Commit Graph
1384 Commits
Author SHA1 Message Date
oliverye-oai edd0d89820 Add a root-only option for background compaction (#59598)
## What changed

Add the disabled-by-default `async_compaction_root_only` feature flag. When enabled, it restricts background compaction to root sessions, excluding non-root agents. Expose the flag in the configuration schema.

## Testing

Add a parameterized test covering root session eligibility and eligibility for review subagents and guardian sessions with the flag enabled and disabled.

GitOrigin-RevId: b4cce3f24f7f418ae61a82a7aef310f849cfd532
2026-11-05 07:59:49 +00:00
cooper-oai 388aee4b26 Use runtime authentication across backend request paths (#59572)
## What changed

- Share the harness runtime identity with analytics delivery and turn-cost queries, and resolve request authentication together with its account-bound transport policy.
- Add runtime authentication to ChatGPT request helpers and `codex cloud`, including environment discovery. Resolve environment credentials for each request and honor CLI configuration overrides when loading cloud authentication.
- Preserve custom-backend authentication, validate runtime destinations, and keep cloud requests from following redirects.
- Forward only organization and project selectors from model-provider headers to API-key analytics requests.

## Testing

Add regression coverage for runtime request signing, credential refresh, account and resource selection, redirect credential handling, custom destinations, and cloud credential-store overrides.

GitOrigin-RevId: 75f33961ee6482881e9ff9817e99d977f31b65b8
2026-11-05 05:09:32 +00:00
cooper-oai 5c0b1ac71f Extend runtime auth to account, plugin, and file requests (#59564)
## What changed

- Use the shared runtime and effective configuration for account API clients and remote plugin requests, including standalone plugin commands.
- Support dual-header and assertion-only authentication for remote plugins and MCP file uploads when runtime agent instances are enabled.
- Keep plugin credentials and caches scoped to the thread's account when resuming with different credentials. Bind request pools and multi-step plugin storage operations to the selected account so account replacement revokes stale operations.
- Protect runtime credentials on cross-origin redirects while preserving configured routing policies and legacy authentication for custom backends.
- Apply server-provided `upload_headers` to file storage uploads.

## Testing

Add coverage for shared runtime registration, both authentication modes, configuration reloads, resumed-thread account selection, account replacement during plugin uploads and downloads, and credential stripping on redirects. Extend file upload tests to verify runtime assertions and storage-only upload headers.

GitOrigin-RevId: 68aef81051a6f9bf919fffda52f372b108fce805
2026-11-05 03:49:13 +00:00
rhan-oai f3074e35c0 Render host-owned V2 subagent rosters in environment context (#59531)
## What changed

- Expose `AgentSessionReader` and `AgentRef` for read-only access to host-owned agent session membership.
- Read direct children when rendering V2 environment context, using the existing roster size limits. Refresh membership when rebuilding context after compaction, and omit the roster if the provider is absent or the read fails.
- Preserve unrecorded initial input across suspension or interruption while waiting for a roster read. Resolve the roster before advancing the compaction window.

## Testing

Add coverage for roster refreshes after compaction, unchanged environments avoiding extra reads, unavailable rosters, recoverable input during blocked reads, and stale compaction results rejected after a concurrent history replacement.

GitOrigin-RevId: 3c7e3822af09e9295d405e53dabb87b020c8f00c
2026-11-05 00:39:39 +00:00
sayan-oai 9fcaeb7335 Use MCP owner permissions for Guardian reviews and score reuse (#59489)
## Why

An MCP server can belong to a different environment from the turn's primary environment. An early Guardian score based on the issuing context must not bypass review when the server owner's permissions differ.

## What changed

- Capture the owning environment's permissions when preparing an MCP call and retain them for server-initiated approvals tied to that still-running call.
- Use the captured permissions consistently in Guardian review prompts and permission context.
- Require a fresh synchronous review when the captured owner permissions do not match the early score's authorization.

## Testing

Extend regression coverage for MCP score reuse across environments and for follow-up approvals from a local MCP server during a remote-primary turn. Assert that reviews use the owner's denied paths and require a fresh review when permissions differ.

GitOrigin-RevId: eb5eeb1259732c57f39765c38e34148fa0787dae
2026-11-04 20:37:39 +00:00
Henry Levy d78f68f8d1 Allow per-turn overrides for direct-only tool namespaces (#59481)
## What changed

- Add `direct_only_tool_namespaces` to `TurnStartOptions`. Omission inherits the thread configuration; an empty list clears direct-only routing for that turn. Steering ignores the option.
- Persist explicit overrides in turn context and restore them when recovering an interrupted turn, including turns started by standalone tool output.
- Keep subsequent turns and spawned or resumed child agents on the thread's routing defaults.

## Testing

Add coverage for routing changes across turns in Responses and Responses Lite, recovery without carrying overrides into subsequent turns, and child-agent inheritance.

GitOrigin-RevId: cea1a997424572d4cd372cd15b098626904c6796
2026-11-04 19:58:49 +00:00
Rohan Mehta 41cbd5b9b2 Add per-turn prompt cache TTL options (#59480)
## What changed

- Expose `PromptCacheTtl` through `TurnStartOptions` and `RecoverTurnRequest`, supporting `30m` and `12h` minimum cache lifetimes.
- Carry the selected TTL through turn execution and compaction into `prompt_cache_options.ttl` on Responses HTTP and WebSocket requests. Omit the field when unset and keep the selection scoped to the turn.
- Include prompt cache options in WebSocket request reuse checks.

