diff --git a/README.md b/README.md
index a321257..c7aaa6e 100644
--- a/README.md
+++ b/README.md
@@ -130,11 +130,13 @@ In `pl.read_excel()` “calamine” engine can be used for reading all major typ
 <p align="right">(<a href="#readme-top">back to top</a>)</p>
 
 ### Worker aware async scheduler
-The project uses the APScheduler library to schedule tasks in the background.
-The APScheduler library is a powerful and flexible in-process task scheduler with Cron-like capabilities.
-It allows you to schedule jobs to run at specific times or intervals, and it supports multiple job stores, triggers, and executors.
-The library is designed to be easy to use and highly configurable, making it suitable for a wide range of use cases.
-It was added to project in version 4.0.0a5 with Redis as event broker and SQLAlchemy as data store.
+The app starts an `AsyncScheduler` as part of FastAPI's lifespan. Each worker shares
+the persistent SQLAlchemy data store in the `scheduler` schema and uses Redis as the
+event broker, allowing APScheduler to coordinate scheduling across workers. The
+`tick-sql-25` schedule runs `app.services.scheduler.tick` every 25 seconds and is
+replaced on startup so its definition stays current. The project uses APScheduler
+4.0.0a6, a pre-release; see the
+[ASGI/FastAPI integration example](https://github.com/agronholm/apscheduler/blob/4.0.0a6/examples/web/asgi_fastapi.py).
 
 <p align="right">(<a href="#readme-top">back to top</a>)</p>
 
@@ -183,7 +185,7 @@ ollama run llama3.2
 <p align="right">(<a href="#readme-top">back to top</a>)</p>
 
 ### Websocket AI chat service
-A general-purpose, websocket-based chat service lives at `/v1/chat/ws`, built with FastAPI 0.139.0
+A general-purpose, websocket-based chat service lives at `/v1/chat/ws`, built with FastAPI 0.159.1
 and Pydantic models for every message shape exchanged over the wire. It follows the
 [Pydantic AI chat-app](https://github.com/pydantic/pydantic-ai/blob/main/docs/web.md) interaction
 pattern - a session id, role-labeled messages, and streamed assistant chunks - so a UI can be wired
@@ -201,10 +203,15 @@ up later without changing the backend contract.
   chat flow.
 
 **Lifespan wiring**: `app/main.py`'s `lifespan` builds the configured `ChatAgent` via
-`build_chat_agent(settings.chat)` and a fresh `ChatSessionManager` on startup (`app.chat_agent`
-/ `app.state.chat_sessions`), and calls `chat_agent.aclose()` on shutdown to release any held
+`build_chat_agent(settings.chat)` and a fresh `ChatSessionManager` on startup. The agent is stored
+on `app.chat_agent`, and the session manager on `app.state.chat_sessions`; the websocket route
+uses that manager to track active sessions. On shutdown, `chat_agent.aclose()` releases any held
 connections.
 
+The WebSocket integration tests use HTTPX2's async `ASGIWebSocketTransport`
+(`httpx2[ws]`) to exercise the application without starting a server. They manage
+the app lifespan explicitly because the transport does not start it.
+
 **Configuration** (`ChatConfig` in `app/config.py`, all overridable via environment variables):
 
 | Env var | Default | Description |
diff --git a/app/api/chat.py b/app/api/chat.py
index a33247e..7d80d2c 100644
--- a/app/api/chat.py
+++ b/app/api/chat.py
@@ -100,7 +100,7 @@ async def chat_websocket(websocket: WebSocket) -> None:
     await websocket.accept()
 
     agent: ChatAgent = websocket.app.chat_agent
-    session_manager: ChatSessionManager = ChatSessionManager()
+    session_manager: ChatSessionManager = websocket.app.state.chat_sessions
     session = await session_manager.create()
 
     await websocket.send_json(Connected(session_id=session.id).model_dump(mode="json"))
diff --git a/app/main.py b/app/main.py
index e3ab516..1817e11 100644
--- a/app/main.py
+++ b/app/main.py
@@ -22,6 +22,8 @@ from app.middleware.profiler import ProfilingMiddleware
 from app.redis import get_redis
 from app.services.auth import AuthBearer
 from app.services.chat_agent import build_chat_agent
+from app.services.chat_session import ChatSessionManager
+from app.services.scheduler import scheduler_lifespan
 
 templates = Jinja2Templates(directory=Path(__file__).parent.parent / "templates")
 
@@ -31,11 +33,9 @@ async def lifespan(app: FastAPI):
     app.logger = get_logger()
     app.redis = await get_redis()
     global_settings.postgres_url.unicode_string()
-    # Chat service: initialize the pluggable model-client adapter. The session
-    # manager is now a singleton and will be auto-instantiated on first access.
-    # See app/services/chat_agent.py to swap the local stub for a real model
-    # client (OpenAI, Ollama, etc.).
+    # Initialize the chat agent and per-application session registry.
     app.chat_agent = build_chat_agent(global_settings.chat)
+    app.state.chat_sessions = ChatSessionManager()
     try:
         # app.postgres_pool = await asyncpg.create_pool(
         #     dsn=postgres_dsn,
@@ -45,7 +45,8 @@ async def lifespan(app: FastAPI):
         # await app.logger.ainfo(
         #     "Postgres pool created", idle_size=app.postgres_pool.get_idle_size()
         # )
-        yield
+        async with scheduler_lifespan():
+            yield
     except Exception as e:
         await app.logger.aerror("Error during app startup", error=repr(e))
         raise
@@ -96,16 +97,4 @@ def create_app() -> FastAPI:
 
 app = create_app()
 
