## What changed
Add an optional `thread_id` to `HostedFileUploadContext` and include it as `codex_thread_id` in file creation requests when provided. Keep MCP app uploads without this field while retaining the thread ID in tool-call metadata.
## Testing
Extend upload tests to check inclusion and omission of `codex_thread_id`, and assert that MCP app calls retain `_meta.threadId` without adding it to file uploads.
GitOrigin-RevId: cff15b35e5130e56973f944b3b849807550d298b
## What changed
Add a default `ModelRequestContributor::budget_exhausted` hook and fail generation with `SessionBudgetExceeded` when any contributor reports an exhausted budget for the thread.
Check budgets before HTTP requests and before WebSocket connection setup, then check again after the connection is established. WebSocket warmup remains available when the budget is exhausted.
## Testing
Add coverage for budget changes during a tool call, verifying that exhaustion prevents the next model request while an available budget allows completion. Extend WebSocket coverage to verify that exhausted generation opens no connection and warmup still permits connection reuse once the budget becomes available.
GitOrigin-RevId: 97ce02dc7f9bd54830e4b73c4f0334a30367eebf
## Why
Item reads with an update boundary previously rejected forked thread histories and only supported an exclusive lower bound on update ordinals.
## What changed
- Add `before_updated_at_ordinal` as an exclusive upper bound, mutually exclusive with `after_updated_at_ordinal`.
- Allow update-filtered and update-ordered item reads across parent and child segments, applying segment bounds and preserving cursor pagination in both directions.
- Require an update boundary for update-order sorting and reject either boundary when using an item anchor or item-ID filtering.
## Testing
Extend thread-store tests to cover upper-bound filtering, descending cursor pagination, forward and backward paging across forked lineages, mutually exclusive boundaries, and ordinals outside SQLite's integer range.
GitOrigin-RevId: 63f7e18fb24a89a41ff8e570cf65963622519a25
## What changed
Move proxy configuration and routing selection into `prepare_managed_network`. Keep native Windows launches direct by stripping managed proxy variables before credential brokering, and select dedicated listeners only for Windows MXC launches.
Consolidate managed network context filtering in `SandboxManager` and simplify proxy preparation for Linux and MXC while preserving the macOS Seatbelt context.
## Testing
Add Windows regression coverage for native launches with both disabled and enabled proxy configurations, verifying that managed proxy variables are removed while caller environment values and API credentials are preserved.
GitOrigin-RevId: 29fe1fb02afb574b4108032b7e8028d19d8270b4
## Why
Incoming delegated messages were rendered as plain text, leaving Markdown formatting visible and links without hyperlink metadata.
## What changed
Use a dedicated history cell to render the message body as Markdown while keeping the source task attribution literal. Preserve hyperlinks and wrapping, and apply the appropriate list spacing for scrollback and the redrawable transcript.
Use the same rendering for live messages, resumed history, and transcript projection.
## Testing
Extend replay coverage to check matching live and replayed output, transcript rendering, hyperlink destinations, and shared source text across wrapped rows. Add a snapshot covering list spacing in scrollback and the redrawable transcript.
GitOrigin-RevId: e9864c7721fa1afa2d3662ab9f57122ab696a51e
## What changed
Allow `model_providers.<id>.gateway_oauth.credential_export` to supply a gateway access token instead of browser OAuth. Run the configured command on first use, cache its token in memory, and deliver it through the configured header or cookie. Export failures surface directly, and command-backed providers do not require an OAuth session.
Require mutually exclusive OAuth and exporter settings, same-origin model catalog access, and rejection of HTTP redirects for command-backed gateway credentials. Share the bounded, timeout-controlled command runner with AWS credential export.
## Testing
Add tests for inference and model listing without OAuth, token caching and recovery after invalid output, preservation of primary authentication, and rejection of mixed OAuth/exporter configuration.
GitOrigin-RevId: e3dbfd8823b1992045f1c1efd56ec98ee17e07a5
## Why
A message's model-visible item ID may be unknown until the recipient records it. Message lifecycle observers need to capture publication state before that point.
## What changed
Remove `message_id` from `MessageLifecycleInput` and give each publication's `ExtensionData` a fresh UUID key. Stop assigning an item ID when publishing messages for lifecycle observers.
## Testing
Update the agent mailbox test to store a contributor-specific token in `ExtensionData` at publication and retrieve it at consumption, verifying that publication state survives queued delivery and recipient eviction.
