ci: fix env-dependent test failures (gated datasets, server extra, router prefix) (#348)

.github/workflows/ci.yml

@@ -25,7 +25,7 @@ jobs:
         uses: astral-sh/setup-uv@v8.0.0
 
       - name: Install dependencies
-        run: uv sync --extra dev --extra framework-comparison
+        run: uv sync --extra dev --extra framework-comparison --extra server
 
       - name: Ruff check
         run: uv run ruff check src/ tests/
@@ -57,7 +57,7 @@ jobs:
         uses: astral-sh/setup-uv@v8.0.0
 
       - name: Install dependencies
-        run: uv sync --extra dev --extra framework-comparison
+        run: uv sync --extra dev --extra framework-comparison --extra server
 
       - name: Build Rust extension
         run: uv run maturin develop --manifest-path rust/crates/openjarvis-python/Cargo.toml

src/openjarvis/server/agent_manager_routes.py

@@ -1303,10 +1303,11 @@ async def _stream_managed_agent(
 
 def create_agent_manager_router(
     manager: AgentManager,
-) -> Tuple[APIRouter, APIRouter, APIRouter, APIRouter]:
+) -> Tuple[APIRouter, APIRouter, APIRouter, APIRouter, APIRouter]:
     """Create FastAPI routers with agent management endpoints.
 
-    Returns a 4-tuple: (agents_router, templates_router, global_router, tools_router).
+    Returns a 5-tuple:
+    ``(agents_router, templates_router, global_router, tools_router, sendblue_router)``.
     """
     agents_router = APIRouter(prefix="/v1/managed-agents", tags=["managed-agents"])
     templates_router = APIRouter(prefix="/v1/templates", tags=["templates"])

tests/evals/test_dataset_splits_integration.py

@@ -23,10 +23,28 @@ def test_train_and_test_are_disjoint_per_provider(mod_name, cls_name):
     mod = importlib.import_module(mod_name)
     ds_cls = getattr(mod, cls_name)
 
-    train_ds = ds_cls()
-    train_ds.load(split="train", seed=42)
-    test_ds = ds_cls()
-    test_ds.load(split="test", seed=42)
+    # Some providers wrap gated HuggingFace datasets (GAIA,
+    # LiveResearchBench) or optional upstream packages (TauBench needs
+    # the ``tau2`` package). When the local environment lacks the
+    # required HF auth token or the optional install, the dataset
+    # raises rather than returns empty — skip cleanly so a CI runner
+    # without secrets doesn't fail the suite. The split-disjointness
+    # invariant under test is still exercised by every provider whose
+    # data IS available.
+    try:
+        train_ds = ds_cls()
+        train_ds.load(split="train", seed=42)
+        test_ds = ds_cls()
+        test_ds.load(split="test", seed=42)
+    except (ImportError, ModuleNotFoundError) as exc:
+        pytest.skip(f"{cls_name}: optional upstream not installed ({exc})")
+    except Exception as exc:  # noqa: BLE001 — HF errors live in many modules
+        msg = str(exc)
+        if "GatedRepo" in type(exc).__name__ or "gated" in msg.lower():
+            pytest.skip(f"{cls_name}: gated dataset, no HF auth ({type(exc).__name__})")
+        if "DatasetNotFound" in type(exc).__name__:
+            pytest.skip(f"{cls_name}: dataset not accessible ({type(exc).__name__})")
+        raise
 
     train_ids = {r.record_id for r in train_ds.iter_records()}
     test_ids = {r.record_id for r in test_ds.iter_records()}

tests/server/test_agent_manager_routes.py

@@ -38,11 +38,8 @@ class TestAgentManagerRoutes:
 
         app = FastAPI()
         routers = create_agent_manager_router(manager)
-        agents_router, templates_router, global_router, tools_router = routers
-        app.include_router(agents_router)
-        app.include_router(templates_router)
-        app.include_router(global_router)
-        app.include_router(tools_router)
+        for r in routers:
+            app.include_router(r)
         return TestClient(app)
 
     def test_list_agents_empty(self, client):
@@ -287,11 +284,8 @@ class TestAgentManagerStreaming:
         app.state.bus = None
 
         routers = create_agent_manager_router(manager)
-        agents_router, templates_router, global_router, tools_router = routers
-        app.include_router(agents_router)
-        app.include_router(templates_router)
-        app.include_router(global_router)
-        app.include_router(tools_router)
+        for r in routers:
+            app.include_router(r)
         return TestClient(app)
 
     def test_send_message_stream(self, manager, stream_client):

tests/server/test_connectors_router.py

@@ -19,7 +19,11 @@ def app():
 
     _app = FastAPI()
     router = create_connectors_router()
-    _app.include_router(router, prefix="/v1")
+    # The router already carries ``prefix="/v1/connectors"``. The real
+    # ``server/app.py`` mounts it with ``include_router(router)`` — adding
+    # a second ``prefix="/v1"`` here would produce ``/v1/v1/connectors``
+    # and every request would 404.
+    _app.include_router(router)
     return TestClient(_app)
 
 
@@ -83,7 +87,12 @@ def test_sync_status(app):
 
 
 def test_trigger_sync(app, tmp_path: Path) -> None:
-    """POST /v1/connectors/obsidian/sync triggers an incremental sync."""
+    """POST /v1/connectors/obsidian/sync triggers an incremental sync.
+
+    The endpoint is intentionally fire-and-forget — it starts the sync in
+    a background thread and returns immediately with ``status=started``.
+    Sync progress is observable via the separate ``GET .../sync`` endpoint.
+    """
     vault = tmp_path / "vault"
     vault.mkdir()
     (vault / "note.md").write_text("# Test note\n\nContent here.")
@@ -91,4 +100,5 @@ def test_trigger_sync(app, tmp_path: Path) -> None:
     resp = app.post("/v1/connectors/obsidian/sync")
     assert resp.status_code == 200
     data = resp.json()
-    assert data["chunks_indexed"] >= 1
+    assert data["connector_id"] == "obsidian"
+    assert data["status"] in {"started", "already_syncing"}