## Testing

Add coverage for both TTL values and omission on a subsequent turn without an override, and extend WebSocket serialization coverage to include prompt cache options.

GitOrigin-RevId: 8529306e6ad6030fe77c5e6bf19f3810717c0c5e
2026-11-04 19:58:24 +00:00
jif bd7cfa3f69 Record saved multi-agent opt-outs in session telemetry (#59478)
## Why

Saved `features.multi_agent = false` settings can be overridden by a higher-precedence session setting. Session telemetry should retain visibility into those saved opt-outs.

## What changed

Add `config.legacy_multi_agent_disabled_sources` to `codex.conversation_starts`, reporting opt-outs from user configuration, project configuration, and managed requirements. Emit comma-separated source labels, or `none` when no opt-outs are found.

## Testing

Add a regression test confirming that a user opt-out is still reported when a CLI override enables multi-agent support for the session.

GitOrigin-RevId: 488bd317d67e9ee87c5903ae51f466e0a63c05eb
2026-11-04 19:57:30 +00:00
Will Emmanuel 031829da6a Carry per-input extension data to user message completion callbacks (#59455)
## What changed

Add `user_input_extension_data` to `TurnInputRequest` and carry it with each queued user input through completion, including inputs that steer an active turn. Expose the data to lifecycle contributors through `on_item_completed_with_metadata`, whose default implementation forwards to `on_item_completed` for compatibility. Keep this data out of model input and preserve the existing serialized input shape when no extension data is present.

## Testing

Add serialization round-trip coverage and extend the concurrent submission test to verify that inputs sharing a client ID retain their own metadata, report completion in the accepting turn, and omit extension data from model requests.

GitOrigin-RevId: f880cf812994a030cb1be9418280ba69e2885808
2026-11-04 17:13:00 +00:00
ioanbudea-OAI 4d62359072 Add opt-in elastic budgeting for synchronous Guardian reviews (#59444)
## What changed

- Add `features.guardianv2.sync_elastic_budgeting` and `sync_elastic_sections` to configure section minimums, per-entry token caps, and optional section maximums. Require and validate section limits when enabled.
- Use the selected budgeting policy consistently for history capture, resume, prompt preparation, and review input budgeting. Keep fixed budgeting as the default.

## Testing

Add configuration validation tests, capture-and-resume coverage for both budgeting modes, and a request-history snapshot verifying retained instructions and transcript section limits.

GitOrigin-RevId: c7b0bdfd3e0e008d75c4d8a204959bcc58eb1bf8
2026-11-04 16:33:22 +00:00
jif 48524f337a Preserve pending agent messages across compaction (#59404)
## Why

Agent messages that arrive after a model request starts must survive compaction until a normal inference request includes them. Compactor output alone does not establish that those messages were delivered.

## What changed

- Persist history sequences and sampling checkpoints in harness metadata, capturing the input prefix for each normal inference request.
- Retain pending agent messages within existing size budgets during local and remote compaction, placing the summary before pending messages and subsequent caller input.
- Preserve sampling progress across history replacement and resume, and drain queued mail after interrupted sampling even when compaction leaves context above the token limit.

## Testing

Add regression coverage for request-boundary sampling checkpoints, pending messages across local compaction and repeated remote compaction with resume, and injected messages that interrupt sampling before mid-turn compaction.

GitOrigin-RevId: 26c04ab96b5a1c3cc1d6625ccc74a0f00201d14c
2026-11-04 13:03:01 +00:00
Ruslan Nigmatullin 71a4c07036 Share context window ID allocation through the agent runtime (#59354)
## What changed

Store a shared `ContextWindowIdGenerator` in `LocalAgentRuntime` and pass it into each session. Use it for both initial context window IDs and subsequent window advances, retaining UUID v7 allocation by default.

Make `AutoCompactWindow` accept caller-provided IDs instead of allocating them directly.

## Testing

Update the window boundary test to verify that advancing installs the supplied ID while preserving the first and previous window IDs.

GitOrigin-RevId: a9da4232fe6ab9a53b75e239cb412ceae0fe3216
2026-11-04 08:13:41 +00:00
Ruslan Nigmatullin 19438adcb6 Introduce a dedicated context window ID type (#59328)
## What changed

Add `ContextWindowId` as a string-backed protocol type with UUIDv7 generation and parsing. Use it throughout compaction state, rollout reconstruction, token budget context, and request metadata in place of raw UUIDs, preserving the serialized string format and UUIDv7 validation.

GitOrigin-RevId: 786b74fcc331707ca5d6765f429b6e5e322d9328
2026-11-04 06:05:16 +00:00
Ruslan Nigmatullin d84c872464 Route response item IDs through a shared generator (#59312)
## What changed

Introduce `ItemIdGenerator` and pass the runtime-owned generator through sessions, model clients, history, hooks, tool outputs, and memory requests. Keep UUIDv7 generation as the default and replace direct calls to `ResponseItemId::new` with the shared callable.

Preserve existing item IDs and backfill missing or empty IDs through the same generator, including forked history before reconstruction and persistence.

## Testing

Extend coverage for UUIDv7 IDs on user messages, agent messages, and additional tools, and assert that forked model context matches the backfilled items persisted to history.

