Commit Graph
209 Commits
Author SHA1 Message Date
jif 36430b3688 Flush completed Guardian reviews before delivering decisions (#46509)
## Why

The parent could receive a Guardian decision before the review's terminal event was flushed, allowing the reviewed action to start while completion was still being saved.

## What changed

Flush the review transcript and `TurnComplete` event together before delivering completion to the parent. Clear the reviewer's active turn before delivery so the parent can immediately request another review. Skip redundant completion flushes for Guardian reviewers while retaining the existing pre-completion flush on errors and cancellation.

## Testing

Add a gated thread-store regression test covering two consecutive reviews in the same reviewer thread. Verify that each save includes `TurnComplete` and that the approved command cannot start while the save is blocked.

GitOrigin-RevId: 9274b08ddba1ea93d4fb4709bbfdc2ccaea866ab
2026-09-18 23:28:13 +00:00
alexsong-oai 4d23af0975 Compose gateway OAuth with primary provider authentication (#46490)
## Why

Providers configured with `gateway_oauth` need gateway credentials alongside primary authentication for inference and model discovery.

## What changed

- Attach gateway tokens through the configured header or cookie while preserving primary authentication, including WebSocket handshake headers.
- Share gateway credential managers across matching provider instances and model discovery so they observe refreshed tokens.
- Reject authentication on gateway setup or token failures, invalid token values, and conflicting auth headers. Mark gateway headers sensitive and avoid exposing issuer error details.
- Include gateway OAuth configuration in model catalog cache identity to prevent reuse across different gateway configurations.

## Testing

Add provider and core integration tests covering combined credentials, header and cookie delivery, token refresh, shared credential state, cache isolation, and request blocking on gateway token or HTTP client initialization failures.

GitOrigin-RevId: cb74125cf7e87660967d49edacf14646cf0b99c4
2026-09-18 22:42:46 +00:00
rhan-oai 3e581ebca8 Support catalog descriptions for all multi-agent V2 tools (#46297)
## Why

Model catalog description overrides only covered `spawn_agent`, leaving the other multi-agent V2 tools with fixed descriptions.

## What changed

Extend `model_messages.tools.multi_agent` description overrides to `send_message`, `followup_task`, `wait_agent`, `interrupt_agent`, and `list_agents`. Resolve each override by tool name across namespaced, plain, and Code Mode exposure.

Missing or null descriptions retain bundled text; empty strings suppress static text without disabling tools. Preserve `spawn_agent` runtime guidance, tool schemas, and execution behavior. Descriptions follow mid-turn model changes.

## Testing

Expand integration coverage to all six tools, including missing, null, empty, and sparse overrides; plain, namespaced, and Code Mode exposure; unchanged V1 behavior; and mid-turn model changes.

GitOrigin-RevId: a2c47eb8efb28c3eeebcc6482e36c188c38a6dd1
2026-09-17 20:05:47 +00:00
felixxia-oai fcf05456bb Publish Guardian cached scores and coverage atomically (#46245)
## Why

Asynchronous score publication updated the risk score, authorization, and tool-call coverage separately. Approval checks could therefore read inconsistent evidence when deciding whether to reuse a cached score.

## What changed

Keep Guardian's cached score and observation state under one lock, and use consistent snapshots for approval checks. Publish successful scores together with their authorization and classified tool-call index. Preserve timestamp ordering, fail-closed precedence on timestamp ties, and per-call oversized-action tracking.

## Testing

Add regression coverage for rejected delayed results and timestamp ties. Add a gated integration test showing that a delayed score becomes stale after intervening tool calls, while a sufficiently fresh score permits cached approval.

GitOrigin-RevId: 8df6fafd5e0ee59aa0f78f541eac562efce3d4d0
2026-09-17 16:06:11 +00:00
jif e269f2164c Include sender user messages in Guardian delegation reviews (#46179)
## Why

A delegated task can omit restrictions from the sender's user messages. Guardian needs that context when reviewing actions in the receiving thread.

## What changed

- When thread-owned Guardian context is enabled, capture up to three recent local user messages when a `send_message_to_thread` delivery from `codex_app` or `codex_tui` is accepted.
- Provide a bounded snapshot to synchronous and asynchronous reviewers, with explicit notices for unavailable evidence and a reminder that historical context does not transfer permission. Keep the snapshot out of the worker prompt.
- Retain snapshots with their deliveries across replay, compaction, and rollback, and strip them from forked agent history.

## Testing

Add coverage for reviewer-only visibility, missing sender provenance, context-budget preservation, checkpoint replay and rollback, fork isolation, and input serialization that rejects harness metadata.

GitOrigin-RevId: 3870ba762aa81689713b56d4cb3b98566bffed88
2026-09-17 10:02:38 +00:00
rhan-oai c5d079470e Allow model catalogs to override the V2 spawn_agent description (#46123)
## What changed

Read the static V2 `spawn_agent` description from `model_messages.tools.multi_agent.spawn_agent.description`, independently of the runtime tool namespace. Missing or null values retain the bundled description; an empty string suppresses it.

Preserve generated model guidance, local usage hints, and tool parameters when applying an override. Resolve the description from the active model so it follows mid-turn model changes.

