From 57ebc4a06f19290540fa2630b31d183ca30c3da3 Mon Sep 17 00:00:00 2001
From: KT <huangjie@evermind.ai>
Date: Tue, 28 Jul 2026 10:57:59 +0800
Subject: [PATCH 1/8] test: run the full unit suite in ci and fix revealed
 failures

CI ran `make test-python` over only 4 hardcoded files, so ~268 of 272
top-level test files never executed in CI and could red-rot unnoticed.
Bare pytest also errored at collection because 7 channel-adapter tests
import optional SDKs at module top level.

- Makefile: test-python now runs the whole unit tree (uv run --all-extras pytest)
- ci.yml: the python job becomes a blocking unit matrix (py3.11-3.13 + macos)
- pyproject: register integration/e2e markers, --strict-markers/--strict-config,
  norecursedirs + default -m "not integration and not e2e"
- guard the 7 channel tests with pytest.importorskip
- relocate 9 integration/benchmark files under tests/integration/ (real_llm marked)
- fix the 44 pre-existing failures the full run revealed: get_event_loop()
  anti-pattern in producer tests (now asyncio.run), stale run_turn stubs missing
  inline_tool_stream, provider_catalog missing LazyProvider/PerModelProvider,
  config-path isolation in tui_rpc_model / tui_commands_error_codes /
  logging_isolation, and a flaky concurrency timing bound

Full unit suite: 4585 passed, 24 deselected (tui e2e), 0 failed.

Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
---
 .github/workflows/ci.yml                      | 34 ++++++++++++++++---
 Makefile                                      |  2 +-
 pyproject.toml                                | 12 ++++++-
 .../test_evolver_launch_e2e.py                |  0
 .../test_sandbox_cli_integration.py           |  0
 .../test_sandbox_cli_real_vm.py               |  0
 .../test_sandbox_integration.py               |  0
 tests/{ => integration}/test_sentinel_e2e.py  |  0
 .../test_token_wise_ablation_real_llm.py}     |  6 ++--
 ...st_token_wise_hermes_vs_raven_real_llm.py} |  8 +++--
 .../test_token_wise_openrouter_real_llm.py}   |  6 ++--
 .../test_token_wise_workload_real_llm.py}     |  6 ++--
 tests/test_channels_dingtalk.py               |  3 ++
 tests/test_channels_feishu.py                 |  2 ++
 tests/test_channels_matrix.py                 |  2 ++
 tests/test_channels_qq.py                     |  4 +++
 tests/test_channels_slack.py                  |  4 +++
 tests/test_channels_telegram.py               |  2 ++
 tests/test_channels_wecom.py                  |  4 +++
 tests/test_cli_cron_handler.py                |  2 +-
 tests/test_cli_tui_logging_isolation.py       |  8 +++++
 tests/test_currently_focused.py               |  2 +-
 tests/test_daily_analysis.py                  |  2 +-
 tests/test_default_context_engine.py          |  8 ++---
 tests/test_sentinel_observations_producer.py  |  2 +-
 tests/test_tui_commands_error_codes.py        | 14 ++++++++
 tests/test_tui_rpc_model.py                   |  3 ++
 tests/test_turn_wire_shape_conformance.py     |  4 ++-
 28 files changed, 117 insertions(+), 23 deletions(-)
 rename tests/{ => integration}/test_evolver_launch_e2e.py (100%)
 rename tests/{ => integration}/test_sandbox_cli_integration.py (100%)
 rename tests/{ => integration}/test_sandbox_cli_real_vm.py (100%)
 rename tests/{ => integration}/test_sandbox_integration.py (100%)
 rename tests/{ => integration}/test_sentinel_e2e.py (100%)
 rename tests/{test_token_wise_agentloop_experiment.py => integration/test_token_wise_ablation_real_llm.py} (98%)
 rename tests/{test_token_wise_hermes_vs_raven.py => integration/test_token_wise_hermes_vs_raven_real_llm.py} (98%)
 rename tests/{test_token_wise_integration_openrouter.py => integration/test_token_wise_openrouter_real_llm.py} (98%)
 rename tests/{test_token_wise_workload_scenarios.py => integration/test_token_wise_workload_real_llm.py} (99%)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index d0d2c9697..eed97bb73 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -46,8 +46,8 @@ jobs:
           fi
           uv run pre-commit run --from-ref "$BASE" --to-ref HEAD
 
-  python:
-    name: python checks
+  lint-python:
+    name: python lint
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v7
@@ -67,8 +67,34 @@ jobs:
       - name: Focused Python lint
         run: make lint-python
 
-      - name: Focused Python tests
-        run: make test-python
+  unit:
+    name: unit (py${{ matrix.python }} / ${{ matrix.os }})
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest]
+        python: ["3.11", "3.12", "3.13"]
+        include:
+          - os: macos-latest
+            python: "3.12"
+    steps:
+      - uses: actions/checkout@v7
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v8.2.0
+        with:
+          enable-cache: true
+          cache-dependency-glob: uv.lock
+
+      - name: Set up Python
+        run: uv python install ${{ matrix.python }}
+
+      - name: Install dependencies
+        run: uv sync --locked --all-extras --dev
+
+      - name: Unit tests (integration + e2e deselected by norecursedirs + addopts)
+        run: uv run --all-extras pytest -q
 
   windows-upgrade:
     name: Windows self-upgrade
diff --git a/Makefile b/Makefile
index 6742f9eb7..aa5019894 100644
--- a/Makefile
+++ b/Makefile
@@ -46,7 +46,7 @@ lint-bridge:
 test: test-python test-tui
 
 test-python:
-	uv run --extra dev pytest tests/test_commit_lint.py tests/test_large_file_check.py tests/test_cli_smoke.py tests/test_litellm_setup.py -q
+	uv run --all-extras pytest -q
 
 test-tui:
 	npm test --prefix ui-tui
diff --git a/pyproject.toml b/pyproject.toml
index f273aaac3..eb54f691e 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -255,9 +255,19 @@ ignore = ["E501"]
 "benchmarks/proactivity_eval/runners/synthesizers/keyword.py" = ["E741"]
 