GitOrigin-RevId: 0dd9ccf0f46f0f1005a6acfc740133b03ce82259
2026-11-04 04:02:00 +00:00
Anthony Tafoya 75aba6768a Add outcome metrics for skill reads and installation listings (#59295)
## What changed

- Record `codex.skills.read.result` and `codex.skills.installations.result` counters with bounded outcome labels for successes and failures.
- Tag both counters with `discovery_version` and installation listings with `source_filter`.
- Make session telemetry available through the step store so skill tools emit metrics with the current step's model metadata.

## Testing

Add coverage for read metrics in both discovery versions, installation listing outcomes and source filters, and extension metrics retaining the current model metadata after a model switch.

GitOrigin-RevId: 0b5418e2b94a89149a20877f32ca2b2e196884e7
2026-11-04 02:29:41 +00:00
Dmitriy Kovalenko 007f668c34 Respect turn filesystem permissions during executor skill discovery (#59291)
## Why

Direct executor skill discovery used unrestricted filesystem access, bypassing the turn's permission context. Cached catalogs also needed to account for changes in filesystem access within a turn.

## What changed

- Bind selected capability roots to callback-scoped filesystem access through `AccessedSelectedCapabilityRoot`.
- Pass accessors built with the turn's granted permissions into skill preparation and discovery, removing the executor skill provider's environment lookup and unrestricted accessor.
- Include filesystem access keys in the direct discovery cache so permission changes trigger rediscovery.

## Testing

Extend tests to verify that restricted discovery returns no skills when sandbox access is unavailable, restoring access restores discovery, and changing or restoring access invalidates the catalog cache.

GitOrigin-RevId: 0f7aa7646f65923f65f0cfc0b387d94bd7f2db05
2026-11-04 02:00:01 +00:00
Dmitriy Kovalenko 0b21a5b526 Respect filesystem permissions when discovering repository skills (#59280)
## Why

Repository skill discovery must reflect current filesystem access, including newly approved grants and narrower permissions. Previously listed skills can remain in conversation history after a turn-scoped grant expires, so catalog updates need to tell the model when availability changes.

## What changed

- Discover repository skills through the selected executor's permission-aware filesystem accessor and refresh snapshots when the working directory or access changes. Use the current step's snapshot for skill selection and invocation.
- Preserve repository boundaries even when project markers are hidden by read restrictions, and report unsupported sandbox backends as skill-loading errors.
- Track grant-dependent catalog entries and emit availability notices when skills disappear or return, even when the metadata budget is exhausted.
- Resolve implicit skill paths through their owning filesystem and exclude remote repository skills from local implicit command attribution.

## Testing

Add coverage for turn and session grants, grant expiry and restoration, hidden and newly created repository markers, cached read denials on Windows, unsupported remote sandboxes, and executor-backed implicit skill paths.

GitOrigin-RevId: ac1af483e73e3130524aeeabe218630435054028
2026-11-04 01:15:38 +00:00
Dmitriy Kovalenko 526f6099f6 Preserve filesystem authority when loading skills (#59279)
## Why

Skill discovery and metadata loading used raw filesystem handles with unrestricted reads, and cache keys distinguished filesystems without accounting for captured permissions.

## What changed

- Pass borrowed `EnvironmentAccess` accessors through skill root resolution, discovery, canonicalization, and metadata reads.
- Include captured permissions in skill snapshot and empty remote root cache keys.
- Add fixed-path watches to `EnvironmentAccess`, falling back to live discovery when read restrictions prevent using unsandboxed executor watches.
- Keep user and plugin roots under host filesystem authority independently of the selected repository executor.

## Testing

Extend remote root cache coverage to verify that changing captured permissions triggers a fresh scan and that restricted accessors decline unsandboxed watches.

GitOrigin-RevId: 97d4bfdfcba78a93fc8845773c6397d4364494e3
2026-11-04 01:13:54 +00:00
Dmitriy Kovalenko afefd0ba95 Enforce current filesystem permissions when reading repository skills (#59278)
## Why

Cached repository skill metadata must not authorize reads after access is revoked or a skill file or directory becomes a symlink to a denied location.

## What changed

Read repository skill instructions through a current environment accessor for the filesystem that discovered them. Reject reads when no matching accessor is available, and cache filesystem identity instead of retaining unrestricted read access.

## Testing

Add regression coverage for revoked permissions, file and directory symlink replacements, and missing or mismatched filesystem accessors.

GitOrigin-RevId: 8dad83b7c7893e0d66e2fe1c0c240715db582005
2026-11-04 01:12:49 +00:00
ioanbudea-OAI c141c31323 Preserve requested approval reviewers across policy overrides (#59272)
## Why

A model requirement can force `auto_review` even when the requested reviewer is `user`. Persisting or sending that effective value back as an explicit choice loses the original selection, so removing the requirement cannot restore it on resume.

## What changed

- Retain `requested_approvals_reviewer` separately from the effective reviewer in configuration and thread settings, with a fallback for older histories.
- Restore the requested reviewer on resume while applying current model requirements and reviewer allowlists.
- Let ordinary TUI turns, continuations, retries, and forks inherit the server's reviewer instead of echoing its effective value as an explicit request.
- Preserve a worker's requested reviewer when reloading it rather than inheriting the sender's selection.

## Testing

Extend resume coverage for explicit `auto_review`, removal of a model requirement, and a tightened reviewer allowlist. Update TUI assertions for omitted reviewer overrides and verify worker reload preserves its reviewer.

GitOrigin-RevId: d8c9c3a7b89e55e63f0e8ae210d6a87cb5547933
2026-11-04 00:31:24 +00:00
Hung 1685f2d621 Use measured durations to budget uploaded audio (#59268)
## Why

Uploaded audio always reserved 12,000 tokens per file, even for short clips. This could discard recordings during compaction that would fit within the retained-message budget.

## What changed

- Add optional audio duration metadata to attachments and cache measured durations per session, resolving metadata for new input and replayed history.
- Use measured durations for history estimates and compaction budgets at 10 tokens per second, rounded up and capped at 12,000 tokens. Keep the full budget for missing or unusable metadata.
- Bound metadata lookups to 250 milliseconds per batch with up to eight concurrent requests, and record accepted input before lookup so interruption preserves the batch.

## Testing

Add coverage for duration rounding and fallback estimates, mixed measured and unknown audio, retention through compaction and cold resume, and interruption during metadata lookup.

GitOrigin-RevId: 26fe320c7024944d9d4c371d5fdcda84442b5cc5
2026-11-04 00:19:12 +00:00
taohong-oai d5830f6fc6 Expand per-turn critical path capture and split inference timing (#59245)
## What changed

- Report sampling and compaction separately as `critical_path_sampling_ms` and `critical_path_compaction_ms`, retaining `critical_path_inference_ms` as their combined compatibility total.
- Enable critical path capture for subagent and Guardian turns when `CriticalPathTiming` is enabled, while continuing to exclude memory consolidation sessions.
- Preserve capture across incoming agent messages and support direct `spawn_agent` and `wait_agent` calls. Attribute these calls to the parent's tool time and keep child work on the child's own turn.

## Testing

Add runtime coverage for independent parent/child timing, incoming agent messages, Guardian turns, feature gating, and memory consolidation exclusion. Update phase accounting and analytics tests to cover the new fields and millisecond truncation.

GitOrigin-RevId: 09d4c0b853e817216bed173e4404d0d782de97ec
2026-11-03 22:29:33 +00:00
Alex Kira a47a518928 Honor Guardian's additional-context setting for tool outputs (#59239)
## What changed

Tag untrusted additional context rendered as tool outputs with history metadata and apply `guardian_include_additional_context` when building Guardian review history. Preserve this metadata across session resume and apply the current session's setting to restored review checkpoints.

Keep ordinary tool outputs in review history and preserve the acting model's additional-context payload.

## Testing

Add coverage for inclusion and exclusion after restart, restored checkpoints after compaction, and asynchronous Guardian review. Verify both line and JSON transcripts and check that metadata does not alter the acting model's request.

GitOrigin-RevId: 8ba270a6aac16398c88db3282e9d666702f462fe
2026-11-03 21:54:52 +00:00
Hung f1b0721c88 Enable uploaded audio with durable rejection recovery (#59234)
## Why

Uploaded audio was rejected even for audio-capable models. Invalid or expired recordings also need to be excluded from later requests so users can continue the conversation.

## What changed

- Accept uploaded audio when the selected model supports audio; otherwise, preserve valid sibling input and report the unsupported attachment before inference.
- Track successful audio decoding by model and provider, and attribute provider rejections against the exact submitted request. Preserve attachments when attribution is ambiguous.
- Persist confirmed rejected file IDs across compaction and resume, omitting them from future requests without rewriting retained user history.
- Check audio availability before remote compaction and retry once without an unambiguously rejected recording. Include guidance to remove or replace affected attachments in audio errors.

## Testing

Add unit and scenario coverage for uploaded audio across follow-up turns, compaction, cold and paginated resume, model switching, ambiguous rejections, expired files, and queued or interrupted input.

GitOrigin-RevId: e701901e7b1c0865c9f2a5918197215f549b06c0
2026-11-03 21:24:16 +00:00
cooper-oai 079e17a422 Wire shared runtime authentication through threads and API clients (#59177)
## What changed

- Share a lazily registered runtime across threads, model discovery, and extension clients when `runtime_agent_instances` is enabled, while honoring each thread's authentication mode and credential owner.
- Bind model catalogs to the effective provider and authentication context while preserving cache policy and live API-key discovery settings.
- Resolve credentials, routing, and transport policy together for each request. Preserve proxy and TLS settings, validate runtime account selectors, and prevent pooled connections from crossing runtime identities.
- Accept supported ChatGPT subdomains when selecting the registration environment.

## Testing

Add integration coverage for shared runtime registration, dual-header and assertion-only modes, disabled and third-party providers, credential replacement, subagents, and delegated review. Add coverage for catalog cache policy and runtime-aware connection reuse.

GitOrigin-RevId: 99dd9d987c29c93597a648f6586f98ac6a932c3c
2026-11-03 15:02:57 +00:00
jif f05aa41da6 Route message-board notifications through agent mailboxes (#59139)
## Why

Board notifications were skipped when recipients had no active turn accepting mailbox delivery. Queue them through the ordinary agent-message path so idle recipients can receive them in a later turn.

## What changed

- Submit board notices as `Op::InterAgentCommunication` with `trigger_turn` disabled, preserving normal mailbox scheduling without reopening finalized answers.
- Hold a residency guard through dispatch of submitted agent messages to coordinate delivery with agent eviction.
- Update board tool descriptions to describe the host's notification queueing and wake policy.