## Testing

Add coverage for sparse and empty catalog values, preservation of generated context and outbound tool schemas, and description updates after mid-turn model changes.

GitOrigin-RevId: 96cfa180f8f374dc868e4fabc337e49380e405da
2026-09-17 05:17:31 +00:00
Rennie 40584fad87 Keep MCP user interaction on the root thread (#46066)
## Why

MCP requests that need human input, including browser sign-in, must be handled by the root thread. Subagents need guidance to hand these blockers to their parent without prompting the user or automatically accepting requests that require input.

## What changed

- Reject interactive MCP elicitations and tool approval prompts in subagents with guidance to ask the parent and wait before retrying.
- Recognize browser authentication and `codex_requires_user_input` metadata even when the form schema is empty, while preserving automatic permission approvals and review decisions.
- Carry user-interaction eligibility through MCP runtime creation and connection reuse, and guard prompt registration.
- Preserve connector authentication diagnostics alongside handoff guidance, subject to normal tool-output limits.

## Testing

Add unit and integration coverage for blocked subagent prompts, root browser authentication, automatic approval and review, connection reuse, and authentication diagnostic preservation and truncation. Add a request-history snapshot for browser-auth handoff guidance.

GitOrigin-RevId: 83e146b6e5c1a2a22a34681ad45615f504b3c96c
2026-09-17 01:18:37 +00:00
ptiet-oai a6d4741d39 Add per-app tool exposure configuration (#46035)
## Why

Server-level tool exposure settings apply to every connector on the apps server. Individual connectors need a way to opt out of deferred discovery without changing exposure for other apps.

## What changed

Add `apps.<connector_id>.omit_tools_from` with support for `code_mode`, `deferred`, and `direct`. Combine connector omissions with server omissions so app settings preserve server restrictions.

For example, `omit_tools_from = ["deferred"]` makes a connector's tools available without tool search, through direct calls or Code Mode as permitted by the active tool mode and remaining restrictions.

Expose the setting in the app-server protocol, JSON schemas, and generated TypeScript and Python types.

## Testing

Add integration coverage for connector-specific exposure and MCP dispatch across tool modes, server restrictions, and direct-only namespaces. Extend config tests to cover populated, absent, and empty omission lists, and add a Code Mode request-history snapshot.

GitOrigin-RevId: cc99ab290c8a878d06d8d7ccf2d1dbd7c46bfa8a
2026-09-16 21:50:07 +00:00
felixxia-oai 0c3a14bbc2 Preserve Guardian authorization evidence across checkpoint migration (#45782)
## Why

Older compaction checkpoints can lack a producer model hash. Enabling thread-owned Guardian context must accommodate those checkpoints without losing user instructions or verified answers.

## What changed

- Select review policy from each history snapshot independently of the session's evidence capture policy.
- Keep legacy review for checkpoints with missing or empty producer hashes when the transcript preserves the evidence. Preserve strict compatibility checks when review depends on retained-only evidence.
- Continue capturing retained instructions and answers during migration, while supplying legacy reviews with runtime answers and subagent reviews with root authorization evidence.
- Record producer model hashes on new compactions regardless of context mode.

## Testing

Add coverage for checkpoint migration across compaction and resume in both history storage modes, snapshot policy stability, retained-evidence preservation, and authorization revalidation for owning sessions and subagents.

GitOrigin-RevId: ec72b4a51d40729173e24ac1b53b7d516ddaa825
2026-09-15 20:08:36 +00:00
Matthew Zeng af1fc2dbff Honor canonical plugin disables for shared connectors (#45755)
## Why

Disabling a connector's canonical plugin could leave its tools available when another enabled plugin contributed the same connector.

## What changed

Retain `canonical_app_id` in the remote installed-plugin cache and use ownership metadata for the current account when building connector snapshots. A disabled canonical owner now excludes its connector even when another plugin contributes it or the owner's bundle is absent from the host. Match owners by plugin name and marketplace, and ignore ownership metadata when the cache no longer matches the current authentication.

Build these snapshots through `PluginsManager` when the plugins feature is enabled; otherwise use an empty snapshot.

## Testing

Add manager coverage for combined local and canonical exclusions, marketplace matching, and authentication changes. Add an integration test showing that disabling a noncanonical contributor preserves shared calendar tools, disabling the canonical owner hides them, and clearing the exclusion restores them.

GitOrigin-RevId: ee2981d1277825fefb671dd7d1078cd7ea5c0ea5
2026-09-15 18:23:08 +00:00
Bryan Ashley b9bfc0aff8 Allow background persistence for steered user input (#45506)
## 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
2026-09-14 19:39:31 +00:00
jif 08d3748cf0 Add managed thread lifetimes with cancellation-safe startup (#45502)
## 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
2026-09-14 19:30:11 +00:00
ningyi-oai 1715e55076 Bind direct tool-call metadata to invocation outputs (#45185)
## 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
2026-09-13 06:53:20 +00:00
pakrym-oai e8271aa8b4 Add scenario snapshots for remote compaction and Code Mode tools (#44934)
## What changed

Add two integration scenarios for `gpt-6-astra` that snapshot request history and settings:

- A multi-turn conversation with local and plugin skills, remote compaction, and a follow-up containing an image.
- A release check combining direct collaboration calls with Code Mode shell commands, MCP calls, image viewing, and patch application, including reading back the edited file.