 [tool.pytest.ini_options]
-asyncio_mode = "auto"
 testpaths = ["tests"]
+norecursedirs = ["tests/integration", "tests/data", "*.egg", ".*", "build", "dist"]
+addopts = [
+    "-ra",
+    "--strict-markers",
+    "--strict-config",
+    "-m",
+    "not integration and not e2e",
+]
+asyncio_mode = "auto"
 markers = [
+    "integration: needs a real external resource (LLM / RPC subprocess / VM / network); deselected by default, run with '-m integration' or by pointing at tests/integration",
+    "e2e: spawns real subprocesses / drives the full stack (slow); deselected by default, run with '-m e2e' (e.g. tests/tui/autotest)",
     "real_llm: hits a real everos runtime (LLM + embedding); skipped unless EVEROS_LLM__API_KEY / EVEROS_EMBEDDING__MODEL are configured",
     "llm_judge: uses an LLM as a grader for extraction quality (extra cost / non-determinism); opt-in",
 ]
diff --git a/tests/test_evolver_launch_e2e.py b/tests/integration/test_evolver_launch_e2e.py
similarity index 100%
rename from tests/test_evolver_launch_e2e.py
rename to tests/integration/test_evolver_launch_e2e.py
diff --git a/tests/test_sandbox_cli_integration.py b/tests/integration/test_sandbox_cli_integration.py
similarity index 100%
rename from tests/test_sandbox_cli_integration.py
rename to tests/integration/test_sandbox_cli_integration.py
diff --git a/tests/test_sandbox_cli_real_vm.py b/tests/integration/test_sandbox_cli_real_vm.py
similarity index 100%
rename from tests/test_sandbox_cli_real_vm.py
rename to tests/integration/test_sandbox_cli_real_vm.py
diff --git a/tests/test_sandbox_integration.py b/tests/integration/test_sandbox_integration.py
similarity index 100%
rename from tests/test_sandbox_integration.py
rename to tests/integration/test_sandbox_integration.py
diff --git a/tests/test_sentinel_e2e.py b/tests/integration/test_sentinel_e2e.py
similarity index 100%
rename from tests/test_sentinel_e2e.py
rename to tests/integration/test_sentinel_e2e.py
diff --git a/tests/test_token_wise_agentloop_experiment.py b/tests/integration/test_token_wise_ablation_real_llm.py
similarity index 98%
rename from tests/test_token_wise_agentloop_experiment.py
rename to tests/integration/test_token_wise_ablation_real_llm.py
index 22ea11c62..deced2827 100644
--- a/tests/test_token_wise_agentloop_experiment.py
+++ b/tests/integration/test_token_wise_ablation_real_llm.py
@@ -41,8 +41,10 @@
 from raven.token_wise.registry import StrategyRegistry
 from raven.token_wise.usage_tracker import UsageTracker
 
-KEY_FILE = Path(__file__).resolve().parent.parent / "raven" / "key.env"
-REPORT_PATH = Path(__file__).resolve().parent.parent / "raven" / "token_wise" / "EXPERIMENT_REPORT.md"
+pytestmark = pytest.mark.real_llm
+
+KEY_FILE = Path(__file__).resolve().parent.parent.parent / "raven" / "key.env"
+REPORT_PATH = Path(__file__).resolve().parent.parent.parent / "raven" / "token_wise" / "EXPERIMENT_REPORT.md"
 MODEL = "anthropic/claude-sonnet-4-5"
 TURNS = 6
 COST_GUARD_USD = 0.50
diff --git a/tests/test_token_wise_hermes_vs_raven.py b/tests/integration/test_token_wise_hermes_vs_raven_real_llm.py
similarity index 98%
rename from tests/test_token_wise_hermes_vs_raven.py
rename to tests/integration/test_token_wise_hermes_vs_raven_real_llm.py
index d07689a69..c4cdc5242 100644
--- a/tests/test_token_wise_hermes_vs_raven.py
+++ b/tests/integration/test_token_wise_hermes_vs_raven_real_llm.py
@@ -45,8 +45,12 @@
 from raven.token_wise.system_and_tail_cache import SystemAndTailCacheStrategy
 from raven.token_wise.usage_tracker import UsageTracker
 
-KEY_FILE = Path(__file__).resolve().parent.parent / "raven" / "key.env"
-REPORT_PATH = Path(__file__).resolve().parent.parent / "raven" / "token_wise" / "EXPERIMENT_REPORT_HERMES_VS_RAVEN.md"
+pytestmark = pytest.mark.real_llm
+
+KEY_FILE = Path(__file__).resolve().parent.parent.parent / "raven" / "key.env"
+REPORT_PATH = (
+    Path(__file__).resolve().parent.parent.parent / "raven" / "token_wise" / "EXPERIMENT_REPORT_HERMES_VS_RAVEN.md"
+)
 MODEL = "anthropic/claude-sonnet-4-5"
 COST_GUARD_USD = 2.00
 _OPENROUTER_PIN = {"provider": {"order": ["Anthropic"], "allow_fallbacks": False}}
diff --git a/tests/test_token_wise_integration_openrouter.py b/tests/integration/test_token_wise_openrouter_real_llm.py
similarity index 98%
rename from tests/test_token_wise_integration_openrouter.py
rename to tests/integration/test_token_wise_openrouter_real_llm.py
index 0dca7582a..13db0c2ab 100644
--- a/tests/test_token_wise_integration_openrouter.py
+++ b/tests/integration/test_token_wise_openrouter_real_llm.py
@@ -34,8 +34,10 @@
 from raven.providers.litellm_provider import LiteLLMProvider
 from raven.token_wise.registry import StrategyRegistry
 
