leaving the visible CLI installation unchanged.
- Coordinate migration with installer locks, refresh package selection after acquiring lifecycle locks, and reject package changes during update downloads.
## Testing
Add coverage for running and stopped migrations, incompatible installers and releases, package-selection races, daemon-only updates, and a capability probe that does not start an updater. Add a shell installer regression test verifying that daemon installation preserves the visible CLI and shell profile.
GitOrigin-RevId: 1ec0f358f8f6ebde7daddb74e880cc8d2235d647
Update the image generation tool guidelines to avoid printing full results
or base64 image data with `text()` or `notify()`. Recommend printing only
small metadata when needed.
GitOrigin-RevId: f62b67b384dd39a1cc1adad0956b6fcf69d09c72
## What changed
Represent images in `ContentItem` and `FunctionCallOutputContentItem` with `ImageReference::Inline`, flattened to preserve the existing `image_url` wire format. Update image producers and consumers and regenerate app-server schemas and SDK artifacts.
Preserve the Python SDK's `InputImageContentItem` and `InputImageFunctionCallOutputContentItem` class names during generation.
## Testing
Add a regression test for stable Python image class names and adapt existing image tests to the shared representation.
GitOrigin-RevId: c38a780ac3314c2ac2deb3afc1b93b94b6f93fec
## What changed
- Add an explicit `registered_core` provisioning mode that registers the app package for both managed sandbox accounts and records their runner aliases. Authenticate callers against the service's package family and installed version.
- Persist runtime ownership and publish readiness only after both registrations and AppData metadata grants succeed. Prevent legacy setup or another owner from replacing registered sandbox accounts, and restrict `refresh_only` requests to an existing ready setup.
- Scope service and pipe names by package identity.
- Preserve registrations across package updates and clean them up when the owner uninstalls the app. Fence provisioning during teardown and defer package removal until native cleanup succeeds and the service exits.
## Testing
Add tests for registration opt-in, ownership and readiness checks, legacy setup isolation, metadata ACL preservation, listener shutdown handling, and the cleanup finalizer's commit barrier and UTF-8 input.
GitOrigin-RevId: 69e52f7463285b51f3bef831f071245f0eca0838
## Why
Guardian reviews must stop when their parent shuts down or their history is reset. Shutdown must also finish reviewer cleanup before closing the parent's persistent history, including when a review is waiting to retry after a rate limit.
## What changed
- Let the Guardian extension own reviewer startup, prewarming, and shutdown through `ThreadManager`, tracking and joining outstanding work during teardown.
- Cancel reviews on history reset or parent shutdown and reject decisions returned after cancellation, including cached extension decisions.
- Keep background prewarm previews from overwriting the active turn's model metadata.
## Testing
Extend regression coverage to verify reviewer cleanup after parent shutdown, prompt shutdown during a 60-second Guardian retry with the network request denied, and preservation of active model review requirements during prewarming.
GitOrigin-RevId: 05efc369b9a1642a10365eac8a09da77f3f6e28c
## What changed
Add `tool_event_type` to tool analytics events, using exact call ID evidence to distinguish `model_tool_call` from `inner_tool_call`. Serialize `null` when evidence is missing or conflicting at emission time; later evidence does not revise emitted events.
Classify sampled code-mode `exec` and `wait` calls as model calls and dispatched child calls as inner calls, without inferring origin from cell associations or parent lineage.
## Testing
Add coverage for exact and ambiguous origin evidence and late sampling evidence. Extend serialization, code-mode, collaborator, and subagent tests to check the classification.
GitOrigin-RevId: a41085b4665b28b20026aa2bcb10ff7fa8af8778
## Why
Linux proxy-routed sandboxing denied standalone Unix sockets even when the effective network policy enabled `dangerously_allow_all_unix_sockets`.
## What changed
- Carry Unix socket permissions in `ManagedNetworkSandboxContext` and pass the prepared context through Linux sandbox launches with `--managed-network`.
- Allow `AF_UNIX` socket creation in proxy-routed mode when `dangerously_allow_all_unix_sockets` is enabled, while preserving network namespace isolation and restrictions on other socket families.
- Keep standalone Unix sockets denied by default and for path-only grants. Default missing fields in older serialized contexts to restrictive values.