GitOrigin-RevId: 9b8a48a817c3dba4b139d1b0d015e8084dcb219b
2026-09-11 23:14:52 +00:00
Nick Steele 654b0a77d0 Add trusted enterprise MCP auth configuration (#44832)
## Why

Enterprise MCP authorization must remain controlled by host, user, or managed configuration. Project settings and plugin declarations must not redirect enterprise credentials or downgrade the selected authentication mode.

## What changed

- Add `auth = "ema_auth"`, shared `mcp_enterprise_managed_auth` IdP configuration, and per-plugin enterprise registration settings.
- Require enterprise registrations to come from one non-project configuration layer, preserve managed IdP precedence, and reject project overrides that change authorization or re-enable disabled enterprise servers.
- Add the disabled-by-default `use_xaa` feature flag and require non-project opt-in or a managed requirement.
- Reject plugin declarations that select `ema_auth`. Keep EMA connections unsupported and block ordinary MCP OAuth login and credential fallback for this mode.
- Extend the configuration schema and serialization for the new settings.

## Testing

Add coverage for IdP precedence, incomplete registrations, alternate credential validation, project-only opt-in, plugin self-declaration rejection, configuration round-tripping, and trusted-project attempts to downgrade enterprise auth to OAuth or ChatGPT auth.

GitOrigin-RevId: f193bdfda74208c98e496a23bf2916955b131a1c
2026-09-11 14:25:56 +00:00
alexsong-oai 1aaa453ce2 Enforce managed model provider selection and definitions (#44650)
## What changed

- Support `model_provider` and `model_providers` in managed requirements. Required selection overrides local and session configuration; each required provider definition replaces the corresponding local entry, including its authentication and headers.
- Merge provider requirement fragments before validation, preserving source-relative `auth.cwd` paths. Reject invalid cloud provider definitions before replacing the cached bundle, including unsupported Bedrock overrides.
- Expose `modelProvider` and `modelProviders` through the configuration requirements API and generated schemas. Reject writes to managed provider settings and omit their local origins, including for provider IDs containing dots.

## Testing

Add tests for requirement precedence, authentication fragment merging, path resolution, cloud cache preservation, and read-only configuration RPCs. Integration tests verify that model discovery and inference use the required gateway and ChatGPT authentication without leaking local headers, and that provider requirements survive configuration rebuilds.

GitOrigin-RevId: d1448604d4ef5662ebb5df7b693712d3b8ad6d54
2026-09-10 21:41:39 +00:00
thomas 6baa076eb6 Allow extensions to select MCP protocol mode per HTTP server (#44571)
## Why

Extension-owned HTTP MCP servers previously inherited the default protocol mode. Extensions need to select a mode for their own server independently of other HTTP servers.

## What changed

- Add `McpServerContribution::SetWithProtocolMode` and re-export `McpProtocolMode` through the extension API.
- Carry the winning registration's protocol override through catalog resolution and materialization, and apply it to Streamable HTTP connections.
- Preserve existing defaults when no override is present. Selecting a protocol mode does not grant host-owned Apps cache access or environment authority.

## Testing

Add coverage for registration precedence and materialization, and extend cache isolation tests to cover explicit protocol selection. Add an integration test verifying that extension servers can select either the legacy or newer protocol while other HTTP servers retain the default mode.

GitOrigin-RevId: 606cecc03094a93258ffc433849f575020b81473
2026-09-10 16:02:59 +00:00
jif 5d3fe48b08 Improve Guardian retries and review failure reporting (#44482)
## Why

Transient rate limits can end automatic approval reviews prematurely, and review failures currently report high risk even when no assessment completed.

## What changed

- Retry rate limits and recoverable exhausted-stream errors, while excluding non-transient HTTP failures.
- Preserve server retry delays after stream retries are exhausted and honor them within the review deadline. Scope retry advice to the current turn so reused sessions cannot apply stale delays.
- Keep failed reviews denied, but leave risk and authorization unset and explain that the review could not complete without declaring the action unsafe.

## Testing

Add an integration test covering rate-limit recovery through approval and tool execution, asserting that the action executes exactly once after approval. Update failure assertions to check absent assessment fields and the review-failure explanation.

GitOrigin-RevId: 1163cfde35c6b8eb23b6f24f1f24461b86ded838
2026-09-10 09:50:59 +00:00
thomas eb680c0558 Give hosted Codex Apps an independent MCP protocol opt-in (#44318)
## Why

Hosted Codex Apps shared the `mcp_2026_07_28` protocol setting with other MCP servers, preventing independent protocol selection.

## What changed

- Add `features.codex_apps_mcp_2026_07_28`, disabled by default, for the host-owned HTTP `codex_apps` server. Enabling it attempts 2026-07-28 discovery with Legacy fallback.
- Keep other servers, including third-party registrations with Apps-like names or URLs and local stdio servers, governed by `mcp_2026_07_28`.
- Support the new flag through `experimentalFeature/enablement/set` and refresh protocol settings in existing sessions. Track protocol mode per server when deciding whether to reuse connections.