## Testing

Update local and HTTP board scenarios to cover queued idle notices, delivery after a finalized answer, and mailbox delivery before a pending tool call. Extend residency coverage to verify that submission waits for the eviction gate and queued messages survive eviction.

GitOrigin-RevId: a77137fc40932f5903852e14742483844e07c9bc
2026-11-03 12:33:26 +00:00
dpatel-oai 33ffb6f905 Normalize saved chat settings after model catalog refreshes (#59111)
## Why

Existing chats can retain reasoning effort and service tier selections that a refreshed provider catalog no longer supports.

## What changed

- Apply fresh provider catalogs to saved settings in open and resumed chats, and normalize incoming settings before capturing the next turn.
- Replace unsupported ranked reasoning efforts with the highest supported level at or below the selection. For unsupported `ultra` or `persistent` selections, use the model's supported default when available. Reset unsupported `priority` and `ultrafast` service tiers to `default`.
- Retire incompatible reasoning effort pins using each step's captured model metadata, preserving running steps and the live pin during compaction.
- Publish catalog updates only from successful provider responses for the current identity, and refresh retained originator catalogs through the periodic refresh worker.

## Testing

Add coverage for setting correction, reasoning effort pins, resumed chats, stale queued settings, retained history, refresh failures, identity changes, and isolated originator notifications.

GitOrigin-RevId: 2ef4605c4acba98db3fe9d0328e2752084e900a2
2026-11-03 10:01:36 +00:00
Adam Perry @ OpenAI 7e8143a897 Tie executor attachments to environment selection lifetimes (#59043)
## Why

Removing an executor must detach its selection even when older turn snapshots still reference it. Connection events queued for a removed attachment must not leak into a later selection of the same executor.

## What changed

- Introduce `EnvironmentAttachment` with a detach signal shared by turn snapshots and a guard owned by the selection.
- Preserve attachments across workspace and configuration changes, while giving reselections and child threads independent attachment lifetimes.
- Stop forwarding connection events on detach and discard buffered events from detached attachments.
- Activate initial attachments after `SessionConfigured`, using configured bindings for startup discovery.

## Testing

Add coverage for discarding buffered connection events on reselection and extend snapshot tests to verify immediate detachment, fresh attachments on reselection, and independent parent and child lifetimes.

GitOrigin-RevId: 6cf882e0d17a296030c13d80c6e2866c80e04638
2026-11-03 02:37:20 +00:00
aojiao-oai 16ddd538bd Expose assistant content token counts in completion notifications (#59041)
## What changed

Add optional `contentTokenCount` metadata to `item/completed` notifications, carrying the producer-supplied token count for each completed assistant message in default and plan modes. Counts include any sampled footer and are `null` when unavailable.

Keep this telemetry separate from stored thread items and omit it from subsequent model requests. Update the protocol schemas and generated TypeScript and Python types.

## Testing

Add coverage for per-message counts, including zero, in default and plan modes; missing and malformed metadata; backward-compatible notification decoding; and exclusion from subsequent model requests.

GitOrigin-RevId: 35973037a42650fa2d7569da8e88cb81215cf971
2026-11-03 02:26:29 +00:00
yitaoh-oai e839b828fb Resolve model-selected search providers with policy-aware fallback (#59036)
## Why

Search discovery must account for the active model's supported tools while honoring host search restrictions. Automatic fallback must also preserve the provider that owns existing browsing references.

## What changed

- Resolve search from model metadata, client configuration, and managed requirements, exposing the selected provider's tools.
- Discover hosted search with the model's supported tools and carry the discovered invocation context into calls.
- Allow native fallback for eligible discovery failures only when search policy permits it and hosted search has not already been used. Do not replay failed hosted invocations through native search.
- Keep native fallback references on native search across later turns, discovery recovery, and resume. Release unused fallback offers and coordinate background calls to prevent conflicting provider use.

## Testing

Add app-server and core scenario coverage for provider selection, discovery failures, search restrictions, reference continuity, resume, and concurrent background search. Include a snapshot covering discovery and call history across a model switch.

GitOrigin-RevId: 0f22605722daaa424c4fa34660654cde5a6db049
2026-11-03 02:08:56 +00:00
sayan-oai 0585dcb5db Preserve sandbox context in prepared MCP calls (#59020)
## Why

MCP permission profiles need the directory context used to interpret them. Prepared calls must retain that context alongside their captured permissions across runtime refreshes.

## What changed

- Store each server's permission source as thread permissions or a captured environment `FileSystemSandboxContext`.
- Capture environment working directories, home and temporary directories, and materialized workspace roots. Retain the host working directory for thread permissions.
- Expose the captured source through `PreparedMcpCall::permission_source()`, while keeping hosted Apps on thread permissions even when a local environment is attached.

## Testing

Extend binding tests to verify that prepared calls retain environment sandbox paths across refreshes and expose the host working directory for thread permissions. Update elicitation coverage to include hosted Apps with an attached local environment.

GitOrigin-RevId: 87e94605daece324786eebc07b62a507d35e5ced
2026-11-03 01:08:08 +00:00
Ben Romano dec38fc9bb Add WebSocket connection metrics with reauthentication labels (#59018)
## What changed

- Record connection attempts and durations as `codex.websocket.connect` and `codex.websocket.connect.duration_ms`, labeled with `success`, `reauth`, and the configured `product_sku` when present.
- Detect reauthentication by comparing the current auth revision with the last successful connection, preserving that revision across connection resets. Include `auth.reauth` in connection logs and traces.
- Populate session telemetry with the configured product SKU.

## Testing

Extend tests to check reauthentication labels after account switches, product SKU labels for restored sessions, and failed-handshake counts, durations, and labels.

GitOrigin-RevId: 53a38b4460ab9021058dfe2fe104d0441993b18a
2026-11-03 00:59:43 +00:00
rhan-oai 52b4526275 Allow hosted child threads to inherit a root session ID (#59000)
## What changed

Expose `InheritedSessionId` so a host can supply a root session ID through `ExtensionDataInit` when starting a fresh child thread. The child retains its distinct thread ID, and the inherited session ID is persisted in rollout metadata and restored on resume.

Reject inherited session IDs for resumed, forked, or native agent threads, and reject an inherited session ID that equals the child thread ID.

## Testing

Add an integration test covering distinct child identity, inherited session metadata, and restoration after shutdown and resume.

GitOrigin-RevId: e3462433f9d8a0bba24251ee2b6d5329557ddb31
2026-11-02 22:45:22 +00:00
Charlie Marsh 4460449921 Keep HTTP fallback active for the rest of the session (#58989)
## What changed

Once Responses requests fall back to HTTP after exhausting WebSocket retries or receiving HTTP 426, keep subsequent requests and turns on HTTP for the remainder of the session. Clear cached WebSocket state and remove background recovery probes, recovery trials, and their acknowledgement timeout.

## Testing

Update fallback tests to verify that a second turn stays on HTTP without additional WebSocket attempts. Adjust retry and `Retry-After` request-count assertions to exclude background probes.

GitOrigin-RevId: a4f6376604527c14db8d60f79bd21c3aaf289f70
2026-11-02 21:37:17 +00:00
Won Park ac21e58790 Respect delivery boundaries for injected agent messages (#58988)
## Why

Injected agent messages need to reach active sampling before stale tool calls execute, while messages queued around a final answer must wait for the next task without requesting another response.

## What changed

- Include injected `agent_message` items in mailbox preemption checks when the matching turn still accepts mail and `DeferMailboxPreemption` is disabled.
- Treat injected agent messages as queue-only mail so they do not reopen delivery after a final answer or force a follow-up response.

## Testing

Add a streaming scenario and context snapshot covering idle injection, active sampling preemption, deferred delivery around a final answer, and preservation of message payloads and ordering before the next task.

GitOrigin-RevId: 70f49cbe5074725c11e1fcb33eea05e0e92e17f7
2026-11-02 21:35:56 +00:00
rhan-oai 40cf168567 Allow hosts to restrict a thread's root-only capabilities (#58984)
## Why

Hosts need to apply subagent-style capability restrictions without changing a thread's session attribution, agent lineage, or permission profile.

## What changed

Add the public `RestrictRootCapabilities` marker to thread startup extension data. Capture the restriction once at startup and apply it alongside the restrictions native subagents already receive. Hosts must supply the marker again when cold-resuming a restricted thread.

Use the restriction to gate user-input and plugin-install requests, async user messaging, legacy MCP tool approvals, and Full Access MCP form input. Exclude inherited user messages from Guardian's retained authorization context for restricted threads, including after cold resume.

## Testing

Add coverage for host-restricted tool registration, MCP approval rejection, Full Access form-input handling, and inherited instruction authorization after cold resume. Update handler tests to exercise the shared restriction flag.

GitOrigin-RevId: 4d6d7102a20dc747431c7511a74a204c773e8420
2026-11-02 21:09:37 +00:00
rhan-oai c725e472e5 Add usage accounting for host-owned agent runtimes (#58975)
## Why

Host-owned runtimes skip controller rollout-budget accounting. Give hosts an optional callback to account for inference spending and signal budget exhaustion.

## What changed

- Add `AgentUsageAccounting`, `AgentUsageReport`, and `UsageAccountingDecision`, captured at startup and supported by host child spawn reservations.
- Await per-response usage reports with thread and turn attribution, including provider budget units and foreground and background compaction usage.
- Map host exhaustion to `SessionBudgetExceeded` and accounting failures to fatal errors without retrying inference. Preserve the original error for incomplete responses.
- Keep controller accounting unchanged and leave accounting disabled for host-owned runtimes without a provider.

## Testing

Add coverage for usage attribution, child accounting, budget exhaustion, nonretryable accounting failures, incomplete responses, and compaction reporting, including background usage charged once to its originating turn.

GitOrigin-RevId: 887703f50fe882497e4e8cb33259072935ec82e0
2026-11-02 20:37:55 +00:00
rhan-oai d893b14475 Add host-owned turn admission to the agent runtime (#58963)
## What changed

Add `AgentRuntimeOwnership::Host` and an `AgentTurnAdmission` provider so hosts can acquire execution capacity before runtime tasks run, including root turns. Require the provider at startup and retain its execution guard for the task lifetime. Controller ownership remains the default; host ownership skips controller rollout-budget accounting, budget reminders, and subagent roster context.

Keep pending admission interruptible without blocking input submission. Report admission failures through the turn lifecycle, record accepted input before exposing the error, and unwind task setup such as review mode. Preserve recoverable input during suspension. On interruption, skip prompts that still require a `UserPromptSubmit` hook rather than recording them without inspection.