## Testing
Add coverage for policy preparation and transport, legacy deserialization, and malformed policy rejection. Add a Linux integration test covering default denial, path-only denial, and explicit allow-all access, while checking that direct TCP access and `AF_NETLINK`/`AF_VSOCK` sockets remain blocked.
GitOrigin-RevId: 2695b945ad3e59fcb3faf7662d852a26650af16c
## What changed
- Share process package-family queries, token user SID extraction, and account-name lookup through `codex_windows_sandbox`, and use them in the provisioning service.
- Bound token query sizes and validate SID pointers, revisions, and lengths before copying SIDs into owned storage.
- Use a drop guard to balance firewall COM initialization, and track the package lifecycle directory guard separately so cleanup releases it while keeping ancestor and home handles pinned.
## Testing
Add tests that verify copied SIDs outlive their query buffers and reject truncated token data, malformed SIDs, and invalid SID pointers.
GitOrigin-RevId: 2b893e4524d3e00cb114df05215beef7b9eebbe9
## What changed
- Add experimental `account/read.workspaceRouting` metadata containing the selected ChatGPT workspace ID, resolved HTTPS backend origin, and routing override (`us`, `us_cr`, or `NO_CONSTRAINT`).
- Discover and cache routing through `accounts/check` for saved logins, new logins, and workspace switches. Return `null` for signed-out accounts, API-only accounts, and saved credentials without a selected workspace.
- Validate discovered origins against required `chatgpt_base_url` origins. Return errors for failed or malformed discovery and retry on later reads.
- Wait for requirements and routing before publishing account updates, including to newly initialized connections. Clear routing on logout, discard stale discovery results, and guard queued notifications against account changes.
## Testing
Add unit and integration coverage for origin resolution and validation, discovery failures and retries, startup discovery, workspace switching, logout, configuration changes during discovery, and authentication changes while notifications wait for queue capacity.
GitOrigin-RevId: 2c5367bb01dd7543b08d374f44444323e40f1981
Sort top-level artifact files by descending size before dispatching them to
`xargs -P2`, so both compression workers stay busy.
GitOrigin-RevId: a459af0a0c50ead1aaf2a7c19563d8100cd63f0e
## Why
The Windows release workflow builds an extra package archive only to extract it again when staging the Python runtime wheel. Reuse the package directory to avoid this round trip.
## What changed
- Allow `stage-runtime` to accept a Codex package directory as well as a `.tar.gz` archive, with the same package layout validation.
- Reject overlapping source and staging directories, symlinks, and non-regular directory entries before staging.
- Stage Windows runtime wheels from the existing package directory and retain the check that voice resources are absent.
## Testing
Add coverage for matching directory and archive output, including file permissions, source preservation, invalid layouts, non-regular entries, overlapping directories, and CLI handling of both source formats.
GitOrigin-RevId: bf6d1e05888367d482d8d51a4a8061f92a9cae8b
## What changed
- Report `windows_mxc` from native MXC availability on Windows.
- Allow MXC TTY launches and managed networking, using dedicated proxy listeners without requiring a shared-ingress restricting SID.
- Reject MXC custom `argv0` and private-desktop launches, and continue failing closed when native MXC is unavailable.
## Testing
Extend the Windows remote sandbox process-write test to cover MXC with both pipes and ConPTY. Retain coverage for rejecting MXC requests when native support is unavailable.
GitOrigin-RevId: 80c5f319b9066d06b26f0a7eb7119a109e6ebef7
## What changed
Replace `ReviewerSessionFactory` with a startup callback installed through
`ReviewerPool::new`. Review requests supply shared setup data and reuse context,
while the pool uses its callback to create both reusable and forked reviewers.
Update Guardian session setup, prewarming, and existing test fixtures to use the
new pool API.
GitOrigin-RevId: fa171503afcee8bdbdf6822573bea8ded50ce117
Add `base64` and `serde_json` as workspace dependencies of
`codex-windows-sandbox-service` and update `Cargo.lock`. Temporarily exclude
both dependencies from `cargo-shear` checks until they are used.
GitOrigin-RevId: f77a0ba13cfa9ba2315dcf00a80163afff32d0b4
## Why
Resuming a thread initialized its collaboration mode to Default, losing the saved Plan mode and its developer instructions. Reconnecting clients also lacked a server-reported mode to reconcile changes made by another client.
## What changed
- Restore the saved collaboration mode from the latest matching `ThreadSettingsApplied` event, falling back to the last legacy `TurnContext`. Apply the effective model and reasoning effort while retaining the saved mode and developer instructions.