## Testing

Add coverage for independent feature and protocol selection, hosted event-stream protocol selection, standalone Legacy defaults and discovery fallback, and runtime enablement refreshing an existing app-server thread.

GitOrigin-RevId: 640e8e251f0c45e5f4829cc2b73c3082cf42f715
2026-09-09 22:22:08 +00:00
felixxia-oai fcd90d8f07 Enforce complete request budgets for Guardian reviews (#44281)
## Why

Review evidence can fit its own limits while exceeding the reviewer context window once history, tools, output format, and reminders are included. Tool continuations can also push an existing review over budget.

## What changed

- Select evidence after tools and turn context are resolved, reserving request overhead and a 256-token margin. Omit optional evidence while preserving required sections, and reject required evidence that cannot fit.
- Keep incoming evidence pending until budget selection succeeds so cancellation and compaction cannot record unselected input in reusable history.
- Check the assembled prompt before sampling, including retries and tool continuations. Retry budget failures with summary compaction once per model step, and retire reviewers whose budgets remain exhausted. Avoid history-resetting token-budget compaction for this recovery.
- Record section costs after budget selection.

## Testing

Add unit and integration coverage for required-evidence rejection, optional-evidence omission, reminder overhead, cancellation safety, incoming-review and tool-continuation compaction, and reviewer retirement after ineffective compaction.

GitOrigin-RevId: 2f7ef191b3f46032d295ad363eb7debf9d7a9684
2026-09-09 19:21:08 +00:00
rka-oai 3dc1e2a584 Always use streamed remote compaction for supported providers (#44255)
## What changed

- Route manual and automatic remote compaction through the v2 implementation.
- Retire the `remote_compaction_v2` feature toggle while continuing to advertise it in the model client beta features header.
- Migrate compaction tests from `/responses/compact` to streamed `/responses` requests with `compaction_trigger` items and remove legacy parity coverage.

## Testing

Add streamed compaction coverage for tool search output trimming and session base instruction token estimates. Update app-server tests to verify started/completed notifications for both local and streamed remote compaction.

GitOrigin-RevId: cea392a6166a9b116ff83708fd9126d5546476e4
2026-09-09 18:04:04 +00:00
jif ce2c2759eb Release persistent writers when session startup is cancelled (#44183)
## Why

Cancelling a session after persistence opens but while MCP startup is still pending can leave its writer held, blocking a subsequent resume.

## What changed

Create `LiveThreadInitGuard` inside the persistence startup future so it protects the live thread while `tokio::join!` waits for other startup work. Dropping the startup future then schedules writer cleanup.

## Testing

Add a regression test that cancels a resume while MCP startup is blocked, verifies the local writer is released, and resumes the same thread successfully.

GitOrigin-RevId: 57d8624e783a7b993579a6386aea35ba845da0e3
2026-09-09 15:54:07 +00:00
Ahmed Ibrahim f046cf35df Scope model catalog caches to the current provider and auth identity (#43906)
## Why

Switching providers or authentication could reuse a previous identity's model catalog, including its default service tier. An in-flight refresh could also overwrite a newer account's catalog after cache storage completed.

## What changed

- Require a matching provider and auth identity for disk and in-memory catalog reuse. Treat legacy entries without an identity as cache misses and fall back to bundled metadata when the in-memory identity no longer matches.
- Recheck identity after fetching and storing a catalog before publishing it.
- Renew cache freshness only when the client version, identity, and ETag match the stored entry.
- Update cache fixtures to include the configured provider and auth identity.

## Testing

Add regression coverage for provider and auth switches, credential rotation, legacy and mismatched cache entries, conditional TTL renewal, and an account switch during cache storage. Verify that API-key requests do not inherit a ChatGPT catalog's default `service_tier`, while explicit tier selection still works.

GitOrigin-RevId: e15d7ffff238d951d7a1c6a68381cbb479994146
2026-09-08 20:37:30 +00:00
Benjamin Carlsson b01c3986fd Add live WebRTC voice conversations to the TUI (#43581)
## What changed

- Add feature-gated `/voice`, `/voice mute`, and `/voice stop` commands with local WebRTC audio and app-server signaling.
- Show live transcripts, conversation status, and microphone and speaker levels. Preserve captions and undelivered answers across thread switches.
- Speak final answers from voice handoffs while keeping delegated reasoning and commentary hidden and typed answers unspoken.
- Retry eligible startup failures once and clean up voice sessions on thread switches and disconnects.
- Stop voice and block late handoffs after a misalignment policy violation. Remove realtime event payloads and spoken text from receipt and debug logs.

## Testing

Add coverage for voice command mapping, rejected starts, disconnect cleanup, delegated final-answer speech, late handoff rejection, and transcript and footer rendering.

GitOrigin-RevId: 11def1f86b0e023b21d0d92072d6636ccad89815
2026-09-07 20:25:30 +00:00
riley-oai 3cd6004dc4 Add capability-gated MCP user-verification handling (#43289)
## What changed

- Handle `openai/userVerification` requests through `openai/elicitation/create` when the client advertises `userVerification` support.
- Validate request fields, size limits, and base64url encoding. Require a correctly shaped, bounded proof for acceptance; cancel invalid acceptance responses and discard proof material on decline or cancellation.
- Handle request and service cancellation, including cancellation received before request dispatch. Bound early-cancellation storage and cancel new verifications when capacity is exceeded.
- Keep verification inactive in Codex: cancel requests without prompting until the typed app-server and UI path is available.

## Testing

Add tests for request and proof validation, malformed modes, cancellation races and capacity limits, timeout-pause cleanup, and independent concurrent requests. Add integration coverage showing that verification does not prompt or prevent tool and turn completion before activation.

GitOrigin-RevId: 6313c1be40fae0b33cc3ce0f8fe025b4124ed2ef
2026-09-06 21:41:16 +00:00
felixxia-oai 56a8470aa0 Record reasoning effort changes in conversation history behind a flag (#43110)
## What changed

Add the disabled-by-default `reasoning_effort_override` feature for OpenAI models with `use_responses_lite` enabled. Append a trusted `configuration_update` after accepted input when no effort is established in surviving history or the resolved effort changes. Preserve the existing history prefix and continue sending the request-level reasoning effort.

Share effort normalization through `ModelInfo::resolve_reasoning_effort` so requests and history updates resolve `ultra` consistently and translate `persistent` to `disabled`. Exclude other custom effort values from history updates, and compare only against harness-authored configuration items.

## Testing

Add integration coverage for effort transitions, deduplication, history prefix and cache-key preservation, alias normalization, and feature/provider/model gating. Add unit coverage for model-specific `ultra` resolution, fallbacks, and `persistent` translation.

GitOrigin-RevId: c3fe7050058076454d8ea20054a618bca9d48c7c
2026-09-05 21:56:43 +00:00
felixxia-oai 5971d42847 Preserve verified answers across history compaction (#42293)
## Why

Compaction replaces the model history window, but host-verified user answers may
still be needed for the rest of the owning thread.

## What changed

- Add bounded, model-invisible retained context for verified
  `request_user_input` answers.
- Persist retained context in rollout events and compaction checkpoints, and
  restore it when a thread resumes.
- Remove answers when their source turns are rolled back, and prevent
  parent-local answers from being inherited by child forks.

## Testing

Add lifecycle coverage for deduplication and size limits, compaction and resume,
rollback, rollout migration, and child forks.

GitOrigin-RevId: c32900d3e8120a9899b71f0576683acd67a4b898
2026-09-02 15:07:11 +00:00
rka-oai 2c79ee6dac Add structured asynchronous user input requests (#42178)
## What changed

- Replace `send_user_message_async` with `request_user_input_async`, accepting one or more questions with optional suggested answers while allowing the turn to continue.
- Attach structured question metadata to asynchronous agent messages while retaining readable fallback text, and preserve it through app-server events, thread history, and generated schemas.
- Continue enabling the new tool for model catalogs that advertise either the old or new tool name.

## Testing

- Cover tool registration, question validation, emitted items, continued turn execution, history preservation, and backwards-compatible deserialization.

GitOrigin-RevId: ffc49aebde2e854c3c50a122aa52805fec3fc6b3
2026-09-02 02:09:35 +00:00
Zahan Malkani 34c4f7e72d Allow per-call sideband endpoints for existing realtime calls (#41923)
## What changed

- Add an in-process `sideband_base_url` override to `ConversationStartTransport::ExistingCall`, while keeping app-server requests on the configured or default endpoint.
- Prefer the per-call endpoint when attaching and reconnecting a sideband, falling back to `experimental_realtime_ws_base_url` when no override is supplied.
- Preserve runtime authentication headers on the selected endpoint without adding bearer authorization.

## Testing

- Cover configured and per-call endpoint selection, authentication headers, call-specific handshake paths, and reconnect behavior.

GitOrigin-RevId: 16c439fe0a9ff8bcdda4ec615bb9983e30935052
2026-08-31 22:29:40 +00:00
Owen Lin 5f79a92e39 Persist response token usage in rollout history (#41912)
## Why

Resumed threads need to continue per-turn and per-thread token totals without scanning arbitrarily far beyond the latest compaction checkpoint.

## What changed

- Add durable `TokenUsageRecord` rollout items with response, turn, thread, session, and root-turn attribution.
- Restore accumulated usage on resume and snapshot the latest record plus the compaction response ID in compaction checkpoints.
- Preserve root-turn lineage in persisted turn context, while ensuring forked child threads start with their own usage totals.

## Testing

- Cover usage accumulation across multiple responses and resumed turns.
- Cover local and remote compaction checkpoints, invalid remote compaction output, rollout reconstruction, and fork isolation.

GitOrigin-RevId: ef9e0c4a9102a08a2c382be4cdac68c84353c90a
2026-08-31 21:15:38 +00:00
jif 1c1e17782a Preserve Guardian review evidence across compaction (#41846)
## Why

Context compaction can replace original conversation items that Guardian still needs to review approval-sensitive actions.