## Testing

Add integration tests for missing-provider startup cleanup, denial and immediate retry, steering during admission, cancellation, execution guard lifetime, submit-hook handling, suspension recovery, and review-mode cleanup. Add a context snapshot verifying retained input reaches the next admitted turn.

GitOrigin-RevId: 9ee85fad2ecfbfb38735bac8ef88888fb179a967
2026-11-02 20:10:37 +00:00
rhan-oai 2fa67f03f3 Allow volatile World State sections to refresh after compaction (#58956)
## Why

Host-provided context such as budget reminders can become stale while compaction runs. Replaying a captured value or retaining its comparison snapshot can prevent the next step from emitting a fresh reminder, even when the remaining budget has changed.

## What changed

- Add `WorldStateSectionContribution::without_full_context()` to omit a section's fragment and comparison snapshot from initial and compacted context rebuilds while preserving ordinary step updates.
- Reset comparison state after prefix compaction for sections excluded at either capture or installation, including when their messages survive in the live suffix.
- Include step World State updates when building debug prompt input.

## Testing

Add regression coverage for budget reminders refreshing after remote compaction within the same threshold, unchanged request prefixes within a context window, refreshes with a retained live suffix, and step reminders in debug prompts.

GitOrigin-RevId: ea4456755e16c93b5d16e8c762b26a33f8571a86
2026-11-02 19:38:59 +00:00
Jiahao f381b64c11 Support caller-owned sideband channels for realtime attachments (#58953)
## What changed

Add `RealtimeSidebandTransport` and `Op::RealtimeConversationAttachStream` so callers can attach Core to an existing realtime call through bounded JSON channels without opening or reconnecting a socket.

Read events independently of outbound backpressure, enforce a 1 MiB message limit, and drain accepted input on detach or channel closure before flushing the transcript. Pending writes are never replayed.

## Testing

Add transport and Core integration tests covering delegation without a callback socket, stalled output, attachment replacement and detach, channel closure, oversized messages, and preservation of queued transcripts in history.

GitOrigin-RevId: 06b7d1dfe2388054438d72d925e510a959dfe106
2026-11-02 19:17:00 +00:00
Charlie Marsh 0a2dda4b5e Recover WebSocket transport after transient HTTP fallback (#58950)
## Why

A transient WebSocket failure previously kept the session on HTTP for its remaining lifetime, even if WebSocket connectivity recovered.

## What changed

- Probe WebSocket connectivity in the background during HTTP fallback, allowing a subsequent request to use a ready connection without delaying HTTP requests.
- Verify recovery with a real request and bound its first-response wait to five seconds. Keep HTTP 426 responses and recovery timeouts on HTTP for the rest of the session.
- Keep retries of an HTTP request on HTTP. If a recovered connection fails before output, replay the request over HTTP without consuming its ordinary retry budget; failures after output use normal retry accounting.
- Rate-limit recovery probes and honor `Retry-After` deadlines.

## Testing

Extend integration coverage for successful recovery within a turn, disconnects before and after output, full-context replay followed by incremental WebSocket requests, persistent HTTP 426 fallback, and retry deadlines.

GitOrigin-RevId: 1c139a4bbce1a67442db5622fd70ff0cc8056c9c
2026-11-02 19:02:50 +00:00
goodman-oai aef7d276a4 Support model catalog overrides for plugin guidance (#58894)
## What changed

- Add `model_messages.plugins` overrides for generic usage guidance, selected-plugin hints, and recommendation introductions, with independent fallback to bundled wording for missing or invalid fields.
- Validate text limits and required template placeholders while keeping plugin identities, available capabilities, and eligibility controlled by the client.
- Track generic guidance changes across model switches and session resume, emitting replacement or retirement notices and supporting legacy snapshots and retained history.

## Testing

Add unit and scenario coverage for sparse overrides, literal text preservation, template validation, UTF-8 size bounds, capability filtering, recommendation fallback, and guidance recovery after resume.

GitOrigin-RevId: 148b658c2d8490c9d0dc208addbee1be5342a5f5
2026-11-02 14:06:17 +00:00
Won Park 05c08d7655 Accept agent follow-ups through native turn input delivery (#58829)
## Why

`start_or_steer_turn` rejected agent messages, preventing callers from submitting follow-up work through the turn input API.

## What changed

Route `TurnInput::InterAgentCommunication` through native agent message delivery and return `TurnInputSubmission::Accepted`, without implying an immediate turn ID. These submissions request work regardless of `trigger_turn`, retain the recipient's settings, and forward turn-start options.

Wrap plaintext messages with agent context while preserving encrypted content. Reject submissions to idle recipients when the host is draining.

## Testing

Add scenario tests and snapshots covering queued-message ordering, encrypted follow-ups, preemption at reasoning boundaries, and follow-ups arriving before or after a final answer. Extend drain-gate coverage to agent submissions with parent-turn lineage.

GitOrigin-RevId: 92d49465f97bb366babbf1b1bc18a1e8a6c6ed1c
2026-11-02 07:27:49 +00:00
Hung 2546265e3e Add audio file references with explicit runtime rejection (#58430)
## What changed

- Allow audio inputs to carry uploaded file references via `fileId` in the app-server API and `file_id` in model content, preserving existing inline audio shapes and Python class names.
- Preserve mixed inline and file-backed audio ordering through history reconstruction and migration, and budget file-backed audio conservatively when duration is unknown.
- Keep uploaded audio disabled at runtime: reject it before inference or raw history injection, preserve accompanying text, and prevent rejected attachments from reaching follow-up or resumed requests.