- Include `collaborationMode` in `thread/resume` responses and update the generated schemas and bindings.
- Use the restored mode in the TUI, including the first prompt after resume. Prefer the server's mode when restoring disconnected input, while preserving the local selection for older servers that omit it.
## Testing
Add regression coverage for persisted and legacy collaboration modes, model and reasoning-effort overrides, the resumed Plan mode display and first prompt, and reconnect behavior with and without a server-reported mode.
GitOrigin-RevId: ed064516e7fae1c1668152ab448f510cbcacfe06
## Why
Inline delegates have no entry in the thread registry, so reviewer creation cannot depend on looking up the parent or waiting for its thread-ready notification.
## What changed
- Capture parent identity, authentication, shared agent control, originator, and inherited instructions in `StartThreadOptions` so `ThreadManager` can start a child without a registered parent.
- Route Guardian reviewer creation through this path, remove the standalone fallback and readiness gate, and require a Guardian extension host.
- Install explicit reviewer hosts in unit tests and the Guardian reviewer extension in the integration test harness, using `ExtensionRegistry::to_builder()` to preserve existing contributors.
## Testing
Extend the thread-manager regression test to remove the parent from the registry before starting a child, then verify inherited lineage, originator, session identity, and authentication, and exclusion from the public thread list.
GitOrigin-RevId: 468ded6fdce7520cb39d55c1a884dcfa5aaea2d9
The provider requirement change test asserts that no traffic reaches the
replacement provider. Start a dedicated server with
`MockServer::builder().start()` and include unexpected request methods and
URL paths in assertion failures.
GitOrigin-RevId: 4020c43c0ec0472fe54d4003bff120349ddd5597
## What changed
Add `auto_review.experimental_policy_template` to override the Guardian prompt template in `config.toml`. Trim the configured value and ignore it when empty. Prefer the override over the model catalog template, retaining the bundled template as the final fallback.
The template's `{{ tenant_policy_config }}` placeholder is replaced with the resolved Guardian policy.
## Testing
Extend tests to cover template deserialization, trimming, precedence over the catalog template, and rendered policy text in Guardian inference requests.
GitOrigin-RevId: 85b4a8fc193a42735354894203ccbd1f738a3b58
## Why
The analytics test helper returned the first analytics request, which could contain unrelated events. Plugin-install assertions need to select the expected event even when it arrives in a later request or shares a batch with other events.
## What changed
Update `wait_for_plugin_analytics_payload` to accept an event type, scan requests until matching events appear, and return the payload with only those events. Update callers to request `codex_plugin_installed` or `codex_plugin_install_failed` as appropriate.
## Testing
Add a regression test that sends an unrelated analytics request followed by a mixed batch and verifies that the helper returns only the expected plugin-install failure event.
GitOrigin-RevId: 41290c63ea1c2a3da1efa78b81222b8da76bcbff
## What changed
Add experimental `thread/start.daybreakEnabled` so clients can set the initial preference for persistent threads. Omitted or null values leave it unset; explicit values are rejected for ephemeral threads.
Return the choice in the start response, `thread/started`, and reads before persistence. Stage it with the initial thread metadata and save it when the thread is persisted. Later changes still use `thread/metadata/update`. The preference does not select `turn/start.cyberAccessProgram` or grant access.
## Testing
Add coverage for true, false, and unset values in responses, notifications, reads, and reads after persistence and restart, plus rejection for ephemeral threads. Update existing metadata and access-program tests to exercise threads with an initial preference.
GitOrigin-RevId: 3bff3dc55a18436067bc2a3f156f5abf52d7321b
## Why
Building MCP search entries eagerly cloned tool specs and normalized schemas even for tools that were never selected.
## What changed
Store search specs in `Arc<ToolSpec>` and let MCP search entries share the handler's spec. Materialize and normalize loadable specs only for selected results, preserving existing result formatting and dynamic-tool cache equality behavior.
## Testing
Add coverage for function, freeform, and namespace specs that verifies shared specs produce equivalent results, retain the source while needed, and release it when the search entry is dropped.
GitOrigin-RevId: 4260f2e2527834d8a856b1528654c6951aab8a7d
## Why
Persisting user input received during an active turn currently blocks the next model request. Stores that support background persistence can overlap this checkpoint with inference.
