From 273a4aa4f2474b9d27a2654f5803d51f65ce3f43 Mon Sep 17 00:00:00 2001 From: jif Date: Thu, 11 Jun 2026 11:07:19 +0100 Subject: [PATCH] core: enable remote compaction v2 by default (#27573) ## Why Remote compaction v2 is ready to become the default for providers that already support remote compaction. Leaving it behind an under-development opt-in keeps eligible sessions on the legacy remote-compaction path. This does not broaden provider eligibility: OpenAI and Azure move to v2, while Bedrock and OSS providers retain their existing local-compaction behavior. ## What changed - Mark `remote_compaction_v2` stable and enable it by default. - Make tests that intentionally cover legacy remote compaction explicitly disable v2. - Update parity coverage so v2 exercises the production default and only legacy mode opts out. ## Verification - `just test -p codex-core auto_compact_runs_after_resume_when_token_usage_is_over_limit auto_compact_counts_encrypted_reasoning_before_last_user auto_compact_runs_when_reasoning_header_clears_between_turns responses_lite_compact_request_uses_lite_transport_contract` --- .../app-server/tests/suite/v2/compaction.rs | 3 +- codex-rs/core/tests/suite/compact.rs | 4 ++ codex-rs/core/tests/suite/compact_remote.rs | 8 ++- .../core/tests/suite/compact_remote_parity.rs | 4 +- codex-rs/core/tests/suite/hooks.rs | 68 ++++++++----------- codex-rs/core/tests/suite/responses_lite.rs | 12 ++-- codex-rs/features/src/lib.rs | 4 +- codex-rs/features/src/tests.rs | 10 --- 8 files changed, 55 insertions(+), 58 deletions(-) diff --git a/codex-rs/app-server/tests/suite/v2/compaction.rs b/codex-rs/app-server/tests/suite/v2/compaction.rs index 690c6d8bf..08ddfa4c6 100644 --- a/codex-rs/app-server/tests/suite/v2/compaction.rs +++ b/codex-rs/app-server/tests/suite/v2/compaction.rs @@ -29,6 +29,7 @@ use codex_app_server_protocol::TurnStartParams; use codex_app_server_protocol::TurnStartResponse; use codex_app_server_protocol::UserInput as V2UserInput; use codex_config::types::AuthCredentialsStoreMode; +use codex_features::Feature; use codex_protocol::models::ContentItem; use codex_protocol::models::ResponseItem; use core_test_support::responses; @@ -150,7 +151,7 @@ async fn auto_compaction_remote_emits_started_and_completed_items() -> Result<() write_mock_responses_config_toml( codex_home.path(), &server.uri(), - &BTreeMap::default(), + &BTreeMap::from([(Feature::RemoteCompactionV2, false)]), REMOTE_AUTO_COMPACT_LIMIT, Some(true), "mock_provider", diff --git a/codex-rs/core/tests/suite/compact.rs b/codex-rs/core/tests/suite/compact.rs index c80d7241a..73bc05df4 100644 --- a/codex-rs/core/tests/suite/compact.rs +++ b/codex-rs/core/tests/suite/compact.rs @@ -2015,6 +2015,7 @@ async fn auto_compact_runs_after_resume_when_token_usage_is_over_limit() { let mut builder = test_codex().with_config(move |config| { set_test_compact_prompt(config); config.model_auto_compact_token_limit = Some(limit); + let _ = config.features.disable(Feature::RemoteCompactionV2); }); let initial = builder.build(&server).await.unwrap(); let home = initial.home.clone(); @@ -2043,6 +2044,7 @@ async fn auto_compact_runs_after_resume_when_token_usage_is_over_limit() { let mut resume_builder = test_codex().with_config(move |config| { set_test_compact_prompt(config); config.model_auto_compact_token_limit = Some(limit); + let _ = config.features.disable(Feature::RemoteCompactionV2); }); let resumed = resume_builder .resume(&server, home, rollout_path) @@ -4023,6 +4025,7 @@ async fn auto_compact_counts_encrypted_reasoning_before_last_user() { config.chatgpt_base_url = chatgpt_base_url; set_test_compact_prompt(config); config.model_auto_compact_token_limit = Some(300); + let _ = config.features.disable(Feature::RemoteCompactionV2); }) .build(&server) .await @@ -4145,6 +4148,7 @@ async fn auto_compact_runs_when_reasoning_header_clears_between_turns() { .with_config(|config| { set_test_compact_prompt(config); config.model_auto_compact_token_limit = Some(300); + let _ = config.features.disable(Feature::RemoteCompactionV2); }) .build(&server) .await diff --git a/codex-rs/core/tests/suite/compact_remote.rs b/codex-rs/core/tests/suite/compact_remote.rs index eee445ea9..2a1d459f9 100644 --- a/codex-rs/core/tests/suite/compact_remote.rs +++ b/codex-rs/core/tests/suite/compact_remote.rs @@ -36,7 +36,7 @@ use core_test_support::responses::start_websocket_server; use core_test_support::skip_if_no_network; use core_test_support::test_codex::TestCodexBuilder; use core_test_support::test_codex::TestCodexHarness; -use core_test_support::test_codex::test_codex; +use core_test_support::test_codex::test_codex as base_test_codex; use core_test_support::test_path_buf; use core_test_support::wait_for_event; use core_test_support::wait_for_event_match; @@ -160,6 +160,12 @@ fn compacted_summary_only_output(summary: &str) -> Vec { }] } +fn test_codex() -> TestCodexBuilder { + base_test_codex().with_config(|config| { + let _ = config.features.disable(Feature::RemoteCompactionV2); + }) +} + fn remote_realtime_test_codex_builder( realtime_server: &responses::WebSocketTestServer, ) -> TestCodexBuilder { diff --git a/codex-rs/core/tests/suite/compact_remote_parity.rs b/codex-rs/core/tests/suite/compact_remote_parity.rs index 1b7c50936..81038337d 100644 --- a/codex-rs/core/tests/suite/compact_remote_parity.rs +++ b/codex-rs/core/tests/suite/compact_remote_parity.rs @@ -529,8 +529,8 @@ async fn build_harness_inner( if hooks { trust_discovered_hooks(config); } - if mode == Mode::V2 { - let _ = config.features.enable(Feature::RemoteCompactionV2); + if mode == Mode::Legacy { + let _ = config.features.disable(Feature::RemoteCompactionV2); } })) .await diff --git a/codex-rs/core/tests/suite/hooks.rs b/codex-rs/core/tests/suite/hooks.rs index 16412c2d2..31e54c5e7 100644 --- a/codex-rs/core/tests/suite/hooks.rs +++ b/codex-rs/core/tests/suite/hooks.rs @@ -34,7 +34,6 @@ use core_test_support::responses::ev_function_call; use core_test_support::responses::ev_message_item_added; use core_test_support::responses::ev_output_text_delta; use core_test_support::responses::ev_response_created; -use core_test_support::responses::mount_compact_json_once; use core_test_support::responses::mount_sse_once; use core_test_support::responses::mount_sse_sequence; use core_test_support::responses::sse; @@ -1441,21 +1440,32 @@ async fn resumed_thread_runs_resume_then_compact_session_start_hooks() -> Result let remote_summary = "remote compact summary"; let resume_context = "remember the resumed reef"; let compact_context = "remember the compacted reef"; - let compacted_history = vec![ - ResponseItem::Message { - id: None, - role: "assistant".to_string(), - content: vec![ContentItem::OutputText { - text: remote_summary.to_string(), - }], - phase: None, - }, - ResponseItem::Compaction { - encrypted_content: "encrypted compact summary".to_string(), - }, - ]; - let compact_mock = - mount_compact_json_once(&server, serde_json::json!({ "output": compacted_history })).await; + let responses_mock = mount_sse_sequence( + &server, + vec![ + sse(vec![ + ev_response_created("resp-1"), + ev_assistant_message("msg-1", "hello before resume"), + ev_completed_with_tokens("resp-1", over_limit_tokens), + ]), + sse(vec![ + serde_json::json!