## Testing

Add coverage for wire-format round trips, SDK compatibility, mixed audio history ordering, and rejection during turn start, steering, task completion, and raw injection.

GitOrigin-RevId: 70895c47a2494c0c9936348ad1250eb1b7b892a6
2026-10-30 20:53:25 +00:00
rhan-oai 6df44f08dd Expose non-consuming input activity waits to hosts (#58411)
## Why

Host-owned tools need to wait for agent messages or steering input with the same behavior as native `wait_agent`, including when native multi-agent support is disabled.

## What changed

- Add `CodexThread::wait_for_input_activity` and export `InputWaitOutcome` through core and core-api, sharing the wait implementation with `wait_agent`.
- Distinguish notifying agent messages from steering input, with steering taking precedence when both are pending. Passive response items do not wake waits.
- Observe input without consuming it. Waits remain active after the original turn ends, support independent observers, and can be cancelled by dropping the future.

## Testing

Add coverage for pending and incoming input, agent-message classification, cancellation, independent observers, timeouts, and input arriving after the original turn completes or is interrupted.

GitOrigin-RevId: d0c0bd776c3b6fd4ba3021603d87eaaf1d93976f
2026-10-30 18:34:10 +00:00
Alex Daley 0e31c29e25 Enable deferred tool world state by default (#58410)
## What changed

Mark `deferred_tool_world_state` as stable and enable it by default. Deferred tool namespaces now appear in the developer `<tools>` context instead of the `tool_search` description. Explicitly disabling the feature preserves discovery metadata in the search description.

## Testing

Update tool discovery tests to cover the default behavior and explicit opt-out, and refresh request snapshots for the deferred namespace context.

GitOrigin-RevId: d58e369ea22754cee593c26a5ca011e41ba56e79
2026-10-30 18:33:14 +00:00
chetan-oai 8cb81f6490 Add an option to suppress incremental model change announcements (#58405)
## What changed

Add the default-off `suppress_model_change_announcements` feature flag. When enabled, incremental context updates omit the model ID announcement after a model switch while still delivering changed base instructions and tool definitions.

Advance the model snapshot even when a switch emits no context update, so subsequent turns recognize the current model.

## Testing

Add coverage for omitted and explicit flag values, snapshot advancement without repeated updates, and request history across model switches, repeated turns, and compaction with announcements suppressed.

GitOrigin-RevId: 89c70f1e9d7981fd945fb8e41b1f98492717a07e
2026-10-30 18:19:38 +00:00
Dmitriy Kovalenko 7a030b9769 Fix context token estimates for responses without usable usage (#58404)
## Why

Retained output from failed responses or responses without usage data can be omitted from context token estimates, delaying automatic compaction even as reasoning accumulates.

## What changed

Track the history boundary covered by the last usable usage report and estimate all items added after it. Spending-only updates preserve that boundary, while successful usage reports cover the request input even when the response contains no output items. Avoid counting unreported reasoning twice and align recomputed context counts and remote compaction estimates with the boundary.

## Testing

Add unit coverage for initial input, empty responses, spending-only updates, and reasoning accounting. Add regression coverage verifying that accumulated unreported reasoning triggers compaction for incomplete responses with or without usage and completed responses without usage.

GitOrigin-RevId: 80c0b800c7236ff40ae53fdd18320e0d06e026ab
2026-10-30 18:12:39 +00:00
rhan-oai cbe16d525c Expose parent context for host-managed child startup (#58390)
## What changed

Replace `PreparedThreadSpawn` with `PreparedChildContext`. `CodexThread::prepare_child` now captures parent inputs from a live tool invocation without requiring a child request. Hosts choose the child's name, role, and history, bind destination services, and apply child settings.

Expose captured instructions, environments, permissions, approval settings, disabled plugins, current inheritable dynamic tools, and an execution-policy snapshot with its original managed requirements. Retain those requirements across MCP configuration refreshes so hosts can assess policy-source compatibility.

Remove `ThreadManager::create_prepared_child`; `CodexThread::create_child` handles reserved child startup on the parent's manager with invocation-bound cancellation and cleanup.

## Testing

Add coverage for execution-policy snapshots and managed requirements after MCP refresh. Update child-spawn tests to verify captured instructions and permissions, current dynamic-tool declarations, and fresh versus forked history.

GitOrigin-RevId: f15b595b44997fa70cd0c9be0e92f470e1f333e1
2026-10-30 17:08:41 +00:00
ioanbudea-OAI dc40446064 Preserve model metadata origin through runtime resolution (#58361)
## What changed

Return `CatalogModel` from model lookups, keeping metadata together with its `Bundled`, `Provider`, `ExplicitConfig`, or `Synthetic` origin through config overrides and runtime snapshots. Replace `used_fallback_model_metadata` checks with explicit `Synthetic` origin checks.

Share the underlying `Arc<ModelInfo>` with extension stores and use copy-on-write for metadata changes. Keep origin tracking outside serialized model metadata.

## Testing

Extend coverage for origin retention across cache reloads, namespaced model lookups, catalog fallback, and config overrides. Verify that step snapshots and extension stores share the same model metadata allocation.

GitOrigin-RevId: bb6130c94a7a795aa168ee24664a1dd453579fba
2026-10-30 15:11:02 +00:00