## What changed

- Retain a bounded, chronological review history independently of the model's compacted history.
- Keep separate limits for user messages and other transcript items so tool traffic cannot evict user instructions.
- Use the retained evidence for synchronous and asynchronous Guardian transcripts, while resetting it after rollback or history reconstruction.

## Testing

- Cover retention through compaction and eviction, reset behavior after rollback, independent size limits, and asynchronous transcript construction.

GitOrigin-RevId: c5851e992820504fc18de8cfec7f6f9f1b69f316
2026-08-31 13:57:59 +00:00
jif 0a12b855a0 Preserve Guardian authorization across history compaction (#41660)
## Why

Compaction and host-injected context can rewrite the model-visible conversation
without changing what the user authorized. Treating those updates as authorization
changes prevents Guardian from reusing an otherwise valid review.

## What changed

- Track a host-owned user-message revision separately from the conversation history
  generation.
- Advance the revision for genuine user messages and history resets, while preserving
  it across compaction and internal context injection.
- Use message content-kind metadata to distinguish host context from user input,
  conservatively treating unknown or incomplete metadata as user authorization.

## Testing

Added coverage that cached Guardian authorization survives compaction and internal
context, but is invalidated by user input and rollback.

GitOrigin-RevId: c7960fa182d1686042c23764d2abc58acbc4b882
2026-08-30 11:56:24 +00:00
jif 8faf7252f0 Test resuming compressed shared rollouts (#41364)
## What changed

Add an end-to-end core test that compresses a parent and forked child rollout,
then resumes the child through paginated history loading. Verify that the
resumed model input:

- retains the persisted compaction checkpoint and inherited history;
- includes child history through the fork boundary;
- excludes obsolete pre-checkpoint and post-fork parent messages; and
- reads the compressed ancestor without materializing its JSONL file.

GitOrigin-RevId: 1624c4d844345fe2671436f37bc39b3e59fb3dea
2026-08-28 16:36:38 +00:00
jif 5eea8d0dd3 Review terminal input against retained permissions (#41328)
## Why

An existing terminal keeps the sandbox, network settings, and grants it had at launch. Later `write_stdin` calls must account for that retained authority, including when permissions change between turns.

## What changed

- Capture each terminal's effective launch permissions and compare them with the current environment before accepting non-empty input.
- Request approval when the terminal bypassed the sandbox, retains additional grants, strict auto-review is active, or its sandbox or network settings differ from the current policy.
- Reject input when current denied-read or environment-owned network restrictions cannot be enforced by the existing terminal, while keeping empty polls and non-TTY interrupts available without review.
- Include the effective sandbox mode and user-visible additional grants in `write_stdin` approval actions without exposing host-owned grant paths.

## Testing

Added unit and end-to-end coverage for retained cross-turn grants, strict auto-review, sandbox and network policy changes, denied reads, Windows sandbox backends, and private grant redaction.

GitOrigin-RevId: e2d71daf9358474c885998786c4928069100711c
2026-08-28 11:38:03 +00:00
jif dc2ccc6843 Make subagents follow the root service tier (#41308)
## What changed

- Share the root thread's current service tier across its agent tree.
- Apply that tier to each subagent request when the selected child model supports it, including remote compaction requests.
- Propagate root service-tier updates to existing and newly spawned subagents without rewriting child-owned settings.
- Remove per-spawn and role-level service-tier overrides so routing remains controlled by the root thread.

## Testing

- Cover supported and unsupported child models, full-history forks, live tier changes, remote compaction, new subagents, and reloaded subagents.

GitOrigin-RevId: 6c299e224e90b41a558788fb7aa4d5f336fa383a
2026-08-28 09:48:13 +00:00
trolle-oai 124e560b93 Make the optional MCP startup grace configurable (#41199)
## What changed

- Add `mcp_optional_startup_grace_ms` with a default of 1,000 ms to control how long tool catalog capture waits for optional MCP servers.
- Treat a value of `0` as disabling the shared grace, so optional servers use their configured `startup_timeout_sec` instead.
- Apply updated grace values during runtime and MCP configuration refreshes, and reset cached startup deadlines when the configured duration changes.

## Testing

- Cover custom, disabled, refreshed, and shared startup grace behavior in MCP connection and core integration tests.

GitOrigin-RevId: 936196215ed6e0a29c9ec72cfee17663978df8ef
2026-08-27 20:31:40 +00:00
jif bce96bcb43 Reject oversized reviewed terminal input (#41159)
## Why

Reviewing a shortened `write_stdin` action while sending the complete input could allow unreviewed bytes to reach an escalated terminal.

## What changed

- Require the fully formatted action and approval reason to fit within 8,000 bytes.
- Reject truncated or oversized actions before requesting approval or writing any bytes to the terminal.
- Document the reviewed stdin size limit in the app-server protocol.

## Testing

Add an integration test that uses JSON escaping to push a formatted action over the limit and verifies that it is neither offered for approval nor executed, while a subsequent valid write still succeeds.