-KEY_FILE = Path(__file__).resolve().parent.parent / "raven" / "key.env"
-REPORT_PATH = Path(__file__).resolve().parent.parent / "raven" / "token_wise" / "EXPERIMENT_REPORT.md"
+pytestmark = pytest.mark.real_llm
+
+KEY_FILE = Path(__file__).resolve().parent.parent.parent / "raven" / "key.env"
+REPORT_PATH = Path(__file__).resolve().parent.parent.parent / "raven" / "token_wise" / "EXPERIMENT_REPORT.md"
 MODEL = "anthropic/claude-sonnet-4-5"
 TURNS = 6
 COST_GUARD_USD = 0.50  # hard cap; abort if we go over
diff --git a/tests/test_token_wise_workload_scenarios.py b/tests/integration/test_token_wise_workload_real_llm.py
similarity index 99%
rename from tests/test_token_wise_workload_scenarios.py
rename to tests/integration/test_token_wise_workload_real_llm.py
index 979add2e4..ded57d347 100644
--- a/tests/test_token_wise_workload_scenarios.py
+++ b/tests/integration/test_token_wise_workload_real_llm.py
@@ -47,8 +47,10 @@
 from raven.token_wise.registry import StrategyRegistry
 from raven.token_wise.usage_tracker import UsageTracker
 
-KEY_FILE = Path(__file__).resolve().parent.parent / "raven" / "key.env"
-REPORT_PATH = Path(__file__).resolve().parent.parent / "raven" / "token_wise" / "EXPERIMENT_REPORT_WORKLOADS.md"
+pytestmark = pytest.mark.real_llm
+
+KEY_FILE = Path(__file__).resolve().parent.parent.parent / "raven" / "key.env"
+REPORT_PATH = Path(__file__).resolve().parent.parent.parent / "raven" / "token_wise" / "EXPERIMENT_REPORT_WORKLOADS.md"
 MODEL = "anthropic/claude-sonnet-4-5"
 COST_GUARD_USD = 1.50
 _OPENROUTER_PIN = {"provider": {"order": ["Anthropic"], "allow_fallbacks": False}}
diff --git a/tests/test_channels_dingtalk.py b/tests/test_channels_dingtalk.py
index f2d12449e..7df29f438 100644
--- a/tests/test_channels_dingtalk.py
+++ b/tests/test_channels_dingtalk.py
@@ -8,6 +8,9 @@
 from unittest.mock import AsyncMock, MagicMock
 
 import pytest
+
+pytest.importorskip("dingtalk_stream")
+
 from dingtalk_stream import AckMessage
 
 from raven.channels.adapters.dingtalk import parsing as p
diff --git a/tests/test_channels_feishu.py b/tests/test_channels_feishu.py
index 8b3484128..b6694115c 100644
--- a/tests/test_channels_feishu.py
+++ b/tests/test_channels_feishu.py
@@ -8,6 +8,8 @@
 
 import pytest
 
+pytest.importorskip("lark_oapi")
+
 from raven.channels.adapters.feishu import cards, content
 from raven.channels.adapters.feishu.channel import FeishuChannel
 
diff --git a/tests/test_channels_matrix.py b/tests/test_channels_matrix.py
index 1a0566b52..0b847b071 100644
--- a/tests/test_channels_matrix.py
+++ b/tests/test_channels_matrix.py
@@ -9,6 +9,8 @@
 
 import pytest
 
+pytest.importorskip("nio")
+
 from raven.channels.adapters.matrix import content
 from raven.channels.adapters.matrix.channel import MatrixChannel
 
diff --git a/tests/test_channels_qq.py b/tests/test_channels_qq.py
index 2ebe75ac8..81429542c 100644
--- a/tests/test_channels_qq.py
+++ b/tests/test_channels_qq.py
@@ -11,6 +11,10 @@
 from types import SimpleNamespace
 from unittest.mock import AsyncMock, MagicMock
 
+import pytest
+
+pytest.importorskip("botpy")
+
 from raven.channels.adapters.qq import parsing as qp
 from raven.channels.adapters.qq.channel import QQChannel
 
diff --git a/tests/test_channels_slack.py b/tests/test_channels_slack.py
index 4ed8f993f..3f888b996 100644
--- a/tests/test_channels_slack.py
+++ b/tests/test_channels_slack.py
@@ -11,6 +11,10 @@
 from types import SimpleNamespace
 from unittest.mock import AsyncMock, MagicMock
 
+import pytest
+
+pytest.importorskip("slack_sdk")
+
 from raven.channels.adapters.slack import parsing as sp
 from raven.channels.adapters.slack.channel import SlackChannel
 
diff --git a/tests/test_channels_telegram.py b/tests/test_channels_telegram.py
index a48df2ccc..5dce98619 100644
--- a/tests/test_channels_telegram.py
+++ b/tests/test_channels_telegram.py
@@ -8,6 +8,8 @@
 
 import pytest
 
+pytest.importorskip("telegram")
+
 from raven.channels.adapters.telegram.channel import (
     TelegramChannel,
     _inbound_ext,
diff --git a/tests/test_channels_wecom.py b/tests/test_channels_wecom.py
index c5e7f5345..7d3086e92 100644
--- a/tests/test_channels_wecom.py
+++ b/tests/test_channels_wecom.py
@@ -6,6 +6,10 @@
 from types import SimpleNamespace
 from unittest.mock import AsyncMock
 