## What changed
- Add `PersistContext::SteeredUserInput` and `allows_background_persistence()` so stores may enqueue these checkpoints, with durability and error reporting enforced by later flush or shutdown operations.
- Use the new context for accepted steered user input and apply the same metadata handling as turn-start persistence.
- Keep tool outputs synchronous, including in mixed input batches, and allow stores to retain synchronous persistence for all contexts.
## Testing
Add gated-store integration tests covering background user-input persistence, synchronous stores, and synchronous tool-output checkpoints. Verify that the next request includes the steered input and waits for persistence when required.
GitOrigin-RevId: c60b7b6c9b483245fd3169306bcf0de248ccdf35
## What changed
- Add spans for one-shot and resumable `exec_command`, `write_stdin`, session creation, and output collection, recording outcomes and output collection stop reasons.
- Correlate calls with conversations, turns, and processes; link stdin interactions to the original exec call and process start requests to executor process IDs. Omit empty turn and call IDs and those longer than 256 bytes.
- Propagate the current tracing span into the spawned one-shot execution task and distinguish timeouts, cancellations, and failures.
GitOrigin-RevId: 722728dc3f5b624e7a4da69fc867c3c672465af0
## Why
Retaining the pseudoconsole's creation pipe handles prevents output readers from seeing EOF while the session remains alive.
## What changed
Drop the creation handles after a successful process spawn and call `ReleasePseudoConsole` when available on Windows 11 24H2 or newer. This lets output close after the last attached client exits while preserving I/O for surviving console descendants. Older Windows versions retain the `ClosePseudoConsole` cleanup path on drop.
## Testing
Add Windows lifecycle tests for output closure after normal exit and termination while retaining the session, plus continued input and output for a surviving console child. These tests skip when `ReleasePseudoConsole` is unavailable.
GitOrigin-RevId: fb1094fb2a570e6fec0cc80d6356f5d7eb1edcbf
## What changed
- Add `NetworkPolicyController` and `NetworkPolicy` APIs for publishing destination policies, checking access, and observing policy changes. Restricted policies permit only `https` and `wss` URLs with exact allowed hosts.
- Add revocable `NetworkPermit` values with cancellation support. Reject stale policy publications, revoke permits when access is removed, and prevent account-bound policy handles from regaining access after invalidation.
- Provide a permit API for SDK transports without destination enforcement that denies access under restricted policies.
- Let `HttpClientFactory` carry a network policy, defaulting to unmanaged access, and include policy identity in factory equality.
## Testing
Add tests for secure host matching, invalidation, stale publication, recovery after policy load failure, account isolation, and SDK permit revocation. Make accepted sockets blocking in redirect test servers to handle macOS socket inheritance, and add a missing read timeout.
GitOrigin-RevId: c79231527bc5e6954cda6581c0b9f43c90335755
## Why
Callers need to tie isolated threads to an explicit lifetime and wait for cleanup even when startup is cancelled or its result is never received.
## What changed
- Add `ThreadManager::start_thread_until` to run an isolated thread until a caller-provided future completes or the thread exits, with a `TaskTracker` covering cleanup and deregistration.
- Retain persistence and session resources across interrupted startup so cleanup can release partially initialized resources or shut down a running session using normal history rules.
- Reject resumed history and startup without explicit session isolation.
## Testing
Add integration tests for cancellation during stalled required MCP initialization, dropping an unconsumed startup result, and owner cancellation that preserves history and parent usability. Also verify that failed duplicate startup leaves the existing thread's writer intact.
GitOrigin-RevId: 47251c1b0c43821ab8f950816c6d341d39cdc823
## What changed
Convert supported expressions inside `$...$` and `\(...\)` to readable Unicode in regular and streaming Markdown. For example, `$\alpha^2 + \beta_{10}$` renders as `α² + β₁₀`. Support a bounded TeX subset including symbols, superscripts, subscripts, square roots, and parenthesized fractions, with verbatim fallback for unsupported expressions.
Exclude code, links, HTML, and display equations from conversion, and avoid interpreting common currency and shell syntax as math. Preserve file citation paths and keep streaming cache boundaries outside display equations.
## Testing
Add snapshots for inline rendering and narrow wrapping, plus regression tests for unsupported input, parser limits, Markdown contexts, file citations, and streaming display boundaries across chunks and widths.