GitOrigin-RevId: f6c96a78a65a136366c8119a7d242dfc2ec05fe6
2026-08-27 14:54:13 +00:00
soyeon-oai ae357e7250 Attach verified access context to eligible plugin MCP calls (#41005)
## What changed

- Fetch ChatGPT account access and attach `cyber_trusted_access` under
  `openai/entitlementContext` when an installed or selected plugin explicitly
  requests it from a local, read-only stdio tool with no arguments.
- Reject caller-supplied entitlement context on direct MCP calls, and fall back
  to an `unknown` status when access cannot be verified or the account changes
  during the lookup.
- Count entitlement lookup time against the MCP tool-call timeout.

## Testing

- Cover eligible calls plus unowned, remote, HTTP, mutable, argument-bearing,
  API-key, undeclared, direct, and account-switch cases in core and app-server
  integration tests.

GitOrigin-RevId: 303b61585ef4530276c373d914a2d0ccef1f4ac2
2026-08-27 00:02:26 +00:00
faizan-oai 94967e03e5 Add per-turn cyber access program selection (#40687)
## What changed

- Add the experimental `turn/start.cyberAccessProgram` option with `standard`,
  `daybreakBlue`, and `daybreakRed` values.
- Forward the selection as `access_programs.cyber` on Responses, WebSocket,
  and remote-compaction requests made with ChatGPT authentication, while
  omitting it for API-key and custom-provider requests.
- Preserve the per-turn selection across recovery, compaction, and child-agent
  turns without making it a persistent thread setting.

## Testing

- Cover app-server forwarding, authentication boundaries, WebSocket reuse,
  turn recovery, compaction, and child-agent inheritance.

GitOrigin-RevId: d2eb468f365b2214c5099bc21741cf30f8bd2eb5
2026-08-25 19:54:52 +00:00
joeflorencio-openai a10c8127f7 Add live turn settings updates (#40653)
## What changed

- Add a feature-gated `TurnSettings` operation that can update the model,
  reasoning effort, reasoning summary, and service tier for subsequent steps of
  a named running turn without changing future thread settings.
- Report whether an update was applied, rejected, or lost its live target.
- Revalidate managed constraints and preserve the turn's admitted approval and
  Guardian safety properties before publishing an updated settings snapshot.

## Testing

- Cover sparse updates, step capture and ordering, target replacement, managed
  policy changes, model metadata safety, and separation from future settings.

GitOrigin-RevId: ecaaaa95b4fd4ec9d62265bef551582dddb36d78
2026-08-25 17:44:50 +00:00
joeflorencio-openai 68301fa45f Snapshot resolved settings for each model step (#40651)
## Why

Thread settings can change while a turn is running, and delegated review or
compaction steps can select a different model. Each request needs a consistent
set of settings resolved against the model that will execute it.

## What changed

- Capture immutable resolved settings for each model step, including model
  metadata, reasoning options, service tier, approvals, and personality.
- Keep in-flight steps on their captured settings while applying thread updates
  to subsequent work.
- Resolve inherited reasoning summaries and filter service tiers against the
  selected model and feature configuration for review and compaction steps.

## Testing

Added coverage for settings updates during paused turns, model changes,
previous-model compaction, review-model defaults, and service-tier filtering.

GitOrigin-RevId: fca4b3b305379f1bf26a5cac75ee0dd1747ac1f6
2026-08-25 17:41:22 +00:00
joeflorencio-openai 1d64085e67 Validate step settings against proposed permissions (#40647)
## Why

Model requirements must be evaluated against the permissions that will apply
after a settings update. Validating against the current profile can reject a
valid model-and-permission update or allow a later permission-only update to
invalidate the selected model.

## What changed

- Group model, reasoning, service-tier, personality, and approval inputs into
  `StepSettings`, with a single apply-and-validate path.
- Apply permission and environment changes before validating step settings, so
  auto-review requirements use the proposed effective permissions.
- Revalidate existing step settings when permission or environment changes alter
  their constraints, and commit the update only when the complete candidate is
  valid.

## Testing

Added unit and integration coverage for atomic model/permission updates,
permission-only revalidation, environment-derived constraints, and both
standalone and turn-start settings operations.

GitOrigin-RevId: f50dfbd300484a09acc51ee4607278f300ca1cd9
2026-08-25 17:27:04 +00:00
joeflorencio-openai 4e883ce36e Keep settings operations bound to their committed snapshot (#40645)
## Why

Post-commit work can overlap with another settings write. Reading the session
again afterward could make a settings operation report or use the later state
instead of the state that it committed.

## What changed

- Return the committed configuration and `ThreadSettingsSnapshot` from session
  settings updates.
- Use that commit result when creating a turn and emitting
  `ThreadSettingsApplied`, while retaining current-state snapshots for
  synthesized fork history.
- Route turn-start updates through the shared settings update path.

## Testing

Add coverage for standalone and turn-start updates that pause after committing,
restore different settings concurrently, and verify that each operation keeps
its original configuration and notification snapshot.