-# --- Unused/experimental code and TODOs ---
-# from apscheduler import AsyncScheduler
-# from apscheduler.datastores.sqlalchemy import SQLAlchemyDataStore
-# from apscheduler.eventbrokers.redis import RedisEventBroker
-# from app.database import engine
-# from app.services.scheduler import SchedulerMiddleware
-# _scheduler_data_store = SQLAlchemyDataStore(engine, schema="scheduler")
-# _scheduler_event_broker = RedisEventBroker(client_or_url=global_settings.redis_url.unicode_string())
-# _scheduler_himself = AsyncScheduler(_scheduler_data_store, _scheduler_event_broker)
-# app.add_middleware(SchedulerMiddleware, scheduler=_scheduler_himself)
 # TODO: every non-GET method should reset cache
-# TODO: scheduler tasks needing DB should access connection pool via request
-# TODO: https://stackoverflow.com/questions/16053364/make-sure-only-one-worker-launches-the-apscheduler-event-in-a-pyramid-web-app-ru
diff --git a/app/services/__init__.py b/app/services/__init__.py
index e69de29..a0dac8d 100644
--- a/app/services/__init__.py
+++ b/app/services/__init__.py
@@ -0,0 +1 @@
+"""Application service implementations."""
diff --git a/app/services/auth.py b/app/services/auth.py
index 0a4f973..6499879 100644
--- a/app/services/auth.py
+++ b/app/services/auth.py
@@ -1,3 +1,5 @@
+"""Bearer-token authentication and Redis-backed token storage."""
+
 import time
 
 import jwt
@@ -11,24 +13,27 @@ from app.models.user import User
 logger = get_logger()
 
 
-async def get_from_redis(request: Request, key: str):
+async def get_from_redis(request: Request, key: str) -> str | None:
+    """Return the Redis value stored under ``key``."""
     return await request.app.redis.get(key)
 
 
-async def set_to_redis(request: Request, key: str, value: str, ex: int):
+async def set_to_redis(request: Request, key: str, value: str, ex: int) -> bool:
+    """Store ``value`` in Redis for ``ex`` seconds."""
     return await request.app.redis.set(key, value, ex=ex)
 
 
 async def verify_jwt(request: Request, token: str) -> bool:
+    """Check whether Redis recognizes the bearer token."""
     payload = await get_from_redis(request, token)
     return bool(payload)
 
 
 class AuthBearer(HTTPBearer):
-    def __init__(self, auto_error: bool = True) -> None:
-        super().__init__(auto_error=auto_error)
+    """Validate bearer credentials against the active Redis token store."""
 
-    async def __call__(self, request: Request):
+    async def __call__(self, request: Request) -> str:
+        """Return a valid token or raise an authorization error."""
         credentials: HTTPAuthorizationCredentials = await super().__call__(request)
         if not credentials:
             raise HTTPException(status_code=403, detail="Invalid authorization code.")
@@ -44,8 +49,8 @@ class AuthBearer(HTTPBearer):
         return credentials.credentials
 
 
-async def create_access_token(user: User, request: Request):
-    # sourcery skip: avoid-builtin-shadow
+async def create_access_token(user: User, request: Request) -> str | None:
+    """Sign a user token and store it in Redis until its configured expiry."""
     payload = {
         "email": user.email,
         "expiry": time.time() + global_settings.jwt_expire,
@@ -55,8 +60,8 @@ async def create_access_token(user: User, request: Request):
         payload, str(user.password), algorithm=global_settings.jwt_algorithm
     )
 
-    _bool = await set_to_redis(
+    stored = await set_to_redis(
         request, token, str(payload), ex=global_settings.jwt_expire
     )
-    if _bool:
+    if stored:
         return token
diff --git a/app/services/chat_agent.py b/app/services/chat_agent.py
index 7ef5725..6a237b6 100644
--- a/app/services/chat_agent.py
+++ b/app/services/chat_agent.py
@@ -16,7 +16,7 @@ from typing import Protocol, runtime_checkable
 from urllib.parse import urlparse
 
 import attrs
-import httpx
+import httpx2
 import orjson
 
 from app.config import ChatConfig
@@ -49,9 +49,11 @@ class LocalEchoAgent:
     """
 
     def __init__(self, stream_delay_seconds: float = 0.02) -> None:
+        """Initialize the delay between streamed words."""
         self.stream_delay_seconds = stream_delay_seconds
 
     def _compose_reply(self, messages: list[ChatMessage]) -> str:
+        """Compose a canned response to the latest user message."""
         last_user = next(
             (m.content for m in reversed(messages) if m.role == "user"), ""
         )
@@ -63,26 +65,20 @@ class LocalEchoAgent:
         return f"You said: {last_user!r}. This is a local echo response (stub agent)."
 
     async def stream_reply(self, messages: list[ChatMessage]) -> AsyncIterator[str]:
+        """Yield the canned reply one word at a time."""
         reply = self._compose_reply(messages)
         for word in reply.split(" "):
             await asyncio.sleep(self.stream_delay_seconds + random.uniform(0, 0.01))
             yield word + " "
 
     async def aclose(self) -> None:
-        return None
+        """Close resources; the local agent holds none."""
 
 
-def _validate_base_url(_instance: OllamaChatAgent, _attribute: attrs.Attribute, value: str) -> None:
-    """Validate that base_url is a valid HTTP(S) URL.
-
-    Args:
-        instance: The OllamaChatAgent instance being initialized.
-        attribute: The attrs attribute descriptor for base_url.
-        value: The URL string to validate.
-
-    Raises:
-        ValueError: If base_url is not a valid HTTP(S) URL.
-    """
+def _validate_base_url(
+    _instance: OllamaChatAgent, _attribute: attrs.Attribute, value: str
+) -> None:
+    """Require an absolute HTTP(S) endpoint URL."""
     parsed = urlparse(value)
     if parsed.scheme not in ("http", "https"):
         msg = f"base_url must be HTTP(S), got scheme '{parsed.scheme}' from '{value}'"
@@ -92,37 +88,15 @@ def _validate_base_url(_instance: OllamaChatAgent, _attribute: attrs.Attribute,
         raise ValueError(msg)
 