GitOrigin-RevId: e254f94994d0976e61cb46ee23297f3248b9a389
## What changed
- When side retention is enabled, let the sidebar chooser create additional side chats from the main chat or reopen existing ones.
- Add numbered, clickable tabs and keyboard navigation in the chooser. Keep the selected tab visible when space is limited.
- Preserve drafts, transcript history, and the sidebar layout when switching between sibling side chats.
- Make `/side` with a prompt request a new side chat, while `/side` without arguments reopens the selected side chat. Preserve existing reuse behavior when side retention is disabled.
## Testing
Extend TUI tests and snapshots to cover creating sibling side chats, switching through the chooser and tabs, retaining drafts and history, and restoring image attachments after failed creation.
GitOrigin-RevId: 4166010c253c40a409390f7020b69478c04cf98c
## Why
Some hosts assign a worker per connection. Idle connection eviction can prevent a quiet session from reaching the worker holding its lease, while retaining a connection after the last session closes unnecessarily keeps that worker allocated.
## What changed
- Disable the HTTP connection pool idle timeout for gRPC code-mode connections.
- Cache URL-based clients weakly and let live sessions own the shared client, releasing their ownership on closure or failure even while the provider remains alive.
- Acquire clients for deferred RPCs only when needed and return the session error if the client has already been released.
## Testing
Add regression tests that verify connection reuse and stored state after 120 seconds of inactivity, and connection closure after the last shared session fails while the provider and session handles remain alive.
GitOrigin-RevId: bd1ff00c4003c03ed173c23dd2ed373ed354ab65
## What changed
Enable `--addrsig` and `--icf=safe` for `x86_64-unknown-linux-musl` release packages in Cargo and Bazel.
Opt Cargo release builds into these flags with `CODEX_MUSL_RELEASE_LINK_FLAGS=1`, selecting `lld` and preserving existing `RUSTFLAGS`. Apply the corresponding flags in Bazel's release package transition for the x86-64 musl platform.
GitOrigin-RevId: ade3c619dad2f03fd93a6fec35477607caec8b31
## What changed
- Add an “Add custom provider” entry to the provider picker and an “Edit configuration” action for editable providers.
- Provide forms for custom gateway names, base URLs, and API key environment variable names, plus supported Bedrock and GCP settings.
- Validate required fields, suggest environment variable names from the provider catalog, and preserve drafts when validation or saving fails.
- Save configuration through the connected app server, write only changed fields, and refresh the provider list after saving. For uncertain creation results, ask users to check the list before adding the provider again.
## Testing
Add editor tests for validation, draft recovery, Vim input, and patches that preserve unchanged authentication settings. Add an integration test covering the edit action and custom provider creation while preserving the default provider, with snapshots of the new UI.
GitOrigin-RevId: 8eb32ed30ed27ab671f08f565af48705f2b9f193
## What changed
- Add `accepted_at_unix_ms` and `accepted_seq` to conversation items, recording the first local `HistoryItemAccepted` event in trace order. Inherited context alone does not count as local acceptance.
- Add `started_code_cell_id` to model-requested `exec` tool calls when a matching code cell exists in the same thread and, when known, turn. Resolve links after replay so deferred cell starts can be matched.
- Default the new optional fields to absent when deserializing older traces and omit them from serialized output when unset.
## Testing
Extend reducer tests to cover duplicate acceptance events with out-of-order clock readings, inherited items without local acceptance, and `exec` calls with and without matching code cells.
GitOrigin-RevId: dfb594630f77d7c899398aa181c88a681587b1df
## Why
Clients need to list models for a configured provider other than the app-server's startup default.
## What changed
Add an optional `modelProvider` parameter to `model/list` and expose it in the generated schemas and client types. Explicit selections use provider definitions loaded at startup and respect current managed provider requirements. Unknown providers return a structured `modelProviderNotConfigured` error.
Use a temporary catalog when selecting another provider, preserving the default catalog and its disk cache while retaining runtime authentication. Requests without a selector keep the existing provider and caching behavior.
## Testing
Add coverage for provider selection, preservation of the default catalog and disk cache, unknown-provider errors, managed requirement enforcement, and runtime authentication for temporary catalogs.
GitOrigin-RevId: 6092df9f6141810c6ac215a8e1fb601f77e54747
## What changed
Update Seatbelt tests to call `create_seatbelt_command_args` directly with explicit filesystem and network policies, and remove the unused legacy-policy wrapper.