GitOrigin-RevId: cee5be9536d2840170893e6eb0c64cd039fb0a96
2026-08-25 17:20:37 +00:00
Andrei Eternal cbfd999db7 Add hooks for interrupted turns (#40511)
## What changed

- Add an `Interrupt` hook event that runs for an active top-level turn before its
  interrupted abort event is emitted.
- Flush the turn transcript before invoking the hook and provide the session,
  turn, transcript, working directory, model, and permission mode in its input.
- Support command and MCP handlers, including asynchronous commands, with a
  one-second default timeout and a three-second maximum.
- Expose the event through hook configuration, managed requirements, app-server
  notifications, generated schemas, analytics, and the TUI hook views.

## Testing

- Cover handler discovery, timeout normalization, output parsing, protocol
  compatibility, TUI rendering, and interrupt execution ordering.

GitOrigin-RevId: 163fa7c098d94ac2775f6d137f8e916f8ea9b6eb
2026-08-25 01:27:18 +00:00
pakrym-oai c4b434aae2 Preserve context annotations in merged messages (#40184)
## What changed

- Carry each contextual fragment's `ContentItemKind` through initial context assembly and world-state updates.
- Attach the classifications to merged response messages through `content_item_kinds`, preserving their order alongside the message content.
- Convert extension prompt fragments into annotated developer fragments and classify built-in developer and guardian instructions.

## Testing

- Add an integration test covering roles and content annotations in the first request.
- Update world-state and persisted-history tests to account for annotation metadata.

GitOrigin-RevId: 71c48dad5b17bffd47ba03263e3eb6d5e74fbf4b
2026-08-23 04:41:11 +00:00
Won Park 6143217c67 Cancel Guardian reviews with their tool calls (#40021)
## What changed

Propagate tool cancellation tokens into Guardian approval reviews so interrupting a tool also aborts its pending review. Apply the same cancellation behavior to server-initiated MCP approval elicitation.

## Testing

Add integration coverage for cancellation from direct tool calls, code-mode turns, background code-mode cells, and MCP elicitation.

GitOrigin-RevId: cbee4b422bbcbf6984e415ef6409cee705beebdc
2026-08-21 21:51:12 +00:00
Krish Chainani 677cfee000 Add end-to-end tests for executor Stop hooks (#40020)
## Testing

- Verify an executor plugin's `Stop` hook starts running after its environment
  attaches and stops after disconnection.
- Confirm hook calls carry the expected session, thread, turn, model, and request
  metadata.
- Reject hooks whose MCP server belongs to a different executor environment.
- Cover the current restriction to the first executor environment and handler.

GitOrigin-RevId: ef13baf61379997f117cc57515363cc9880d3724
2026-08-21 21:45:26 +00:00
jif d12a7f3fd8 Preserve root user authorization in subagent Guardian reviews (#39975)
## Why

MultiAgent V2 workers can be reviewed after the root conversation has received
new user authorization that is absent from the worker's own transcript. Guardian
reviews need that genuine user context without treating forwarded or
assistant-authored claims as authorization.

## What changed

- Add the bounded root conversation to Guardian prompts for worker reviews.
- Preserve user and assistant roles, and explicitly treat only root user messages
  as authorization evidence.
- Exclude summaries, review artifacts, and assistant commentary from the root
  conversation evidence.
- Apply the same evidence to both core Guardian prompts and Guardian V2
  classification.

## Testing

Add an integration test covering late root-user authorization, forged role text,
forwarded agent claims, and filtered synthetic authorization.

GitOrigin-RevId: 8846de69ba214a88f58f70675798d324e213427c
2026-08-21 17:12:42 +00:00
jif 8a40095ea3 Standardize shell execution on unified exec (#39757)
## What changed

- Remove the legacy `shell_command` handler and runtime, leaving `exec_command`
  and `write_stdin` as the shell execution tools.
- Treat legacy `shell_command` model metadata as `unified_exec`, and normalize
  legacy user opt-outs so they do not disable command execution. Managed feature
  requirements and `shell_tool` can still disable it.
- Preserve shell approvals, sandboxing, zsh-fork support, and output truncation
  through the unified execution path.

## Testing

- Cover legacy configuration and model-metadata compatibility.
- Exercise unified shell execution, approvals, truncation, and `apply_patch`
  serialization across the app-server and core test suites.

GitOrigin-RevId: 5c2fd6164fc3519cdae4944cb9db276b8467311c
2026-08-20 17:46:05 +00:00
Jeremy Rose bc3545b805 Validate linked worktrees before inheriting project trust (#39616)
## Why

A checkout could point its `.git` file at a trusted repository's worktree
directory without proving that the repository had registered that checkout.
This could cause project configuration from an unrelated checkout to be treated
as trusted.

## What changed

- Verify the linked worktree's `gitdir` backlink, `commondir`, registered
  checkout, and main checkout ownership before resolving the main repository's
  trust key.
- Reject missing, oversized, symlinked, mismatched, or swapped Git metadata.
- Preserve valid linked worktrees that use path aliases, separate Git
  directories, or non-UTF-8 POSIX paths.

## Testing

Add resolver and config-loading coverage for forged worktrees, metadata races,
case-sensitive paths, moved worktrees, and host MCP startup from project config.

GitOrigin-RevId: 6052a7d10ad2d613436f20175c356abdef8c758e
2026-08-20 06:29:32 +00:00