 
-def _validate_timeout(_instance: OllamaChatAgent, _attribute: attrs.Attribute, value: float) -> None:
-    """Validate that timeout is positive.
-
-    Args:
-        instance: The OllamaChatAgent instance being initialized.
-        attribute: The attrs attribute descriptor for timeout.
-        value: The timeout value in seconds.
-
-    Raises:
-        ValueError: If timeout is not positive.
-    """
+def _validate_timeout(
+    _instance: OllamaChatAgent, _attribute: attrs.Attribute, value: float
+) -> None:
+    """Require a positive request timeout."""
     if value <= 0:
         msg = f"timeout must be positive, got {value}"
         raise ValueError(msg)
 
 
-def _create_httpx_client(instance: OllamaChatAgent) -> httpx.AsyncClient:
-    """Factory function to create the httpx.AsyncClient with validated config.
-
-    This function is called from __attrs_post_init__ after all field validators
-    have run, ensuring the client is created with validated configuration.
-
-    Args:
-        instance: The OllamaChatAgent instance being initialized.
-
-    Returns:
-        An initialized httpx.AsyncClient configured with base_url and timeout.
-    """
-    return httpx.AsyncClient(base_url=instance.base_url, timeout=instance.timeout)
-
-
 @attrs.define(slots=True, eq=False, hash=False)
 class OllamaChatAgent:
     """Streams chat completions from an OpenAI-compatible endpoint.
@@ -132,9 +106,6 @@ class OllamaChatAgent:
     a ready-to-swap-in replacement for :class:`LocalEchoAgent` once a real
     model should be used.
 
-    Attrs Configuration:
-        - slots=True: Memory-efficient attribute storage (~40-50% reduction)
-        - eq=False, hash=False: Instances are not comparable (contain async resources)
     """
 
     model: str = attrs.field(
@@ -160,20 +131,15 @@ class OllamaChatAgent:
             "constraints": "Must be positive",
         },
     )
-    _client: httpx.AsyncClient = attrs.field(
+    _client: httpx2.AsyncClient = attrs.field(
         init=False,
         repr=False,
         metadata={"description": "Internal HTTP client for API communication"},
     )
 
     def __attrs_post_init__(self) -> None:
-        """Initialize the HTTP client after field validation.
-
-        This hook is called by attrs after __init__ completes and all field
-        validators have run. It's used to initialize the internal _client
-        field which depends on validated configuration.
-        """
-        self._client = _create_httpx_client(self)
+        """Create the HTTPX2 client after attrs validates the fields."""
+        self._client = httpx2.AsyncClient(base_url=self.base_url, timeout=self.timeout)
 
     async def stream_reply(self, messages: list[ChatMessage]) -> AsyncIterator[str]:
         """Stream chat completion responses from the configured model.
@@ -195,13 +161,11 @@ class OllamaChatAgent:
             async for line in response.aiter_lines():
                 if not line.startswith("data: ") or line == "data: [DONE]":
                     continue
-                try:
-                    data = orjson.loads(line[6:])
-                    content = (
-                        data.get("choices", [{}])[0].get("delta", {}).get("content", "")
-                    )
-                except Exception:
-                    content = ""
+                data = orjson.loads(line.removeprefix("data: "))
+                choices = data.get("choices", [])
+                if not choices:
+                    continue
+                content = choices[0].get("delta", {}).get("content", "")
                 if content:
                     yield content
 
diff --git a/app/services/chat_session.py b/app/services/chat_session.py
index 4de6ab6..8b8b3bd 100644
--- a/app/services/chat_session.py
+++ b/app/services/chat_session.py
@@ -13,46 +13,54 @@ from uuid import UUID, uuid4
 import attrs
 
 from app.schemas.chat import ChatMessage
-from app.utils.singleton import SingletonMetaNoArgs
 
 
 @attrs.define(slots=True)
 class ChatSession:
+    """Conversation history associated with one websocket connection."""
+
     id: UUID = attrs.field(factory=uuid4)
     messages: list[ChatMessage] = attrs.field(factory=list)
 
     def add(self, message: ChatMessage) -> None:
+        """Append a message to this conversation."""
         self.messages.append(message)
 
     def history(self) -> list[ChatMessage]:
+        """Return a copy of the conversation history."""
         return list(self.messages)
 
 
-class ChatSessionManager(metaclass=SingletonMetaNoArgs):
+class ChatSessionManager:
     """Tracks active chat sessions keyed by their opaque session id.
 
-    Implemented as a singleton to ensure exactly one instance per application,
-    maintaining a consistent registry of all active websocket chat sessions.
+    Each application lifespan owns its own manager, maintaining a consistent
+    registry of active websocket chat sessions for that application.
     """
 
     def __init__(self) -> None:
+        """Create an empty, lock-protected session registry."""
         self._sessions: dict[UUID, ChatSession] = {}
         self._lock = asyncio.Lock()
 
     async def create(self) -> ChatSession:
+        """Create and register a new conversation."""
         session = ChatSession()
         async with self._lock:
             self._sessions[session.id] = session
         return session
 
     async def get(self, session_id: UUID) -> ChatSession | None:
+        """Return the matching session, if it is active."""
         async with self._lock:
             return self._sessions.get(session_id)
 
     async def remove(self, session_id: UUID) -> None:
+        """Remove a session from the active registry."""
         async with self._lock:
             self._sessions.pop(session_id, None)
 
     async def count(self) -> int:
+        """Return the number of active sessions."""
         async with self._lock:
             return len(self._sessions)
diff --git a/app/services/llm.py b/app/services/llm.py
index ceaa455..cc1f4a3 100644
--- a/app/services/llm.py
+++ b/app/services/llm.py
@@ -1,25 +1,28 @@
-from collections.abc import AsyncGenerator
+"""Streaming client for OpenAI-compatible chat completion endpoints."""
 