GitOrigin-RevId: 828bf7b48345b05bc6c2fb5cb2798412c3b71dbe
## Why
Remote app servers cannot read image paths on the TUI host. Image attachments need to carry their contents when submitting to a remote workspace.
## What changed
- Prepare local images as data URLs off the event loop for new turns and steers, preserving source pixels for model-specific resizing and enforcing a 32 MiB image transport budget.
- Show preparation progress, preserve queued message order, and recover drafts on preparation failures, cancellation, disconnects, or thread switches. Ignore stale preparation completions.
- Avoid duplicate user messages from server receipts and duplicate image labels in history.
## Testing
Add regression tests for portable image contents, receipt deduplication, draft restoration, responsive input during preparation, and cancellation across thread and connection changes.
GitOrigin-RevId: 6404615eae0b06c141af96b4229501c61f7fcfee
## What changed
Add an `instructions.load` tracing span with `provider = "global"` around loading user instructions from the Codex home directory. Skip recording function arguments and add the `tracing` dependency to `codex-home`.
GitOrigin-RevId: 1b11e4b18f7a0dbee742a092aba580a184dc5fbf
## Why
Configuration requirements did not report which login methods the running app server permits after applying managed policy, forced login settings, and workspace restrictions.
## What changed
- Add `allowedLoginMethods` to `configRequirements/read`, using the running authentication manager's effective policy rather than newly read authentication settings.
- Return requirements when login methods are restricted even without managed requirements, while preserving `requirements: null` for the unrestricted default.
- Update protocol schemas and generated TypeScript and Python types. An empty list permits no login method; older servers may omit the field.
## Testing
Add coverage for managed and forced login restrictions, workspace intersections, policy reporting after requirements files change, invalid login methods, and API-only Amazon Bedrock without ChatGPT requests. Extend tests for conflicting authentication requirements and cloud policy precedence.
GitOrigin-RevId: 56c0767a74143e793aac2ac165d0cbe98a09469b
## What changed
Restrict `run_before_review_deadline_with_cancel` and its re-export to `codex-guardian-reviewer`. Move its timeout, abort, and successful-completion tests from core into the reviewer's deadline module, and remove the standalone `run_before_review_deadline` tests from core.
GitOrigin-RevId: dd9f1ed571a40a4bd66b08c88f3ee2be071f4870
## What changed
Extract tool observation and evidence capture into `observation.rs`, background classification into `classification.rs`, and score tracking and failure handling into `score.rs`. Keep lifecycle hooks in `extension.rs` and pass captured evidence through a `Classification` struct, preserving the existing snapshot and background task boundaries.
## Testing
Move the fail-closed score-ordering test into `score_tests.rs` and extend it to verify that a failed sample replaces an equally dated score while preserving newer scores.
GitOrigin-RevId: b25b9e828cce78fb2be522d7209346b63403d824
## What changed
- Remove `AgentSpawner` and `AgentSpawnFuture` from the extension API, along with the Guardian wrapper, thread lifecycle context, and app-server injection plumbing.
- Define `InternalSessionSpawnFuture` directly as a boxed future instead of aliasing `AgentSpawnFuture`.
- Raise the workspace `rustls` minimum version to `0.23.45`.
GitOrigin-RevId: b7319dee41bfb869479afeb7555a6f050c4d00a5
## What changed
- Update `rustls` from 0.23.36 to 0.23.45 and `rustls-webpki` from 0.103.13 to 0.103.15.
- Update `aws-lc-rs` from 1.16.2 to 1.18.1 and `aws-lc-sys` from 0.39.0 to 0.45.0, including its new `pkg-config` dependency.
- Refresh the corresponding dependency metadata in `MODULE.bazel.lock`.
GitOrigin-RevId: 7b37b3cea5485c0b5bd11d229b2949aec399afd9
## Why
Session initialization can be cancelled after a persistence writer is installed but before acquisition returns. Cleanup must wait for acquisition to finish so it can discard the writer.
## What changed
Extend `LiveThreadInitGuard` to own in-flight acquisition and finish it before discarding persistence, including when the guard is dropped. Use the guard for thread creation, resume, and inherited model context initialization.
## Testing
Add a regression test that cancels acquisition after writer installation, verifies cleanup waits for handoff, and confirms the writer is removed afterward.
GitOrigin-RevId: 09041fa5cd675d082a86a4cc8a719c572afe11a8