+import pytest
+
+pytest.importorskip("wecom_aibot_sdk")
+
 from raven.channels.adapters.wecom.channel import WecomChannel
 
 
diff --git a/tests/test_cli_cron_handler.py b/tests/test_cli_cron_handler.py
index b73524ecb..6e4706273 100644
--- a/tests/test_cli_cron_handler.py
+++ b/tests/test_cli_cron_handler.py
@@ -498,7 +498,7 @@ async def test_repl_assembly_cron_renders_once_via_clioutlet_not_broadcast(fake_
     patch_cron_config(["*"])
 
     class _CronEchoLoop:
-        async def run_turn(self, req, emit, drain, *, stream) -> TurnOutcome:
+        async def run_turn(self, req, emit, drain, *, stream, inline_tool_stream=False) -> TurnOutcome:
             await emit(Text(content=f"cron-reply<{req.conversation}>", source=req.source))
             return TurnOutcome(usage=Usage(0, 0, 0), explicit_reply=True)
 
diff --git a/tests/test_cli_tui_logging_isolation.py b/tests/test_cli_tui_logging_isolation.py
index de1f99624..fdc3963f4 100644
--- a/tests/test_cli_tui_logging_isolation.py
+++ b/tests/test_cli_tui_logging_isolation.py
@@ -65,6 +65,14 @@ def _isolate_logging(tmp_path, monkeypatch):
     saved_named_handlers = {name: list(logging.getLogger(name).handlers) for name in _KNOWN_TTY_LEAKING_LOGGERS}
     saved_disable = logging.root.manager.disable
 
+    # litellm's stderr handlers are global and installed once at litellm import; a
+    # prior test that called redirect_loguru_to_file may have stripped them. Ensure
+    # the precondition (a stderr handler exists to be stripped) holds regardless of
+    # suite order — the finally-block restores the snapshot, dropping what we add.
+    for name in _KNOWN_TTY_LEAKING_LOGGERS:
+        if not _tty_stream_handlers(name):
+            logging.getLogger(name).addHandler(logging.StreamHandler(sys.stderr))
+
     try:
         yield
     finally:
diff --git a/tests/test_currently_focused.py b/tests/test_currently_focused.py
index 86b12459e..2536e0900 100644
--- a/tests/test_currently_focused.py
+++ b/tests/test_currently_focused.py
@@ -90,7 +90,7 @@ def _seed_episodes(
 
 
 def _run(coro):
-    return asyncio.get_event_loop().run_until_complete(coro)
+    return asyncio.run(coro)
 
 
 # ===========================================================================
diff --git a/tests/test_daily_analysis.py b/tests/test_daily_analysis.py
index 73bb7c85c..839d07f7e 100644
--- a/tests/test_daily_analysis.py
+++ b/tests/test_daily_analysis.py
@@ -78,7 +78,7 @@ def config() -> DailyAnalysisConfig:
 
 
 def _run(coro):
-    return asyncio.get_event_loop().run_until_complete(coro)
+    return asyncio.run(coro)
 
 
 def _provider_with_result(payload: dict) -> MagicMock:
diff --git a/tests/test_default_context_engine.py b/tests/test_default_context_engine.py
index f792ce4da..3b92e5714 100644
--- a/tests/test_default_context_engine.py
+++ b/tests/test_default_context_engine.py
@@ -195,8 +195,8 @@ async def test_skill_and_memory_run_concurrently(
         self,
         builder: ContextBuilder,
     ) -> None:
-        slow_source = _StubSource("local", hits=[], delay_s=0.10)
-        slow_backend = _StubBackend(recall_response=[], delay_s=0.10)
+        slow_source = _StubSource("local", hits=[], delay_s=0.20)
+        slow_backend = _StubBackend(recall_response=[], delay_s=0.20)
         eng = _engine(
             builder,
             router=SkillForgeRouter([slow_source]),
@@ -210,8 +210,8 @@ async def test_skill_and_memory_run_concurrently(
             turn=_turn(),
         )
         elapsed = time.monotonic() - t0
-        # Serial would be ~0.20 s. Concurrent ~0.10 s. Loose bound 0.15.
-        assert elapsed < 0.15
+        # Serial would be ~0.40 s. Concurrent ~0.20 s. Loose bound 0.30.
+        assert elapsed < 0.30
 
     async def test_track_ids_passed_to_recall(
         self,
diff --git a/tests/test_sentinel_observations_producer.py b/tests/test_sentinel_observations_producer.py
index 0143e9bb9..738494bb4 100644
--- a/tests/test_sentinel_observations_producer.py
+++ b/tests/test_sentinel_observations_producer.py
@@ -119,7 +119,7 @@ def _seed_topic_fires(
 
 
 def _run(updater: AttentionUpdater):
-    return asyncio.get_event_loop().run_until_complete(updater.update())
+    return asyncio.run(updater.update())
 
 
 # ===========================================================================
diff --git a/tests/test_tui_commands_error_codes.py b/tests/test_tui_commands_error_codes.py
index 488ecc069..e33756aae 100644
--- a/tests/test_tui_commands_error_codes.py
+++ b/tests/test_tui_commands_error_codes.py
@@ -19,6 +19,20 @@
 # ---------------------------------------------------------------------------
 