({ + "type": "response.output_item.done", + "item": { + "type": "compaction", + "encrypted_content": remote_summary, + } + }), + ev_completed("resp-compact"), + ]), + sse(vec![ + ev_response_created("resp-2"), + ev_assistant_message("msg-2", "hello after resume"), + ev_completed("resp-2"), + ]), + ], + ) + .await; let mut builder = test_codex() .with_pre_build_hook(move |home| { @@ -1479,37 +1489,19 @@ async fn resumed_thread_runs_resume_then_compact_session_start_hooks() -> Result .clone() .context("rollout path")?; - mount_sse_once( - &server, - sse(vec![ - ev_response_created("resp-1"), - ev_assistant_message("msg-1", "hello before resume"), - ev_completed_with_tokens("resp-1", over_limit_tokens), - ]), - ) - .await; initial.submit_turn("hello before resume").await?; - assert!(compact_mock.requests().is_empty()); + assert_eq!(responses_mock.requests().len(), 1); let mut resume_builder = test_codex().with_config(move |config| { config.model_auto_compact_token_limit = Some(limit); trust_discovered_hooks(config); }); let resumed = resume_builder.resume(&server, home, rollout_path).await?; - let follow_up = mount_sse_once( - &server, - sse(vec![ - ev_response_created("resp-2"), - ev_assistant_message("msg-2", "hello after resume"), - ev_completed("resp-2"), - ]), - ) - .await; - resumed.submit_turn("hello after resume").await?; - assert_eq!(compact_mock.requests().len(), 1); - let developer_messages = follow_up.single_request().message_input_texts("developer"); + let requests = responses_mock.requests(); + assert_eq!(requests.len(), 3); + let developer_messages = requests[2].message_input_texts("developer"); assert!( developer_messages .iter() diff --git a/codex-rs/core/tests/suite/responses_lite.rs b/codex-rs/core/tests/suite/responses_lite.rs index 5b47035ba..b137d5bc7 100644 --- a/codex-rs/core/tests/suite/responses_lite.rs +++ b/codex-rs/core/tests/suite/responses_lite.rs @@ -177,10 +177,14 @@ async fn responses_lite_compact_request_uses_lite_transport_contract() -> Result let compact_mock = responses::mount_compact_json_once(&server, serde_json::json!({ "output": [] })).await; - let mut builder = test_codex().with_model_info_override("gpt-5.4", |model_info| { - model_info.use_responses_lite = true; - model_info.supports_parallel_tool_calls = true; - }); + let mut builder = test_codex() + .with_model_info_override("gpt-5.4", |model_info| { + model_info.use_responses_lite = true; + model_info.supports_parallel_tool_calls = true; + }) + .with_config(|config| { + let _ = config.features.disable(Feature::RemoteCompactionV2); + }); let test = builder.build(&server).await?; test.submit_turn("Compact this conversation").await?; diff --git a/codex-rs/features/src/lib.rs b/codex-rs/features/src/lib.rs index afb5bbac2..2f8e91d70 100644 --- a/codex-rs/features/src/lib.rs +++ b/codex-rs/features/src/lib.rs @@ -1249,8 +1249,8 @@ pub const FEATURES: &[FeatureSpec] = &[ FeatureSpec { id: Feature::RemoteCompactionV2, key: "remote_compaction_v2", - stage: Stage::UnderDevelopment, - default_enabled: false, + stage: Stage::Stable, + default_enabled: true, }, FeatureSpec { id: Feature::WorkspaceDependencies, diff --git a/codex-rs/features/src/tests.rs b/codex-rs/features/src/tests.rs index b7d158ca3..0a4f925f4 100644 --- a/codex-rs/features/src/tests.rs +++ b/codex-rs/features/src/tests.rs @@ -150,16 +150,6 @@ fn request_permissions_tool_is_under_development() { assert_eq!(Feature::RequestPermissionsTool.default_enabled(), false); } -#[test] -fn remote_compaction_v2_is_under_development() { - assert_eq!(Feature::RemoteCompactionV2.stage(), Stage::UnderDevelopment); - assert_eq!(Feature::RemoteCompactionV2.default_enabled(), false); - assert_eq!( - feature_for_key("remote_compaction_v2"), - Some(Feature::RemoteCompactionV2) - ); -} - #[test] fn terminal_resize_reflow_is_experimental_and_enabled_by_default() { assert_eq!(