-import httpx
+from collections.abc import AsyncIterator
+
+import httpx2
 import orjson
 
 
 class StreamLLMService:
+    """Send prompts to an LLM endpoint and yield newline-delimited JSON events."""
+
     def __init__(self, base_url: str = "http://localhost:11434/v1") -> None:
+        """Configure the endpoint and default model."""
         self.base_url = base_url
         self.model = "llama3.2"
 
-    async def stream_chat(self, prompt: str) -> AsyncGenerator[bytes]:
-        """Stream chat completion responses from LLM."""
-        # Send the user a message first
+    async def stream_chat(self, prompt: str) -> AsyncIterator[bytes]:
+        """Yield the prompt followed by streamed model response chunks."""
         user_msg = {
             "role": "user",
             "content": prompt,
         }
         yield orjson.dumps(user_msg) + b"\n"
 
-        # Open client as context manager and stream responses
-        async with httpx.AsyncClient(base_url=self.base_url) as client:
+        async with httpx2.AsyncClient(base_url=self.base_url) as client:
             async with client.stream(
                 "POST",
                 "/chat/completions",
@@ -32,21 +35,16 @@ class StreamLLMService:
             ) as response:
                 async for line in response.aiter_lines():
                     if line.startswith("data: ") and line != "data: [DONE]":
-                        try:
-                            json_line = line[6:]  # Remove "data: " prefix
-                            data = orjson.loads(json_line)
-                            content = (
-                                data.get("choices", [{}])[0]
-                                .get("delta", {})
-                                .get("content", "")
-                            )
-                            if content:
-                                model_msg = {"role": "model", "content": content}
-                                yield orjson.dumps(model_msg) + b"\n"
-                        except Exception:
-                            pass
+                        data = orjson.loads(line.removeprefix("data: "))
+                        choices = data.get("choices", [])
+                        if not choices:
+                            continue
+                        content = choices[0].get("delta", {}).get("content", "")
+                        if content:
+                            model_msg = {"role": "model", "content": content}
+                            yield orjson.dumps(model_msg) + b"\n"
 
 
-# FastAPI dependency
 def get_llm_service(base_url: str | None = None) -> StreamLLMService:
+    """Build the service using the supplied URL or its local default."""
     return StreamLLMService(base_url=base_url or "http://localhost:11434/v1")
diff --git a/app/services/scheduler.py b/app/services/scheduler.py
index 6352fb6..be2ff5c 100644
--- a/app/services/scheduler.py
+++ b/app/services/scheduler.py
@@ -1,18 +1,24 @@
+"""Background database health-check scheduling."""
+
+from collections.abc import AsyncIterator
+from contextlib import asynccontextmanager
 from datetime import datetime
 
-from apscheduler import AsyncScheduler
+from apscheduler import AsyncScheduler, ConflictPolicy
+from apscheduler.datastores.sqlalchemy import SQLAlchemyDataStore
+from apscheduler.eventbrokers.redis import RedisEventBroker
 from apscheduler.triggers.interval import IntervalTrigger
-from attrs import define
 from rotoger import get_logger
 from sqlalchemy import text
-from starlette.types import ASGIApp, Receive, Scope, Send
 
-from app.database import AsyncSessionFactory
+from app.config import settings as global_settings
+from app.database import AsyncSessionFactory, engine
 
 logger = get_logger()
 
 
-async def tick():
+async def tick() -> bool:
+    """Run the scheduler's lightweight database heartbeat."""
     async with AsyncSessionFactory() as session:
         stmt = text("select 1;")
         await logger.ainfo(f">>>> Be or not to be...{datetime.now()}")
@@ -21,26 +27,20 @@ async def tick():
         return True
 
 
-@define
-class SchedulerMiddleware:
-    app: ASGIApp
-    scheduler: AsyncScheduler
+@asynccontextmanager
+async def scheduler_lifespan() -> AsyncIterator[None]:
+    """Start the scheduler and stop it when the app lifespan ends."""
+    scheduler = AsyncScheduler(
+        SQLAlchemyDataStore(engine, schema="scheduler"),
+        RedisEventBroker(client_or_url=global_settings.redis_url.unicode_string()),
+    )
 
-    async def __call__(self, scope: Scope, receive: Receive, send: Send) -> None:
-        """
-        Handles the incoming request and schedules tasks if the scope type is 'lifespan'.
-
-        Args:
-            scope (Scope): The ASGI scope dictionary containing request information.
-            receive (Receive): The ASGI receive callable.
-            send (Send): The ASGI send callable.
-        """
-        if scope["type"] == "lifespan":
-            async with self.scheduler:
-                await self.scheduler.add_schedule(
-                    tick, IntervalTrigger(seconds=25), id="tick-sql-25"
-                )
-                await self.scheduler.start_in_background()
-                await self.app(scope, receive, send)
-        else:
-            await self.app(scope, receive, send)
+    async with scheduler:
+        await scheduler.add_schedule(
+            tick,
+            IntervalTrigger(seconds=25),
+            id="tick-sql-25",
+            conflict_policy=ConflictPolicy.replace,
+        )
+        await scheduler.start_in_background()
+        yield
diff --git a/app/services/smtp.py b/app/services/smtp.py
index 3d09932..16734f7 100644
--- a/app/services/smtp.py
+++ b/app/services/smtp.py
@@ -1,3 +1,5 @@
+"""Email delivery through a configured SMTP server."""
+
 import smtplib
 from email.mime.multipart import MIMEMultipart
 from email.mime.text import MIMEText
diff --git a/pyproject.toml b/pyproject.toml
index 9b88b57..966be12 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -12,7 +12,7 @@ dependencies = [
     "uvicorn[standard]==0.54.0",
     "asyncpg==0.32.0",
     "alembic==1.20.1",
-    "httpx==0.28.1",
+    "httpx2==2.16.0",
     "uvloop==0.23.0",
     "httptools==0.9.0",
     "rich==15.0.0",
@@ -29,6 +29,7 @@ dependencies = [
     "websockets==17.3",
     "jinja2==3.1.6",
     "whenever==0.10.5",
+    "sniffio==1.3.1",
 ]
 
 [dependency-groups]
@@ -46,6 +47,8 @@ test = [
     "inline-snapshot==0.36.1",
     "dirty-equals==0.11",
     "polyfactory==3.5.0",
+    "httpx==0.28.1",
+    "httpx2[ws]==2.16.0",
 ]
 
 [tool.pyrefly.errors]
diff --git a/tests/api/test_chat_ws.py b/tests/api/test_chat_ws.py
index 8867df6..629d30d 100644
--- a/tests/api/test_chat_ws.py
+++ b/tests/api/test_chat_ws.py
@@ -1,81 +1,106 @@
 """Tests for the websocket chat service using the local stub agent.
 
-Uses Starlette's synchronous ``TestClient`` websocket support (rather than
-the ``httpx.AsyncClient`` fixture used elsewhere) since it triggers the
-FastAPI lifespan on ``__enter__``/``__exit__``, wiring up
-``app.chat_agent`` / ``app.state.chat_sessions`` exactly like a real
-server run.
+Uses HTTPX2's async ASGI WebSocket transport. The app lifespan is managed
+explicitly because the transport does not trigger lifespan events itself.
 """
 
+from collections.abc import AsyncIterator
+from contextlib import asynccontextmanager
+
 import pytest
-from fastapi.testclient import TestClient
+from httpx2 import AsyncClient
+from httpx2.websockets import ASGIWebSocketTransport
 
 from app.main import app
 
 pytestmark = pytest.mark.anyio
 
 
-async def test_chat_ws_full_conversation_flow() -> None:
-    with (
-        TestClient(app) as client,
-        client.websocket_connect("/v1/chat/ws") as websocket,
-    ):
+@pytest.fixture()
+async def chat_client(
+    monkeypatch: pytest.MonkeyPatch,
+) -> AsyncIterator[AsyncClient]:
+    class RedisStub:
+        async def aclose(self) -> None:
+            pass
+
+    async def get_redis_stub() -> RedisStub:
+        return RedisStub()
+
+    @asynccontextmanager
+    async def scheduler_lifespan_stub() -> AsyncIterator[None]:
+        yield
+
+    monkeypatch.setattr("app.main.get_redis", get_redis_stub)
+    monkeypatch.setattr("app.main.scheduler_lifespan", scheduler_lifespan_stub)
+
+    async with app.router.lifespan_context(app):
+        async with AsyncClient(
+            transport=ASGIWebSocketTransport(app),
+            base_url="http://testserver",
+        ) as client:
+            yield client
+
+
+async def test_chat_ws_full_conversation_flow(chat_client: AsyncClient) -> None:
+    assert await app.state.chat_sessions.count() == 0
+
+    async with chat_client.websocket("ws://testserver/v1/chat/ws") as websocket:
         # 1. connect
-        connected = websocket.receive_json()
+        connected = await websocket.receive_json()
         assert connected["type"] == "connected"
         session_id = connected["session_id"]
+        assert await app.state.chat_sessions.count() == 1
 
         # 2. start conversation with a system prompt
-        websocket.send_json(
+        await websocket.send_json(
             {"type": "start", "system_prompt": "You are a helpful assistant."}
         )
-        started = websocket.receive_json()
-        print(f"{started=}")
+        started = await websocket.receive_json()
         assert started == {"type": "started", "session_id": session_id}
 
         # 3. send a user message and aggregate the streamed assistant chunks
-        websocket.send_json({"type": "user_message", "content": "hello"})
+        await websocket.send_json({"type": "user_message", "content": "hello"})
 
-        event = websocket.receive_json()
+        event = await websocket.receive_json()
         aggregated = ""
         while event["type"] == "assistant_chunk":
             assert event["index"] >= 0
             aggregated += event["content"]
-            event = websocket.receive_json()
+            event = await websocket.receive_json()
 
         assert event["type"] == "assistant_message"
         assert event["message"]["role"] == "assistant"
         assert event["message"]["content"] == aggregated
-        print(f"{event=}")
-        print(f"{aggregated=}")
         assert "hello" in aggregated.lower()
 
         # 4. request full history (system + user + assistant)
-        websocket.send_json({"type": "history_request"})
-        history = websocket.receive_json()
+        await websocket.send_json({"type": "history_request"})
+        history = await websocket.receive_json()
         assert history["type"] == "history"
         roles = [m["role"] for m in history["messages"]]
         assert roles == ["system", "user", "assistant"]
 
         # 5. gracefully end the conversation
-        websocket.send_json({"type": "end"})
+        await websocket.send_json({"type": "end"})
+
+    assert await app.state.chat_sessions.count() == 0
 
 
-async def test_chat_ws_invalid_event_returns_error() -> None:
-    with (
-        TestClient(app) as client,
-        client.websocket_connect("/v1/chat/ws") as websocket,
-    ):
-        websocket.receive_json()  # connected
+async def test_chat_ws_invalid_event_returns_error(chat_client: AsyncClient) -> None:
+    async with chat_client.websocket("ws://testserver/v1/chat/ws") as websocket:
+        await websocket.receive_json()  # connected
 
-        websocket.send_json({"type": "not_a_real_event"})
-        error = websocket.receive_json()
+        await websocket.send_json({"type": "not_a_real_event"})
+        error = await websocket.receive_json()
         assert error["type"] == "error"
 
-        websocket.send_json({"type": "end"})
+        await websocket.send_json({"type": "end"})
 
 
-async def test_chat_ws_ask_question_and_print_model_answer() -> None:
+async def test_chat_ws_ask_question_and_print_model_answer(
+    chat_client: AsyncClient,
+) -> None:
     """Ask a real question over the websocket chat and print the model's answer.
 
     Runs against whatever ``ChatAgent`` is wired up via ``app.state.chat_agent``
@@ -85,30 +110,27 @@ async def test_chat_ws_ask_question_and_print_model_answer() -> None:
     """
     question = "What is the capital of France?"
 
-    with (
-        TestClient(app) as client,
-        client.websocket_connect("/v1/chat/ws") as websocket,
-    ):
+    async with chat_client.websocket("ws://testserver/v1/chat/ws") as websocket:
         # 1. connect
-        connected = websocket.receive_json()
+        connected = await websocket.receive_json()
         assert connected["type"] == "connected"
         session_id = connected["session_id"]
 
         # 2. start conversation
-        websocket.send_json(
+        await websocket.send_json(
             {"type": "start", "system_prompt": "You are a helpful assistant."}
         )
-        started = websocket.receive_json()
+        started = await websocket.receive_json()
         assert started == {"type": "started", "session_id": session_id}
 
         # 3. ask the question and aggregate the streamed answer
-        websocket.send_json({"type": "user_message", "content": question})
+        await websocket.send_json({"type": "user_message", "content": question})
 
-        event = websocket.receive_json()
+        event = await websocket.receive_json()
         answer = ""
         while event["type"] == "assistant_chunk":
             answer += event["content"]
-            event = websocket.receive_json()
+            event = await websocket.receive_json()
 
         assert event["type"] == "assistant_message"
         answer = event["message"]["content"]
@@ -120,5 +142,4 @@ async def test_chat_ws_ask_question_and_print_model_answer() -> None:
         assert answer.strip() != ""
 
         # 4. gracefully end the conversation
-        websocket.send_json({"type": "end"})
-
+        await websocket.send_json({"type": "end"})
diff --git a/tests/test_ollama_chat_agent_attrs.py b/tests/test_ollama_chat_agent_attrs.py
index 19250f2..3866677 100644
--- a/tests/test_ollama_chat_agent_attrs.py
+++ b/tests/test_ollama_chat_agent_attrs.py
@@ -1,5 +1,6 @@
 """Test suite for refactored OllamaChatAgent with attrs features."""
-import httpx
+
+import httpx2
 import pytest
 
 from app.config import ChatConfig
@@ -54,14 +55,14 @@ class TestOllamaChatAgentAttrsFeatures:
         assert isinstance(agent.timeout, float)
         assert agent.timeout == 30.0
 
-    def test_factory_initialization_of_client(self):
-        """Test that _client is created via factory function."""
+    def test_client_initialization_after_field_validation(self):
+        """Test that the async client uses the validated configuration."""
         agent = OllamaChatAgent(model="llama3.2", base_url="http://localhost:11434/v1")
         assert hasattr(agent, "_client")
-        assert isinstance(agent._client, httpx.AsyncClient)
-        # httpx normalizes URLs by adding a trailing slash
+        assert isinstance(agent._client, httpx2.AsyncClient)
+        # httpx2 normalizes URLs by adding a trailing slash
         assert str(agent._client.base_url) == "http://localhost:11434/v1/"
-        assert agent._client.timeout == httpx.Timeout(60.0)
+        assert agent._client.timeout == httpx2.Timeout(60.0)
 
     def test_slots_enabled_no_dict(self):
         """Test that slots=True prevents __dict__ attribute."""
@@ -72,12 +73,8 @@ class TestOllamaChatAgentAttrsFeatures:
 
     def test_equality_disabled(self):
         """Test that eq=False means instances are not equal even with same values."""
-        agent1 = OllamaChatAgent(
-            model="llama3.2", base_url="http://localhost:11434/v1"
-        )
-        agent2 = OllamaChatAgent(
-            model="llama3.2", base_url="http://localhost:11434/v1"
-        )
+        agent1 = OllamaChatAgent(model="llama3.2", base_url="http://localhost:11434/v1")
+        agent2 = OllamaChatAgent(model="llama3.2", base_url="http://localhost:11434/v1")
         # With eq=False, only identity comparison works
         assert agent1 != agent2
         assert agent1 == agent1
@@ -112,9 +109,7 @@ class TestOllamaChatAgentAttrsFeatures:
 
     def test_https_urls_accepted(self):
         """Test that HTTPS URLs are properly accepted."""
-        agent = OllamaChatAgent(
-            model="llama3.2", base_url="https://api.example.com/v1"
-        )
+        agent = OllamaChatAgent(model="llama3.2", base_url="https://api.example.com/v1")
         assert agent.base_url == "https://api.example.com/v1"
 
     def test_metadata_annotations_present(self):
@@ -133,4 +128,3 @@ class TestOllamaChatAgentAttrsFeatures:
 
 if __name__ == "__main__":
     pytest.main([__file__, "-v"])
-
diff --git a/tests/test_scheduler.py b/tests/test_scheduler.py
new file mode 100644
index 0000000..a21db8b
--- /dev/null
+++ b/tests/test_scheduler.py
@@ -0,0 +1,62 @@
+import pytest
+from apscheduler import ConflictPolicy
+from apscheduler.triggers.interval import IntervalTrigger
+
+import app.services.scheduler as scheduler_service
+
+
+@pytest.mark.anyio
+async def test_scheduler_lifespan_starts_and_stops_scheduler(monkeypatch) -> None:
+    events = []
+    data_store = object()
+    event_broker = object()
+
+    class SchedulerStub:
+        async def __aenter__(self):
+            events.append("enter")
+            return self
+
+        async def __aexit__(self, *_args: object) -> None:
+            events.append("exit")
+
+        async def add_schedule(
+            self, func, trigger, *, id: str, conflict_policy: ConflictPolicy
+        ) -> None:
+            events.append(("schedule", func, trigger, id, conflict_policy))
+
+        async def start_in_background(self) -> None:
+            events.append("start")
+
+    scheduler = SchedulerStub()
+
+    def create_data_store(actual_engine, *, schema):
+        assert actual_engine is scheduler_service.engine
+        assert schema == "scheduler"
+        return data_store
+
+    def create_event_broker(*, client_or_url):
+        assert client_or_url == (
+            scheduler_service.global_settings.redis_url.unicode_string()
+        )
+        return event_broker
+
+    def create_scheduler(actual_data_store, actual_event_broker):
+        assert actual_data_store is data_store
+        assert actual_event_broker is event_broker
+        return scheduler
+
+    monkeypatch.setattr(scheduler_service, "SQLAlchemyDataStore", create_data_store)
+    monkeypatch.setattr(scheduler_service, "RedisEventBroker", create_event_broker)
+    monkeypatch.setattr(scheduler_service, "AsyncScheduler", create_scheduler)
+
+    async with scheduler_service.scheduler_lifespan():
+        assert events[0] == "enter"
+        event_type, func, trigger, schedule_id, conflict_policy = events[1]
+        assert event_type == "schedule"
+        assert func is scheduler_service.tick
+        assert isinstance(trigger, IntervalTrigger)
+        assert schedule_id == "tick-sql-25"
+        assert conflict_policy is ConflictPolicy.replace
+        assert events[2:] == ["start"]
+
+    assert events[-1] == "exit"
diff --git a/uv.lock b/uv.lock
index e44ae8f..ce7e279 100644
--- a/uv.lock
+++ b/uv.lock
@@ -330,7 +330,7 @@ dependencies = [
     { name = "fastexcel" },
     { name = "granian" },
     { name = "httptools" },
-    { name = "httpx" },
+    { name = "httpx2" },
     { name = "jinja2" },
     { name = "polars" },
     { name = "pydantic", extra = ["email"] },
@@ -341,6 +341,7 @@ dependencies = [
     { name = "redis" },
     { name = "rich" },
     { name = "rotoger" },
+    { name = "sniffio" },
     { name = "sqlalchemy" },
     { name = "uvicorn", extra = ["standard"] },
     { name = "uvloop" },
@@ -358,6 +359,8 @@ dev = [
 ]
 test = [
     { name = "dirty-equals" },
+    { name = "httpx" },
+    { name = "httpx2", extra = ["ws"] },
     { name = "inline-snapshot" },
     { name = "polyfactory" },
     { name = "pytest" },
@@ -374,7 +377,7 @@ requires-dist = [
     { name = "fastexcel", specifier = "==0.21.0" },
     { name = "granian", specifier = "==2.8.5" },
     { name = "httptools", specifier = "==0.9.0" },
-    { name = "httpx", specifier = "==0.28.1" },
+    { name = "httpx2", specifier = "==2.16.0" },
     { name = "jinja2", specifier = "==3.1.6" },
     { name = "polars", specifier = "==2.0.0" },
     { name = "pydantic", extras = ["email"], specifier = "==2.14.1" },
@@ -385,6 +388,7 @@ requires-dist = [
     { name = "redis", specifier = "==8.1.0" },
     { name = "rich", specifier = "==15.0.0" },
     { name = "rotoger", specifier = "==0.3.0" },
+    { name = "sniffio", specifier = "==1.3.1" },
     { name = "sqlalchemy", specifier = "==2.1.6" },
     { name = "uvicorn", extras = ["standard"], specifier = "==0.54.0" },
     { name = "uvloop", specifier = "==0.23.0" },
@@ -402,6 +406,8 @@ dev = [
 ]
 test = [
     { name = "dirty-equals", specifier = "==0.11" },
+    { name = "httpx", specifier = "==0.28.1" },
+    { name = "httpx2", extras = ["ws"], specifier = "==2.16.0" },
     { name = "inline-snapshot", specifier = "==0.36.1" },
     { name = "polyfactory", specifier = "==3.5.0" },
     { name = "pytest", specifier = "==9.1.1" },
@@ -511,6 +517,19 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/7e/f5/f66802a942d491edb555dd61e3a9961140fd64c90bce1eafd741609d334d/httpcore-1.0.9-py3-none-any.whl", hash = "sha256:2d400746a40668fc9dec9810239072b40b4484b640a8c38fd654a024c7a1bf55", size = 78784, upload-time = "2025-04-24T22:06:20.566Z" },
 ]
 
+[[package]]
+name = "httpcore2"
+version = "2.16.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "h11" },
+    { name = "truststore" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/c0/61/afa9d6b7452f9eca73f73d45324eb0bb52f098fd5e6049bf6b8263be9b4e/httpcore2-2.16.0.tar.gz", hash = "sha256:c72cd0413563b17e1cd8354677dff887466f8ec2551679759e93126534428643", size = 69886, upload-time = "2026-10-21T07:09:45.31Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/af/c2/ce417c397afa5181f0e3c6ff01a4d01639414e1adef08730d97571297151/httpcore2-2.16.0-py3-none-any.whl", hash = "sha256:5c4358319f2602d984ad93182c9fee5160f22e9da7babdda870a5e7c4e2dc34f", size = 85070, upload-time = "2026-10-21T07:09:42.802Z" },
+]
+
 [[package]]
 name = "httptools"
 version = "0.9.0"
@@ -560,6 +579,36 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/2a/39/e50c7c3a983047577ee07d2a9e53faf5a69493943ec3f6a384bdc792deb2/httpx-0.28.1-py3-none-any.whl", hash = "sha256:d909fcccc110f8c7faf814ca82a9a4d816bc5a6dbfea25d6591d6985b8ba59ad", size = 73517, upload-time = "2024-12-06T15:37:21.509Z" },
 ]
 
+[[package]]
+name = "httpx2"
+version = "2.16.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio", marker = "sys_platform != 'emscripten'" },
+    { name = "httpcore2", marker = "sys_platform != 'emscripten'" },
+    { name = "httpx2-jsfetch", marker = "sys_platform == 'emscripten'" },
+    { name = "idna" },
+    { name = "truststore", marker = "sys_platform != 'emscripten'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/43/73/5ca2cedf5e4e50423d1c258877865a6b62793633b561c58499f8aeb33b7e/httpx2-2.16.0.tar.gz", hash = "sha256:0d54c98128e8f2ff7fbbde0fca8c359e5f032edeaaaac443cdb7f31f685bcd7f", size = 103116, upload-time = "2026-10-21T07:09:46.348Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7b/3d/2e4bc4fd1ef9770871669adb348d612a6e4786b303a23f9935b81c48dd60/httpx2-2.16.0-py3-none-any.whl", hash = "sha256:fd2c5d7909e8d86fe92d53020c310ca0e4e431b838c207c5754b9bbf0308d49a", size = 97880, upload-time = "2026-10-21T07:09:44.01Z" },
+]
+
+[package.optional-dependencies]
+ws = [
+    { name = "wsproto" },
+]
+
+[[package]]
+name = "httpx2-jsfetch"
+version = "1.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/cd/c4/0e5636363151a2a1795e0a77617168b9ca438e1748ec05fc9b5687f93d64/httpx2_jsfetch-1.0.tar.gz", hash = "sha256:70a0e3eabfef7cce5ad9c629f7d01ca05e418f586646f4ddf14782e4c1454c60", size = 6872, upload-time = "2026-08-07T00:13:07.492Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/9b/43/832f631d32e4f1211caa2ba368317739fe71f0b8530e4c9d15dc454bac2a/httpx2_jsfetch-1.0-py3-none-any.whl", hash = "sha256:cb916b707601e69a07721aabc8f3f6659be3a6893bc1ff5c6f9e02241df2da32", size = 6382, upload-time = "2026-08-07T00:13:06.567Z" },
+]
+
 [[package]]
 name = "idna"
 version = "3.20"
@@ -1177,6 +1226,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/b7/ce/149a00dd41f10bc29e5921b496af8b574d8413afcd5e30dfa0ed46c2cc5e/six-1.17.0-py2.py3-none-any.whl", hash = "sha256:4721f391ed90541fddacab5acf947aa0d3dc7d27b2e1e8eda2be8970586c3274", size = 11050, upload-time = "2024-12-04T17:35:26.475Z" },
 ]
 
+[[package]]
+name = "sniffio"
+version = "1.3.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/a2/87/a6771e1546d97e7e041b6ae58d80074f81b7d5121207425c964ddf5cfdbd/sniffio-1.3.1.tar.gz", hash = "sha256:f4324edc670a0f49750a81b895f35c3adb843cca46f0530f79fc1babb23789dc", size = 20372, upload-time = "2024-02-25T23:20:04.057Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e9/44/75a9c9421471a6c4805dbf2356f7c181a29c1879239abab1ea2cc8f38b40/sniffio-1.3.1-py3-none-any.whl", hash = "sha256:2f6da418d1f1e0fddd844478f41680e794e6051915791a034ff65e5f100525a2", size = 10235, upload-time = "2024-02-25T23:20:01.196Z" },
+]
+
 [[package]]
 name = "sqlalchemy"
 version = "2.1.6"
@@ -1276,6 +1334,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/ad/66/0d785f0bc5e4315a96c989bb476d0fc07ea4f85132550c7b156ca2035d52/traitlets-5.16.1-py3-none-any.whl", hash = "sha256:f775618166caa0396c8e337099240f2bd3e5e917d203b2e6fbe21a58d3cb1f6b", size = 86211, upload-time = "2026-08-03T08:32:34.48Z" },
 ]
 
+[[package]]
+name = "truststore"
+version = "0.10.5"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/ee/9f/c5201d42a484c061e528825fc8e2d565f5abd50a4ced6fb7d29c4ec99b2b/truststore-0.10.5.tar.gz", hash = "sha256:30d36967ccaded5cbb38d602c433f53600036c79d502f4533a49b60a03bbefcd", size = 28091, upload-time = "2026-10-12T22:27:31.808Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/51/e9/3a7820be2bb0fe53b6bc9c3be26d3d1158004e4c3ab953aa6840b955b1e9/truststore-0.10.5-py3-none-any.whl", hash = "sha256:9aaaedaefaf06d8b206278cf8b5012bc897f485a874503501e12d776df78951c", size = 19017, upload-time = "2026-10-12T22:27:30.377Z" },
+]
+
 [[package]]
 name = "tryceratops"
 version = "2.4.1"
@@ -1530,3 +1597,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/31/06/c969504aa9887010bac12484df2f03aa0e7ca5ada615a989da04e8f33046/whenever-0.10.5-cp314-cp314t-win_amd64.whl", hash = "sha256:06c32f8931bbc52b6b3f513e831aeb936e231f038b2a66cd11d4a00c131b82e6", size = 544821, upload-time = "2026-08-07T15:02:39.579Z" },
     { url = "https://files.pythonhosted.org/packages/f2/84/c937c7ec9de2aa59fbcdd60b52a8da60069636d2252c7461e43d389c04a4/whenever-0.10.5-py3-none-any.whl", hash = "sha256:193091633fc1dece4c86e7303cbd4039320b4f45c69dc5ae69b8f711bca82603", size = 129302, upload-time = "2026-08-07T15:03:29.285Z" },
 ]
+
+[[package]]
+name = "wsproto"
+version = "1.3.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "h11" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/c7/79/12135bdf8b9c9367b8701c2c19a14c913c120b882d50b014ca0d38083c2c/wsproto-1.3.2.tar.gz", hash = "sha256:b86885dcf294e15204919950f666e06ffc6c7c114ca900b060d6e16293528294", size = 50116, upload-time = "2025-11-20T18:18:01.871Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a4/f5/10b68b7b1544245097b2a1b8238f66f2fc6dcaeb24ba5d917f52bd2eed4f/wsproto-1.3.2-py3-none-any.whl", hash = "sha256:61eea322cdf56e8cc904bd3ad7573359a242ba65688716b0710a5eb12beab584", size = 24405, upload-time = "2025-11-20T18:18:00.454Z" },
+]