Reuse `FileSystemSandboxPolicy::read_only()` in read-only policy conversion and test setup instead of constructing equivalent policies inline or through legacy conversion.
GitOrigin-RevId: 2e2b3342425a3477986f41135caa18003b697fe4
## What changed
Remove `FileSystemSandboxContext::from_legacy_sandbox_policy` and migrate its test callers to `from_permission_profile`. Construct read-only, disabled, and externally enforced sandbox contexts directly with `PermissionProfile`.
GitOrigin-RevId: e330b89d060061eda106232eab554be5ccd06149
## What changed
- Expose an optional `AgentServiceTier` provider so host-owned children, including detached children, inherit the root's committed tier during spawn preparation, startup prewarm, and each inference step.
- Reject per-turn tier overrides for children using the shared provider. Root updates apply to subsequent requests without changing requests already in flight.
- Preserve independent configured tiers and role overrides when no provider is installed, including across eviction and reload. Spawned children inherit the parent's thread preference rather than its per-turn override.
- Publish the latest committed root tier only after successful runtime registration, preventing failed duplicate starts from changing shared settings.
## Testing
Add coverage for detached-child WebSocket prewarm and between-step tier changes, shared-child override rejection, independent-tier inheritance and reload, and settings updates before registration.
GitOrigin-RevId: f0f3676813a1c59bd77baf95121e2d13e0163a8e
## Why
WebSocket telemetry reparses text frames to extract event kinds that the transport already decodes.
## What changed
Pass decoded event kinds to telemetry and report receive failures separately. Keep JSON parsing for timing metrics and as a fallback when stream event decoding fails. Remove the `tokio-tungstenite` dependency from `codex-otel` and restrict it to tests in `codex-core`.
Limit wrapped error parsing to error events or failed stream decoding, preserving HTTP error forwarding when unrelated stream fields are malformed.
## Testing
Update runtime summary coverage to include receive failures and extend the WebSocket error forwarding test with a malformed stream field.
GitOrigin-RevId: e1f9d200fd6d30ec124a2354bdbc01f3082ddde3
## What changed
- Remove startup prefetching of remote catalogs, featured plugins, and recommendations while retaining installed-plugin refreshes and bundle downloads.
- Fetch featured plugin IDs on demand, concurrently with marketplace fetching when listing the remote global catalog, and reuse the cached results.
- Retire `plugin_discovery_prefetch` as an ignored compatibility flag and remove forced background catalog refreshes.
## Testing
Update app-server tests to verify that startup downloads installed bundles without discovery requests and that two `plugin/list` calls share one featured-plugin fetch.
GitOrigin-RevId: aa10474f3429102d624c83472e81ec85e3b470ae
## Why
Starting a side chat waits for fork and setup requests. Keep the workspace responsive during those requests and let users return to an existing side chat through the sidebar chooser.
## What changed
- Add a clickable `Side chat` tab and `+` chooser, plus the configurable `focus_sidebar_tabs` action bound to `leader t` by default.
- Create workspace side chats in the background, showing loading and failure states with `Enter` to retry and `Esc` to go back. Preserve opening prompts and avoid taking focus when the user has moved away.
- Retain inherited permission profiles and defer startup notifications until the side chat is registered on its owning backend.
- Quiet the parent composer while the chooser has focus and prevent completion popups or repeated activation keys from consuming chooser input.
## Testing
Extend tests and snapshots for loading and failure states, returning to an existing chat, draft preservation, inherited permissions, focus transitions, and deferred MCP startup notifications.
GitOrigin-RevId: 244bcd3f170819e2260f21dc1041cf8899fe048c
## What changed
- Advance the pinned `tokio-tungstenite` and `tungstenite` fork revisions and update the Cargo and Bazel lockfiles.
- Use `OAuthCredentialsStore::File` in the gateway credential test to avoid spending the model discovery deadline on keyring-backed decryption.
GitOrigin-RevId: 2862850215abce478136137b9b68a37f3caaeed4
## What changed
- Expose `RequiredCyberAccessPrograms` so hosts can require an explicit allowed program on a thread. Snapshot the requirement before extension callbacks; hosts must reapply it on cold resume.
- Validate resolved `access_programs` before HTTP and WebSocket model requests, rejecting omitted or disallowed programs. Explicit `Standard` remains distinct from omission. Reject memory summarization and realtime requests when a requirement is present.