 
+@pytest.fixture(autouse=True)
+def _hermetic_preagentloop(monkeypatch):
+    """Isolate the AgentLoop-ctor error path from ambient config.
+
+    _build_tui_agent_loop builds a lazy provider before constructing AgentLoop;
+    with no provider configured (clean CI, or a prior test that reset config
+    state) make_lazy_provider raises typer.Exit(1) before AgentLoop is reached,
+    so the ctor-error classification under test never runs. Stub it so these
+    tests exercise only what they claim to."""
+    from unittest.mock import MagicMock
+
+    monkeypatch.setattr("raven.cli._helpers.make_lazy_provider", lambda *a, **k: MagicMock())
+
+
 def _patch_agent_loop_to_raise(monkeypatch: pytest.MonkeyPatch, exc: BaseException) -> None:
     """Replace ``AgentLoop`` constructor so that calling it raises ``exc``.
 
diff --git a/tests/test_tui_rpc_model.py b/tests/test_tui_rpc_model.py
index 72197f597..fe3128419 100644
--- a/tests/test_tui_rpc_model.py
+++ b/tests/test_tui_rpc_model.py
@@ -27,6 +27,9 @@
 @pytest.fixture
 def fake_home(monkeypatch, tmp_path) -> Path:
     monkeypatch.setattr(Path, "home", lambda: tmp_path)
+    # Clear any process-wide config-path override a prior test left set, so
+    # get_config_path() falls back to the patched Path.home (monkeypatch restores it).
+    monkeypatch.setattr("raven.config.loader._current_config_path", None)
     return tmp_path
 
 
diff --git a/tests/test_turn_wire_shape_conformance.py b/tests/test_turn_wire_shape_conformance.py
index 2f67f3215..47eff934d 100644
--- a/tests/test_turn_wire_shape_conformance.py
+++ b/tests/test_turn_wire_shape_conformance.py
@@ -85,7 +85,9 @@ async def test_message_complete_payload_has_turn_id_and_usage_only() -> None:
     class _StubAgent:
         tools: dict = {}
 
-        async def run_turn(self, req, emit, drain, *, stream, usage_sink=None, text_sink=None):
+        async def run_turn(
+            self, req, emit, drain, *, stream, inline_tool_stream=False, usage_sink=None, text_sink=None
+        ):
             from raven.spine import Text, TurnOutcome, Usage
 
             await emit(Text(content="stub content (must not appear on wire)", source=req.source))

From c732541f4006c81bc47d458fa3d20003d9058968 Mon Sep 17 00:00:00 2001
From: KT <huangjie@evermind.ai>
Date: Tue, 28 Jul 2026 12:03:30 +0800
Subject: [PATCH 2/8] test: add temporary ci diagnostic and fix windows job
 pytest-asyncio

Temporary diagnostic to capture the CI console environment; will be
removed before merge. Also fixes the Windows self-upgrade job, whose
isolated pytest hit --strict-config on the unknown asyncio_mode option.

Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
---
 .github/workflows/ci.yml |  5 ++++-
 tests/test_zz_ci_diag.py | 24 ++++++++++++++++++++++++
 2 files changed, 28 insertions(+), 1 deletion(-)
 create mode 100644 tests/test_zz_ci_diag.py

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index eed97bb73..8fe9781bb 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -109,7 +109,10 @@ jobs:
         run: uv python install 3.12
 
       - name: Test real uv self-upgrade
-        run: uvx --python 3.12 --from "pytest>=8,<10" pytest --noconftest tests/integration/test_cli_upgrade_real_uv.py -q
+        # pytest-asyncio is needed even though this file has no async test: the
+        # isolated interpreter still reads the project's pytest config, whose
+        # --strict-config turns the then-unknown asyncio_mode into a hard error.
+        run: uvx --python 3.12 --from "pytest>=8,<10" --with pytest-asyncio pytest --noconftest tests/integration/test_cli_upgrade_real_uv.py -q
 
   tui:
     name: TUI checks
diff --git a/tests/test_zz_ci_diag.py b/tests/test_zz_ci_diag.py
new file mode 100644
index 000000000..a029e20c1
--- /dev/null
+++ b/tests/test_zz_ci_diag.py
@@ -0,0 +1,24 @@
+"""Temporary CI diagnostic; removed before merge."""
+
+from __future__ import annotations
+
+import os
+import sys
+
+from typer.testing import CliRunner
+
+from raven.cli.commands import app
+
+
+def test_zz_ci_diag() -> None:
+    from rich.console import Console
+
+    env = {k: os.environ.get(k) for k in ("TERM", "COLORTERM", "NO_COLOR", "FORCE_COLOR", "CI", "COLUMNS", "LANG")}
+    con = Console()
+    r = CliRunner().invoke(app, ["doctor", "--help"])
+    print(f"DIAG env={env}", file=sys.stderr)
+    print(f"DIAG console width={con.width} color={con.color_system} term={con.is_terminal}", file=sys.stderr)
+    print(f"DIAG exit={r.exit_code} len={len(r.stdout)}", file=sys.stderr)
+    print(f"DIAG stdout={r.stdout[:600]!r}", file=sys.stderr)
+    if r.exception:
+        print(f"DIAG exception={r.exception!r}", file=sys.stderr)

From 711a2e38c5e8d5211c3376bf072e9cdee92bedee Mon Sep 17 00:00:00 2001
From: KT <huangjie@evermind.ai>
Date: Tue, 28 Jul 2026 12:08:49 +0800
Subject: [PATCH 3/8] test: make ci diagnostic fail so its output is captured

Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
---
 tests/test_zz_ci_diag.py | 11 ++++-------
 1 file changed, 4 insertions(+), 7 deletions(-)

diff --git a/tests/test_zz_ci_diag.py b/tests/test_zz_ci_diag.py
index a029e20c1..a92415b92 100644
--- a/tests/test_zz_ci_diag.py
+++ b/tests/test_zz_ci_diag.py
@@ -3,7 +3,6 @@
 from __future__ import annotations
 
 import os
-import sys
 
 from typer.testing import CliRunner
 
@@ -16,9 +15,7 @@ def test_zz_ci_diag() -> None:
     env = {k: os.environ.get(k) for k in ("TERM", "COLORTERM", "NO_COLOR", "FORCE_COLOR", "CI", "COLUMNS", "LANG")}
     con = Console()
     r = CliRunner().invoke(app, ["doctor", "--help"])
-    print(f"DIAG env={env}", file=sys.stderr)
-    print(f"DIAG console width={con.width} color={con.color_system} term={con.is_terminal}", file=sys.stderr)
-    print(f"DIAG exit={r.exit_code} len={len(r.stdout)}", file=sys.stderr)
-    print(f"DIAG stdout={r.stdout[:600]!r}", file=sys.stderr)
-    if r.exception:
-        print(f"DIAG exception={r.exception!r}", file=sys.stderr)
+    raise AssertionError(
+        f"DIAG env={env} | console width={con.width} color={con.color_system} term={con.is_terminal} "
+        f"| exit={r.exit_code} len={len(r.stdout)} exc={r.exception!r} | stdout={r.stdout[:800]!r}"
+    )

From 629ffce95b463856d4dd7c752ee32eaf39b59ec4 Mon Sep 17 00:00:00 2001
From: KT <huangjie@evermind.ai>
Date: Tue, 28 Jul 2026 12:21:56 +0800
Subject: [PATCH 4/8] ci: keep colour out of the unit job and drop the python
 matrix

The first full-suite CI run failed 16 CLI/theme tests that pass locally.
Typer's rich_utils force-enables colour when GITHUB_ACTIONS is set, and its
option highlighter then emits escape codes between the two dashes of a flag,
so assertions like `"--probe" in result.stdout` cannot match. Setting
TERM=dumb for the unit job restores plain rendering; tests that assert on
colour build their own forced-terminal Console and are unaffected.

Also drop the version/OS matrix down to a single ubuntu / py3.12 cell:
pyproject declares requires-python >= 3.12, so the 3.11 cell was testing an
unsupported version, and all four cells reported identical results. The
matrix block is kept so more versions are one list entry away. Windows keeps
its dedicated self-upgrade job.

Removes the temporary diagnostic test added to capture the CI environment.

Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
---
 .github/workflows/ci.yml | 10 ++++++----
 tests/test_zz_ci_diag.py | 21 ---------------------
 2 files changed, 6 insertions(+), 25 deletions(-)
 delete mode 100644 tests/test_zz_ci_diag.py

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 8fe9781bb..8a28f4c4e 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -74,10 +74,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-latest]
-        python: ["3.11", "3.12", "3.13"]
-        include:
-          - os: macos-latest
-            python: "3.12"
+        python: ["3.12"]
     steps:
       - uses: actions/checkout@v7
 
@@ -94,6 +91,11 @@ jobs:
         run: uv sync --locked --all-extras --dev
 
       - name: Unit tests (integration + e2e deselected by norecursedirs + addopts)
+        # TERM=dumb: typer's rich_utils force-enables colour when GITHUB_ACTIONS is
+        # set, and the highlighter then splits option names with escape codes, so
+        # `"--flag" in result.stdout` assertions fail only on CI.
+        env:
+          TERM: dumb
         run: uv run --all-extras pytest -q
 
   windows-upgrade:
diff --git a/tests/test_zz_ci_diag.py b/tests/test_zz_ci_diag.py
deleted file mode 100644
index a92415b92..000000000
--- a/tests/test_zz_ci_diag.py
+++ /dev/null
@@ -1,21 +0,0 @@
-"""Temporary CI diagnostic; removed before merge."""
-
-from __future__ import annotations
-
-import os
-
-from typer.testing import CliRunner
-
-from raven.cli.commands import app
-
-
-def test_zz_ci_diag() -> None:
-    from rich.console import Console
-
-    env = {k: os.environ.get(k) for k in ("TERM", "COLORTERM", "NO_COLOR", "FORCE_COLOR", "CI", "COLUMNS", "LANG")}
-    con = Console()
-    r = CliRunner().invoke(app, ["doctor", "--help"])
-    raise AssertionError(
-        f"DIAG env={env} | console width={con.width} color={con.color_system} term={con.is_terminal} "
-        f"| exit={r.exit_code} len={len(r.stdout)} exc={r.exception!r} | stdout={r.stdout[:800]!r}"
-    )

From 15c727bbb256f8c8805f777aeafafe0f38f67e3a Mon Sep 17 00:00:00 2001
From: KT <huangjie@evermind.ai>
Date: Tue, 28 Jul 2026 12:36:50 +0800
Subject: [PATCH 5/8] test: spawn subprocesses explicitly and widen the unit
 matrix to three OSes

The first green-tests CI run still exited 139: three cross-process locking
tests start workers with the platform default start method, which is fork on
Linux. pytest leaves the parent multi-threaded, so the forked children left
the interpreter segfaulting at exit. They now use an explicit spawn context,
matching what macOS already did and what CPython is moving toward.

The unit matrix gains macos-latest and windows-latest on py3.12. macOS is
expected to pass -- its cell reported the same failures as ubuntu before the
colour fix and never segfaulted. Windows has never run this suite, so this
run is a measurement: fail-fast is off, and the ubuntu and macos cells report
independently.

Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
---
 .github/workflows/ci.yml      |  2 +-
 tests/test_atomic_io.py       |  5 ++++-
 tests/test_memory_locking.py  | 12 ++++++++----
 tests/test_session_manager.py |  5 ++++-
 4 files changed, 17 insertions(+), 7 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 8a28f4c4e..fccb1d8dd 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -73,7 +73,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest]
+        os: [ubuntu-latest, macos-latest, windows-latest]
         python: ["3.12"]
     steps:
       - uses: actions/checkout@v7
diff --git a/tests/test_atomic_io.py b/tests/test_atomic_io.py
index e3b18d778..3a45e2091 100644
--- a/tests/test_atomic_io.py
+++ b/tests/test_atomic_io.py
@@ -38,7 +38,10 @@ def test_locked_append_concurrent_writers_lose_nothing(tmp_path: Path):
     """Two processes appending concurrently: every line lands, and the
     lines of one locked_append call stay contiguous (turn-block invariant)."""
     path = tmp_path / "s.jsonl"
-    procs = [multiprocessing.Process(target=_append_worker, args=(str(path), w)) for w in range(WRITERS)]
+    # spawn, not the Linux default fork: pytest leaves the parent multi-threaded,
+    # and forking from there segfaults the interpreter at exit.
+    ctx = multiprocessing.get_context("spawn")
+    procs = [ctx.Process(target=_append_worker, args=(str(path), w)) for w in range(WRITERS)]
     for p in procs:
         p.start()
     for p in procs:
diff --git a/tests/test_memory_locking.py b/tests/test_memory_locking.py
index 98078694b..0c9aac7b1 100644
--- a/tests/test_memory_locking.py
+++ b/tests/test_memory_locking.py
@@ -5,9 +5,10 @@
 
 from __future__ import annotations
 
+import multiprocessing
 import sys
 import time
-from multiprocessing import Process, Value
+from multiprocessing import Value
 from pathlib import Path
 
 import pytest
@@ -72,9 +73,12 @@ def test_lock_serializes_across_processes(tmp_path: Path) -> None:
     store = MemoryStore(tmp_path)
     store.write_long_term("init\n")
 
-    ready = Value("i", 0)
-    done = Value("i", 0)
-    p = Process(
+    # spawn, not the Linux default fork: pytest leaves the parent multi-threaded,
+    # and forking from there segfaults the interpreter at exit.
+    ctx = multiprocessing.get_context("spawn")
+    ready = ctx.Value("i", 0)
+    done = ctx.Value("i", 0)
+    p = ctx.Process(
         target=_worker_acquire_and_hold,
         args=(str(tmp_path), 0.3, ready, done),
     )
diff --git a/tests/test_session_manager.py b/tests/test_session_manager.py
index c82e69760..6db033e81 100644
--- a/tests/test_session_manager.py
+++ b/tests/test_session_manager.py
@@ -245,7 +245,10 @@ def test_concurrent_writers_lose_no_turns(tmp_path: Path):
     """Two processes saving the same session: both turn blocks land,
     each block's messages contiguous (tool_call/result adjacency)."""
     key = "tui:race01"
-    procs = [multiprocessing.Process(target=_turn_worker, args=(str(tmp_path), key, w)) for w in range(2)]
+    # spawn, not the Linux default fork: pytest leaves the parent multi-threaded,
+    # and forking from there segfaults the interpreter at exit.
+    ctx = multiprocessing.get_context("spawn")
+    procs = [ctx.Process(target=_turn_worker, args=(str(tmp_path), key, w)) for w in range(2)]
     for p in procs:
         p.start()
     for p in procs:

From 23d44c7b1ed7cf5fb4140478a83f5d42e8d53410 Mon Sep 17 00:00:00 2001
From: KT <huangjie@evermind.ai>
Date: Tue, 28 Jul 2026 13:12:48 +0800
Subject: [PATCH 6/8] ci: install only the extras the suite needs and drop the
 windows cell

The ubuntu cell passed all 4611 tests but still exited 139. The hazard is
already documented in cli/agent_commands.py: native runtimes loaded by the
agent loop (lancedb's Rust/tokio thread, torch) segfault during interpreter
finalization, which production dodges with a hard exit. Nothing in the repo
imports torch or transformers -- a whole-tree grep finds them only in the
extra declaration, three comments, and a doc note -- so the retrieval extra
is dead weight that only slows CI and adds a crash surface. The unit job now
installs channels + tools + sandbox; the suite still collects 4611 tests.

The windows cell is removed again: boxlite ships wheels for macOS and Linux
only (it is a microVM runtime built on KVM / Hypervisor.framework), so
installing the sandbox extra on Windows fails outright and no test ever runs.
The missing `sys_platform != 'win32'` marker on that dependency is a real
packaging bug, tracked separately -- Windows can rejoin the matrix once it
lands. macOS stays, having passed on its first run.

Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
---
 .github/workflows/ci.yml | 9 ++++++---
 Makefile                 | 2 +-
 2 files changed, 7 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index fccb1d8dd..b4f071f43 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -73,7 +73,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
+        os: [ubuntu-latest, macos-latest]
         python: ["3.12"]
     steps:
       - uses: actions/checkout@v7
@@ -87,8 +87,11 @@ jobs:
       - name: Set up Python
         run: uv python install ${{ matrix.python }}
 
+      # Everything the suite imports, minus `retrieval`: torch/transformers have
+      # no consumer in raven/, and their native runtimes segfault at interpreter
+      # finalization (see the same hazard noted in cli/agent_commands.py).
       - name: Install dependencies
-        run: uv sync --locked --all-extras --dev
+        run: uv sync --locked --extra channels --extra tools --extra sandbox --dev
 