- Keep required programs scoped to the parent thread by preventing Guardian reviewers from inheriting its program, including Guardian v2 reviews.
## Testing
Add HTTP and WebSocket coverage for allowed programs, omission, disallowed programs, unsupported providers, and connection reuse. Extend Guardian tests to verify parent-only program scope and that removing the public attachment does not erase the startup snapshot.
GitOrigin-RevId: 9306a9af0e03e7a1728dfeb610cd7ec7517367fd
## What changed
- Configure Bazel to apply `opusic_sys_install_headers.patch` to `opusic-sys` with `-p1`.
- Run the cloud-managed MCP OAuth test with `--no-browser` and keep stdin open until the child exits so EOF cannot race the HTTP callback.
- Serialize the cancelled-refresh test with other auth environment tests and temporarily clear the access-token environment variable.
GitOrigin-RevId: ad76b72ae98320bd0e3ff3231e1ed874332ce38c
## Why
Registry `401` responses during harness key validation previously produced generic relay resets, causing recovery to retry rejected executor credentials instead of surfacing the authentication failure.
## What changed
- Send an authenticated Noise handshake rejection for registry `401` responses without opening an application transport.
- Propagate the rejection as `EnvironmentRegistryAuth` and preserve it in the terminal connection state so recovery stops with a specific authentication error.
- Preserve registry HTTP status codes so `403` responses for stale harness grants and transient service failures can continue through recovery.
## Testing
Add tests that reject tampered authentication rejections and exercise session recovery: `401` stops after one rejected recovery attempt, while `403`, `502`, and `503` retry with fresh harness grants and reconnect successfully.
GitOrigin-RevId: 9a55412e583146c094948618c7c9aeb4fb4b953e
## What changed
- Select fresh authentication for each HTTP operation on eligible OpenAI MCP connections while preserving the connection's credential source and account network policy. Stop redirects when requests carry runtime assertions.
- Pass the owning thread's authentication and runtime context through app discovery, connector metadata, and trusted-access requests. Include the runtime authentication mode in MCP connection and tool-cache identities.
- Give resumed actors analytics clients bound to their own authentication context, and serialize stored-credential reloads with token refresh while preserving host-provided credentials.
## Testing
Add coverage for fresh request assertions, revoked account policies, redirect handling, resumed-actor isolation, and MCP uploads in dual-header and assertion-only modes.
GitOrigin-RevId: f06900a6db61cd6cf53c3aa8075afeec4d5706d9
## What changed
Add `CloudRequestHandle::submit_user_message` to submit a message with a stable ID using `SubmissionMode::Steer`. Return a `CloudUserMessageReceipt` that distinguishes started, steered, and duplicate admissions while preserving server-provided turn status and duplicate identity.
Report preflight rejections and invalid requests as `-32602`, while preserving uncertain admission outcomes as errors without automatic retries. Keep existing `turn/start` callers on `StartIfIdle` with their existing response and error behavior.
## Testing
Add receipt mapping tests and a mock-server worker test covering submission modes, single admission attempts, uncertain outcomes, input validation, and disconnected reply streams.
GitOrigin-RevId: eba120cf47cf0297b072ea373341ead663fb9bc9
## Why
Legacy `.sandbox-bin` ACLs grant `SYSTEM` Modify access without `WRITE_DAC`, preventing the provisioning service from reapplying the bin policy.
## What changed
- Grant `SYSTEM` inheritable `WRITE_DAC` access when protecting the sandbox bin so it can update permissions on the directory and its children.
- Before legacy provisioning, repair the old ACL through an authenticated owner handle when `SYSTEM` is denied access. Upgrade only the expected SYSTEM entry in a protected, allow-only ACL, reject unexpected permissions, and verify that `SYSTEM` can reopen the directory with `WRITE_DAC`.
## Testing
Extend ACL coverage to check `SYSTEM` permissions on the bin and existing and newly created runner files. Add an ignored Windows integration test requiring LocalSystem and user-owned legacy homes to cover successful repair, repeated repair, and rejection of an explicit SYSTEM deny entry.
GitOrigin-RevId: 2303cdf5182947c2d044c79ac7027a16f0373c7d
## Why
Explicit cyber access program selections were dropped for requests using `CodexAuth::Headers`, preventing the server from evaluating the requested program.
## What changed
Forward explicit program selections for header-authenticated requests, leaving entitlement decisions to the server.
## Testing
Add integration tests covering explicit selections and omission over HTTP and a reused WebSocket connection, preservation of actor and account headers, and propagation of backend HTTP 403 denials without dropping the selected program.
GitOrigin-RevId: 5e549e87a1420ce8a5f1ce5c704cc6b7630af3ba
## What changed
- Open a sidebar chooser when no side chat exists, with keyboard and mouse controls to start a new side chat while preserving the parent draft.
- Animate sidebar opening and closing, including reversing an unfinished transition, when animations are enabled.
- Show context-aware workspace shortcuts in a floating panel after a short delay, and render presets in the same overlay without moving the composer.
- Prevent chooser input from accepting an approval dialog before it has been painted.
## Testing
Add coverage for chooser activation and draft preservation, sidebar animation reversal, shortcut and preset snapshots, and approval input handling.
GitOrigin-RevId: 014112622a94798f4e611756c55112dfd8016051
## What changed
Require the `openai` provider ID as well as `GuardianNativeMultiAgentHistory` to enable native agent history and review of multi-agent calls. Apply the same check to Guardian prompt construction, collaboration tool registration, and Guardian v2 configuration.
## Testing
Add coverage for fresh and resumed sessions with the feature enabled and disabled on OpenAI, and enabled on Bedrock and LiteLLM. Verify that only OpenAI sessions with the feature enabled request native review and include the agent call as evidence.
GitOrigin-RevId: bfbe547b0544e66763dca7d1e1376f805994bcb2
## Why
Versions of `ignore` differ in whether parallel walks report traversal errors at the depth boundary, making the test's expectation depend on the walker implementation.
## What changed
Keep the denied directory within the bounded scan depth in `unreadable_glob_traversal_errors_fail_closed`. Require both bounded and unbounded scans to fail with the denied path and permission error, replacing the CPU-count-dependent assertion.
GitOrigin-RevId: 4ed97253247e0c032648ebc6d4be2a97c3ab261a
## What changed
Extend `guardian_review` to assess recorded MCP tool calls using the current tool catalog, bindings, and approval policy without executing the call or remembering an approval. Return no assessment when policy or a remembered approval skips review, and reject tools blocked by app configuration.
## Testing
Add a scenario test that compares direct and live MCP review inputs, verifies that neither dispatches the tool, and checks that the current app approval policy can skip review.
GitOrigin-RevId: 05746dd0804a0f96d7946468701d5222171ff64f
## Why
Multi-agent V2 still supplied native tool usage guidance when native tools were disabled, leaving instructions for unavailable tools in model requests.
## What changed
Skip bundled and model catalog role guidance for both root agents and subagents when `disable_native_tools` is enabled. Preserve explicitly configured usage hints.
## Testing
Add coverage for omitted bundled and catalog guidance and preserved configured hints with native tools disabled. Update the request snapshot and assert that spawned agents do not receive native tool call guidance.
GitOrigin-RevId: 6f2facabc95e78cbf64af19636a4daf488bbcc51
## Why
Idle reads from local MCP servers on Windows can occupy Tokio's blocking threads and delay unrelated filesystem operations.
## What changed
Use IOCP-backed named pipes for local MCP server stdin, stdout, and stderr on Windows. Connect child endpoints before launch, restrict pipe access to the owner and SYSTEM, and reject remote clients. Preserve the shared legacy and modern protocol framing.
## Testing
Add Windows regression tests to verify that idle MCP reads leave blocking I/O available in both protocol modes and that child output handles support querying the pipe mode.
GitOrigin-RevId: ca39b940a54427a2cd7bab5291011b0b09143b3f
## What changed
Expose `AgentTurnOutcomeReporter` for hosts to install through `StartThreadOptions::thread_extension_init`. Await its callback after emitting the terminal turn event, independently of native parent notifications, so hosts can handle result delivery and retries.
Track outstanding outcome handoffs and wait for them during graceful shutdown before running thread-stop hooks. Prevent new turns from starting once handoff registration closes.
## Testing
Add integration coverage for completed and errored outcomes, turn lineage, queued follow-up turns, and shutdown waiting for handoffs. Extend Guardian tests to verify that strict safety stops reach the reporter while default interruptions remain silent.
GitOrigin-RevId: 0d825f6208f0250969734488881f32fd1dcd7509
## What changed
- Add Space to set the default provider from Model providers settings, persisting `model_provider` on the connected server. Keep provider details available and disable selection when server configuration or client launch settings fix the provider.
- Apply the choice to `/new` while preserving the current chat. Refresh account details for new chats and reject account data that does not match the started provider.
- Block stale `/model` choices during saves and after confirmed or uncertain provider changes, including across reconnects. Tell users to restart the app server and TUI to refresh the model list, and discard queued picker actions from another backend or chat.
## Testing
Add coverage for provider saves, retry after disconnection, concurrent-save prevention, stale model choices, and account/provider matching, with updated provider-picker snapshots.
GitOrigin-RevId: af6d149f7de7a2eaa09a31935f7088faa64d6a59
Document that `AgentMessageBoard` notification delivery follows the host's
scheduling and wake policy and may buffer notifications for idle recipients.
Link to `MessageBoardHost` for the delivery contract.
GitOrigin-RevId: fb085fe2c43c74b85b4e6fe471e7d4ae216bf3db
## What changed
Replace external `rg` scans and the fallback walker with an in-process `ignore` walker for bubblewrap deny-glob expansion. Remove scanner lookup and dependency tracking from sandbox integrity checks.
Include hidden and ignored files, respect scan depth limits, and fail closed on traversal errors. Anchor patterns to their static search roots and preserve literal leading `#`, leading `!`, trailing spaces, and unclosed character classes. Follow explicit symlink roots while skipping child symlinks, and mask canonical targets of matched paths.
## Testing
Extend unit and sandbox integration coverage for glob syntax, depth limits, traversal failures, symlink handling, commands launched from `/`, and operation without executing `rg`.
GitOrigin-RevId: 1767d7f37acb149457642bf24ee5971a28696438
## Why
Matches in parent directory names can crowd matching filenames out of limited search results. For example, searching for `ViewModel` should keep `ChatViewModel.kt` visible even when many directories are named `viewmodel`.
## What changed
- Rank basename matches before applying the result limit for single positive fuzzy queries without path separators.
- Prefer exact filenames, then the strongest exact directory match, then other matching filenames. Leave additional directory matches ranked by their paths.
- Compute highlight indices from the basename for promoted results and return scores that reflect the new ranking.
- Rank snapshots in resumable time slices so new queries and cancellation can interrupt the work.
## Testing
Add regression coverage for filename matches surviving the result limit, exact file and directory priority, path queries, and resuming partial rankings. Update app-server score expectations while preserving highlight indices.
GitOrigin-RevId: 41b3420581b4f6490859e809245370b737889544
## Why
A tool can start before an assistant message finishes streaming. The TUI needs to incorporate the completed text without duplicating the reply or leaving Markdown split across interrupted fragments. Some streams also include initial text in `item/started` before sending deltas.
## What changed
- Display initial assistant text from `item/started`, avoiding duplicate starts for the same message.
- Track interrupted replies by turn and item ID, then replace their flushed fragments with the completed message at the reply's original position, preserving intervening tool order.
- Re-render the completed Markdown and update transcript views and scrollback.
## Testing
Extend streaming tests to cover initial text, resumed deltas, Markdown spanning a tool interruption, and matching live and replayed transcripts. Expand completion-repair cases to cover missing text and resumed fragments.
GitOrigin-RevId: 7f35e52d0bca2100baf722d7ef4922cd127bbc45
## Why
Reasoning summaries can arrive with item start or completion events rather than streamed deltas. The TUI can miss or truncate these summaries, and replay after a voice handoff can hide reasoning that started before the handoff.
## What changed
- Initialize reasoning from item start payloads and reconcile live and replayed completions with their full summaries. Preserve streamed text when completion payloads are empty, and continue honoring `show_raw_agent_reasoning`.
- Retain active reasoning payloads across refreshes without resetting an already active item when its start event is replayed.
- Preserve the typed origin of reasoning started before a voice handoff and filter later delegated reasoning starts from buffered replay.
## Testing
Extend regression coverage for summaries supplied at start or completion, partial streamed summaries, active reasoning after refresh, and replay that retains pre-handoff summaries while hiding delegated reasoning even after the handoff marker is evicted.
GitOrigin-RevId: 724b9fc58e10f4d87dabd3d9930e252270855a4f
## What changed
- Map thread listing and creation, paginated turn and item history, and turn submission and interruption to Cloud operations.
- Mark listed and activated threads as using paginated history, and warn when history items cannot be displayed.
- Reject unsupported filters, custom settings, and structured input. Require stable submission IDs and avoid returning a running turn for duplicate or queued submissions.
- Forward optional `turn_trigger` values through the Cloud client.
## Testing
Add coverage for resume-picker pagination, unsupported filters, submission IDs and settings, duplicate and queued admissions, and `turn_trigger` wire encoding.
GitOrigin-RevId: dcbd87d8190786da65e343a3a0f8d849b24a5b26
Give each independent MCP server and session setup its own test identity
and timeout budget by replacing nested loops with `test_case::test_matrix`.
Preserve all 64 configuration combinations and the existing assertions for
direct tool exposure, tool search availability, and advertised tool calls.
GitOrigin-RevId: 74ef56d34b6bab3038b70d7d84183f7d104788a2
Move channel notification formatting into `PostPreview::notification_text()` and use it in the local message-board host. Preserve the notification fields, preview text, and `read_post` hint for truncated posts.
GitOrigin-RevId: fd3596c45194c958817a24f8fc35a886f766a84b
## What changed
Remove `summarize_sandbox_policy` and its public export from `codex-utils-sandbox-summary`, leaving `summarize_permission_profile` as the summary API. Migrate the external-sandbox, read-only, and workspace-write summary tests to permission profiles.
GitOrigin-RevId: 2ea039a2573e05c17edb8356aa7e04e065ef592a
## What changed
Add a `Housekeeper` that uses an isolated reviewer runtime to route captured original evidence into topic channels and update their discovery descriptions. Keep originals intact and batch inputs within the selected model's tool-output budget.
Validate routing responses before writing to the message board. Advance the evidence cursor only after all writes succeed, stop the worker after an error, and return the runtime disposition so callers can manage session reuse.
## Testing
Add integration coverage for routing originals to multiple channels, updating descriptions, preserving capture boundaries, batching within the input budget, and leaving the cursor unchanged on write failure.
GitOrigin-RevId: 670508c7ecf2ddffdd2946f1b25dd7e071318c5a
## Why
Evidence capture previously reloaded the full history without tracking consumer progress. Live consumers need to read new originals from activation onward and acknowledge them after downstream writes succeed.
## What changed
- Add an in-memory `EvidenceCursor` that starts at the current complete tail without backfilling, captures a fixed unread range, and supports partial acknowledgements. Reads leave progress unchanged.
- Add `ThreadStore::read_original_history` to read original records and their source positions through a fixed cutoff. Local storage supports append-only paginated root histories, including compressed rollouts, and defers unfinished records.
- Reject unsupported histories and invalid ranges, and retain source ordinals as evidence IDs.
## Testing
Add coverage for partial records, compressed suffixes, fixed cutoffs, mismatched thread positions, and truncation. Extend evidence tests to cover activation without backfill, partial acknowledgements, later appends, stale captures, and unchanged cursor progress after capture errors.
GitOrigin-RevId: c7a13749fedc4f44fadddc9dd9875043bccebd95
## What changed
Add `EvidenceSnapshot::capture` to read original evidence from `ThreadStore` into an immutable snapshot with root-history ordinals. Preserve accepted user messages, assistant messages, tool activity, verified answers, delivered assistant messages, and host-annotated approvals without changing their contents or provenance.
Exclude generated context and compaction output. Reject forked or reverted histories, inherited or ambiguously attributed user messages, and incomplete retained evidence. Capture currently supports append-only root histories and requires the host to flush pending originals first.
## Testing
Add an integration test covering content preservation, generated-context filtering, approval provenance, isolation from later appends, and rejection of ambiguous user messages.
GitOrigin-RevId: f464a2714c2678bcffe7abf0a8224196bfe48692
## What changed
When `features.guardian_denial_reconsideration` is enabled, retry the first terminal denial once using the unchanged inference request. Buffer provisional output so only the selected response enters reviewer history, while allowing tool calls to proceed normally.
Preserve the original denial if the retry fails or times out, and continue to honor cancellation. Account for token usage from both attempts, including failed reviews.
Remove `auto_review.experimental_denial_reconsideration_prompt` and warn when it remains in configuration.
## Testing
Update the Guardian scenario and snapshot to verify identical retry inputs, retained tool evidence, and reuse of the selected assessment in a subsequent review, with asynchronous compaction enabled.
GitOrigin-RevId: cc4e303d40a02acc824c8b4b954852325e64ad52
lt-oai