       - name: Unit tests (integration + e2e deselected by norecursedirs + addopts)
         # TERM=dumb: typer's rich_utils force-enables colour when GITHUB_ACTIONS is
@@ -96,7 +99,7 @@ jobs:
         # `"--flag" in result.stdout` assertions fail only on CI.
         env:
           TERM: dumb
-        run: uv run --all-extras pytest -q
+        run: uv run --extra channels --extra tools --extra sandbox pytest -q
 
   windows-upgrade:
     name: Windows self-upgrade
diff --git a/Makefile b/Makefile
index aa5019894..8c689eb0f 100644
--- a/Makefile
+++ b/Makefile
@@ -46,7 +46,7 @@ lint-bridge:
 test: test-python test-tui
 
 test-python:
-	uv run --all-extras pytest -q
+	uv run --extra channels --extra tools --extra sandbox pytest -q
 
 test-tui:
 	npm test --prefix ui-tui

From 2e0cf64ddf8d1968314d749f64fa810a36b7cb70 Mon Sep 17 00:00:00 2001
From: KT <huangjie@evermind.ai>
Date: Tue, 28 Jul 2026 14:27:46 +0800
Subject: [PATCH 7/8] test: hard-exit the pytest session past the lancedb
 finalization hazard

Dropping the retrieval extra was not enough: the ubuntu cell still exits 139
after a fully green run, because lancedb -- a transitive dependency, not
something the extras control -- leaves its Rust/tokio LanceDBBackgroundEventLoop
daemon thread running, and finalizing the interpreter around it segfaults.

raven.cli._exit already documents and guards exactly this for the CLI. Its
docstring notes that CliRunner keeps normal exit semantics in tests, which is
true per command, but the pytest process itself finalizes with that thread
live once any test has built the agent loop. The session now reuses the same
guard: pytest_sessionfinish records the real exit status and
pytest_unconfigure flushes and hard-exits with it when the hazard is live.

Verified that the guard preserves exit codes rather than masking failures: a
passing run exits 0, a failing run exits 1, and a failing run that also loaded
lancedb still exits 1.

Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
---
 tests/conftest.py | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/tests/conftest.py b/tests/conftest.py
index 9e82742a6..f4326328c 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -9,6 +9,27 @@
 import pytest
 
 
+def pytest_unconfigure(config: pytest.Config) -> None:
+    """Hard-exit past interpreter finalization when lancedb's runtime is live.
+
+    Tests that build the agent loop leave lancedb's Rust/tokio daemon thread
+    running with no shutdown hook, and finalizing the interpreter around it
+    segfaults (exit 139) after a fully green run. The CLI guards its own exit
+    the same way; see raven.cli._exit for the analysis.
+    """
+    from raven.cli._exit import flush_and_hard_exit, lancedb_finalization_hazard
+
+    if not lancedb_finalization_hazard():
+        return
+    session = getattr(config, "_raven_exitstatus", 0)
+    flush_and_hard_exit(int(session))
+
+
+def pytest_sessionfinish(session: pytest.Session, exitstatus: int) -> None:
+    """Stash the real exit status so pytest_unconfigure can preserve it."""
+    session.config._raven_exitstatus = int(exitstatus)  # type: ignore[attr-defined]
+
+
 @pytest.fixture(autouse=True)
 def _restore_loguru_enabled_state():
     """Undo any ``loguru.logger.disable("raven")`` left over from a

From 532a1a5840b3c84a2ea869f9b08ad3688f639c28 Mon Sep 17 00:00:00 2001
From: KT <huangjie@evermind.ai>
Date: Tue, 28 Jul 2026 14:43:17 +0800
Subject: [PATCH 8/8] test: gate the pytest hard-exit on CI instead of the
 lancedb probe

The previous guard keyed on lancedb's background thread, but the unit tier
never opens a lancedb connection, so the probe returned false and the run
still exited 139. The crash therefore comes from another native teardown --
the run also reports an asyncio BaseSubprocessTransport finalized during GC,
and only Linux is affected.

Rather than chase each runtime, the guard now hard-exits whenever CI is set,
after the session has recorded its real status. Local runs are untouched, so
nothing masks an exit-time error during development, and the recorded status
is preserved in both cases: verified that a passing run exits 0 and a failing
run exits 1, with and without CI set.

Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
---
 tests/conftest.py | 25 ++++++++++++++++---------
 1 file changed, 16 insertions(+), 9 deletions(-)

diff --git a/tests/conftest.py b/tests/conftest.py
index f4326328c..994463e8a 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -10,19 +10,26 @@
 
 
 def pytest_unconfigure(config: pytest.Config) -> None:
-    """Hard-exit past interpreter finalization when lancedb's runtime is live.
+    """On CI, hard-exit past interpreter finalization once the run is over.
 
-    Tests that build the agent loop leave lancedb's Rust/tokio daemon thread
-    running with no shutdown hook, and finalizing the interpreter around it
-    segfaults (exit 139) after a fully green run. The CLI guards its own exit
-    the same way; see raven.cli._exit for the analysis.
+    Native runtimes pulled in by the suite (lancedb's Rust/tokio thread, asyncio
+    subprocess transports finalized during GC) segfault Py_FinalizeEx on Linux,
+    turning a fully green run into exit 139. raven.cli._exit guards the CLI the
+    same way; the pytest process needs its own guard because it finalizes with
+    those runtimes live.
+
+    Local runs keep normal semantics so nothing masks an exit-time error, and
+    the recorded status is preserved either way -- a failing run still exits
+    non-zero.
     """
-    from raven.cli._exit import flush_and_hard_exit, lancedb_finalization_hazard
+    import os
 
-    if not lancedb_finalization_hazard():
+    if not os.environ.get("CI"):
         return
-    session = getattr(config, "_raven_exitstatus", 0)
-    flush_and_hard_exit(int(session))
+
+    from raven.cli._exit import flush_and_hard_exit
+
+    flush_and_hard_exit(int(getattr(config, "_raven_exitstatus", 0)))
 
 
 def pytest_sessionfinish(session: pytest.Session, exitstatus: int) -> None: