fix(models): synchronize canonical runtime selection

This commit is contained in:
Xubin Ren
2026-08-16 11:50:56 +08:00
parent 0a6ee1c539
commit 731b8fc2ed
15 changed files with 265 additions and 16 deletions
+9 -3
View File
@@ -533,9 +533,15 @@ class AgentLoop:
self.subagents.max_iterations = self.max_iterations
def invalidate_runtime_config(self) -> None:
"""Invalidate runtime config and notify clients to refresh its catalog."""
"""Invalidate runtime config for lazy refresh at the next admission."""
self.runtime_resolver.invalidate()
self._publish_runtime_selection(self.runtime_resolver.runtime)
def refresh_runtime_config(self) -> LLMRuntime:
"""Refresh runtime config now and publish the canonical selection."""
self.runtime_resolver.invalidate()
runtime = self.runtime_resolver.admit()
self._publish_runtime_selection(runtime)
return runtime
def runtime_for_session(
self,
@@ -1787,7 +1793,7 @@ class AgentLoop:
session.provider_state = None
self.sessions.save(session)
ctx.input_persisted_early = True
ctx.delivery.record_runtime(runtime)
await ctx.delivery.runtime_admitted(runtime)
ctx.request_context = self._request_context_for_turn(ctx)
if ctx.kind is TurnKind.USER:
+9 -1
View File
@@ -189,7 +189,15 @@ class TurnDelivery:
started_at=started_at,
)
def record_runtime(self, runtime: LLMRuntime) -> None:
async def runtime_admitted(self, runtime: LLMRuntime) -> None:
"""Record the immutable runtime and expose it at the lifecycle seam."""
if self.route.publish_lifecycle:
await self.runtime_event_publisher.turn_runtime_admitted(
self.delivery_message,
self.session_key,
runtime,
)
return
self.runtime_event_publisher.record_turn_runtime(self.session_key, runtime)
def record_latency(self, latency_ms: int | None) -> None:
+2 -1
View File
@@ -84,9 +84,10 @@ class RuntimeModelUpdatedEvent(OutboundEvent):
@dataclass(frozen=True)
class TurnModelUpdatedEvent(OutboundEvent):
"""The fallback model currently handling one chat turn."""
"""The canonical preset and concrete model handling one chat turn."""
model: str
model_preset: str | None = None
def outbound_message_for_event(
+30
View File
@@ -40,6 +40,14 @@ class SessionTurnStarted:
context: RuntimeEventContext
@dataclass(frozen=True)
class TurnRuntimeAdmitted:
"""The immutable model runtime selected for one admitted turn."""
context: RuntimeEventContext
runtime: LLMRuntime
@dataclass(frozen=True)
class TurnRunStatusChanged:
"""Visible run status changed for a turn."""
@@ -85,6 +93,7 @@ class RuntimeModelChanged:
RuntimeEvent = (
SessionTurnStarted
| TurnRuntimeAdmitted
| SessionTurnPersisted
| TurnRunStatusChanged
| TurnCompleted
@@ -93,6 +102,7 @@ RuntimeEvent = (
)
RuntimeEventType = (
type[SessionTurnStarted]
| type[TurnRuntimeAdmitted]
| type[SessionTurnPersisted]
| type[TurnRunStatusChanged]
| type[TurnCompleted]
@@ -204,6 +214,26 @@ class RuntimeEventPublisher:
)
)
async def turn_runtime_admitted(
self,
msg: InboundMessage,
session_key: str,
runtime: LLMRuntime,
) -> None:
"""Record and publish the runtime selected for one turn."""
self.record_turn_runtime(session_key, runtime)
await self.bus.publish(
TurnRuntimeAdmitted(
context=self._context(
channel=msg.channel,
chat_id=msg.chat_id,
session_key=session_key,
metadata=msg.metadata,
),
runtime=runtime,
)
)
async def run_status_changed(
self,
msg: InboundMessage,
+4
View File
@@ -1407,6 +1407,7 @@ class WebSocketChannel(BaseChannel):
await self.send_turn_model_updated(
msg.chat_id,
model_name=event.model,
model_preset=event.model_preset,
)
return
if isinstance(event, GoalStateSyncEvent):
@@ -1774,6 +1775,7 @@ class WebSocketChannel(BaseChannel):
chat_id: str,
*,
model_name: Any,
model_preset: Any = None,
) -> None:
"""Notify one chat's subscribers which model is handling its current request."""
conns = list(self._subs.get(chat_id, ()))
@@ -1788,6 +1790,8 @@ class WebSocketChannel(BaseChannel):
"chat_id": chat_id,
"model_name": model_name.strip(),
}
if isinstance(model_preset, str) and model_preset.strip():
body["model_preset"] = model_preset.strip()
raw = json.dumps(body, ensure_ascii=False)
for connection in conns:
await self._safe_send_to(connection, raw, label=" turn_model_updated ")
@@ -1642,7 +1642,10 @@ async def test_send_scopes_turn_model_updates_to_the_subscribed_chat() -> None:
channel="websocket",
chat_id="chat-1",
content="",
event=TurnModelUpdatedEvent(model="deepseek/deepseek-chat"),
event=TurnModelUpdatedEvent(
model="deepseek/deepseek-chat",
model_preset="Deep Research",
),
)
)
@@ -1651,6 +1654,7 @@ async def test_send_scopes_turn_model_updates_to_the_subscribed_chat() -> None:
"event": "turn_model_updated",
"chat_id": "chat-1",
"model_name": "deepseek/deepseek-chat",
"model_preset": "Deep Research",
}
chat_two.send.assert_not_awaited()
+1 -1
View File
@@ -658,7 +658,7 @@ def _run_gateway(
return agent.model.strip() or None
def _webui_refresh_runtime_config() -> None:
agent.invalidate_runtime_config()
agent.refresh_runtime_config()
def _webui_skill_state_action(disabled_skills: set[str]) -> None:
config.agents.defaults.disabled_skills = sorted(disabled_skills)
+28 -1
View File
@@ -33,6 +33,7 @@ from nanobot.bus.runtime_events import (
SessionTurnStarted,
TurnCompleted,
TurnRunStatusChanged,
TurnRuntimeAdmitted,
)
from nanobot.providers.base import LLMProvider
from nanobot.providers.fallback_provider import FallbackModelObserver
@@ -459,7 +460,14 @@ def build_webui_fallback_model_observer(bus: MessageBus) -> FallbackModelObserve
outbound_message_for_event(
channel=context.channel,
chat_id=chat_id,
event=TurnModelUpdatedEvent(model=model),
event=TurnModelUpdatedEvent(
model=model,
model_preset=(
context.runtime.model_preset
if context.runtime is not None
else None
),
),
metadata=context.metadata,
)
)
@@ -486,6 +494,10 @@ class WebuiTurnCoordinator:
self._handle_run_status_changed,
TurnRunStatusChanged,
),
runtime_events.subscribe(
self._handle_turn_runtime_admitted,
TurnRuntimeAdmitted,
),
runtime_events.subscribe(
self._handle_turn_completed_event,
TurnCompleted,
@@ -537,6 +549,21 @@ class WebuiTurnCoordinator:
started_at=event.started_at,
)
async def _handle_turn_runtime_admitted(self, event: TurnRuntimeAdmitted) -> None:
if not self._is_websocket_event(event.context):
return
await self.bus.publish_outbound(
outbound_message_for_event(
channel=event.context.channel,
chat_id=event.context.chat_id,
event=TurnModelUpdatedEvent(
model=event.runtime.model,
model_preset=event.runtime.model_preset,
),
metadata=event.context.metadata,
)
)
async def _handle_turn_completed_event(self, event: TurnCompleted) -> None:
if not self._is_websocket_event(event.context):
return
+9 -2
View File
@@ -1646,6 +1646,11 @@ class ModelSettingsHandler:
self.settings = settings
self.logger = logger
def _refresh_runtime_config(self) -> None:
"""Make a successful model-settings mutation visible to live clients now."""
if self.settings.refresh_runtime_config is not None:
self.settings.refresh_runtime_config()
async def handle(
self,
action: str,
@@ -1655,6 +1660,7 @@ class ModelSettingsHandler:
try:
if action == "agent-update":
payload = self.settings.mutate(operations.update_agent, request.query)
self._refresh_runtime_config()
return SettingsRouteResult.success(
payload,
decorate_restart=True,
@@ -1667,8 +1673,7 @@ class ModelSettingsHandler:
request.query,
rename_model_preset=self.settings.rename_model_preset,
)
if self.settings.refresh_runtime_config is not None:
self.settings.refresh_runtime_config()
self._refresh_runtime_config()
return SettingsRouteResult.success(payload, decorate_restart=True)
mutation = {
@@ -1680,6 +1685,7 @@ class ModelSettingsHandler:
}.get(action)
if mutation is not None:
payload = self.settings.mutate(mutation, request.query)
self._refresh_runtime_config()
return SettingsRouteResult.success(payload, decorate_restart=True)
if action == "provider-update":
@@ -1690,6 +1696,7 @@ class ModelSettingsHandler:
payload, image_restart_cleared = await operations.apply_image_runtime_change(
payload
)
self._refresh_runtime_config()
return SettingsRouteResult.success(
payload,
decorate_restart=True,
+51 -4
View File
@@ -226,7 +226,7 @@ def test_named_default_refresh_is_used_by_sessions_without_override(tmp_path: Pa
@pytest.mark.asyncio
async def test_config_invalidation_notifies_clients_before_session_runtime_refresh(
async def test_config_invalidation_defers_canonical_notification_until_default_refresh(
tmp_path: Path,
) -> None:
provider = _provider("model-a")
@@ -266,12 +266,59 @@ async def test_config_invalidation_notifies_clients_before_session_runtime_refre
runtime = loop.runtime_for_session(session)
await asyncio.sleep(0)
assert [(event.model, event.model_preset) for event in published] == [
("model-a", "fast"),
]
assert published == []
assert runtime.model == "model-b"
assert loop.model_presets["fast"].model == "model-b"
assert loop.llm_runtime().model == "model-b"
await asyncio.sleep(0)
assert [(event.model, event.model_preset) for event in published] == [
("model-b", "fast"),
]
@pytest.mark.asyncio
async def test_config_refresh_publishes_renamed_canonical_preset(tmp_path: Path) -> None:
provider = _provider("model-a")
catalog = {"fast": ModelPresetConfig(model="model-a")}
default_name = "fast"
published: list[RuntimeModelChanged] = []
def load_preset(name: str) -> ProviderSnapshot:
return ProviderSnapshot(
provider=provider,
model=catalog[name].model,
context_window_tokens=16_000,
signature=(name, catalog[name].model),
model_preset=name,
)
loop = AgentLoop(
bus=MessageBus(),
provider=provider,
workspace=tmp_path,
model="model-a",
context_window_tokens=16_000,
provider_signature=("fast", "model-a"),
provider_snapshot_loader=lambda: load_preset(default_name),
model_presets=catalog,
preset_catalog_loader=lambda: catalog,
model_preset="fast",
preset_snapshot_loader=load_preset,
)
loop.runtime_events.subscribe(published.append, RuntimeModelChanged)
catalog["Codex"] = catalog.pop("fast")
default_name = "Codex"
runtime = loop.refresh_runtime_config()
await asyncio.sleep(0)
assert (runtime.model, runtime.model_preset) == ("model-a", "Codex")
assert [(event.model, event.model_preset) for event in published] == [
("model-a", "Codex"),
]
def test_next_turn_captures_generation_changed_after_previous_admission(
tmp_path: Path,
+32
View File
@@ -10,6 +10,7 @@ from nanobot.bus.runtime_events import (
SessionTurnStarted,
TurnCompleted,
TurnRunStatusChanged,
TurnRuntimeAdmitted,
)
@@ -123,6 +124,37 @@ async def test_runtime_event_publisher_consumes_turn_metadata_on_complete() -> N
assert second.runtime is None
@pytest.mark.asyncio
async def test_runtime_event_publisher_exposes_admitted_runtime() -> None:
bus = RuntimeEventBus()
seen: list[object] = []
publisher = RuntimeEventPublisher(bus)
msg = InboundMessage(
channel="websocket",
sender_id="user",
chat_id="chat-a",
content="hello",
)
runtime = object()
bus.subscribe(seen.append)
await publisher.turn_runtime_admitted(msg, "websocket:chat-a", runtime) # type: ignore[arg-type]
await publisher.turn_completed(
channel="websocket",
chat_id="chat-a",
session_key="websocket:chat-a",
metadata=None,
)
admitted = seen[0]
completed = seen[1]
assert isinstance(admitted, TurnRuntimeAdmitted)
assert admitted.runtime is runtime
assert admitted.context.chat_id == "chat-a"
assert isinstance(completed, TurnCompleted)
assert completed.runtime is runtime
@pytest.mark.asyncio
async def test_runtime_event_publisher_emits_persisted_turn_attributes() -> None:
bus = RuntimeEventBus()
+52
View File
@@ -7,7 +7,11 @@ import pytest
from nanobot.agent.tools.context import RequestContext, request_context
from nanobot.bus.events import InboundMessage
from nanobot.bus.outbound_events import GoalStatusEvent, TurnModelUpdatedEvent
from nanobot.bus.runtime_events import RuntimeEventBus, RuntimeEventContext, TurnRuntimeAdmitted
from nanobot.providers.base import GenerationSettings
from nanobot.session import webui_turns as wth
from nanobot.session.manager import SessionManager
from nanobot.utils.llm_runtime import LLMRuntime
from nanobot.webui.metadata import WEBSOCKET_TURN_OWNER_METADATA_KEY
@@ -144,10 +148,18 @@ async def test_fallback_model_is_scoped_to_its_websocket_chat() -> None:
bus.publish_outbound = AsyncMock()
observer = wth.build_webui_fallback_model_observer(bus)
runtime = LLMRuntime(
provider=MagicMock(),
model="openai/gpt-4.1",
generation=GenerationSettings(),
context_window_tokens=16_000,
model_preset="Deep Research",
)
with request_context(
RequestContext(
channel="websocket",
chat_id="chat-model",
runtime=runtime,
metadata={"webui": True},
)
):
@@ -159,6 +171,46 @@ async def test_fallback_model_is_scoped_to_its_websocket_chat() -> None:
assert outbound.metadata == {"webui": True}
assert isinstance(outbound.event, TurnModelUpdatedEvent)
assert outbound.event.model == "deepseek/deepseek-chat"
assert outbound.event.model_preset == "Deep Research"
@pytest.mark.asyncio
async def test_admitted_runtime_publishes_chat_scoped_model_and_preset(tmp_path) -> None:
bus = MagicMock()
bus.publish_outbound = AsyncMock()
runtime_events = RuntimeEventBus()
coordinator = wth.WebuiTurnCoordinator(
bus=bus,
sessions=SessionManager(tmp_path),
schedule_background=lambda coro: coro.close(),
)
coordinator.subscribe(runtime_events)
runtime = LLMRuntime(
provider=MagicMock(),
model="openai-codex/gpt-5.6",
generation=GenerationSettings(),
context_window_tokens=262_144,
model_preset="Codex",
)
await runtime_events.publish(
TurnRuntimeAdmitted(
context=RuntimeEventContext(
channel="websocket",
chat_id="chat-model",
session_key="websocket:chat-model",
metadata={"webui": True},
),
runtime=runtime,
)
)
outbound = bus.publish_outbound.await_args.args[0]
assert outbound.channel == "websocket"
assert outbound.chat_id == "chat-model"
assert isinstance(outbound.event, TurnModelUpdatedEvent)
assert outbound.event.model == "openai-codex/gpt-5.6"
assert outbound.event.model_preset == "Codex"
@pytest.mark.asyncio
+30 -2
View File
@@ -307,6 +307,18 @@ async def test_oauth_completion_reads_websocket_payload(
@pytest.mark.parametrize(
("route_path", "function_name", "payload", "expected_query"),
[
(
"/api/settings/update",
"update_agent_settings",
{"model_preset": "Codex"},
{"model_preset": ["Codex"]},
),
(
"/api/settings/model-configurations/create",
"create_model_configuration",
{"name": "Codex", "model": "openai-codex/gpt-5.6"},
{"name": ["Codex"], "model": ["openai-codex/gpt-5.6"]},
),
(
"/api/settings/model-configurations/delete",
"delete_model_configuration",
@@ -325,10 +337,22 @@ async def test_oauth_completion_reads_websocket_payload(
{"order": ["backup"]},
{"order": ['["backup"]']},
),
(
"/api/settings/provider/create",
"create_provider_settings",
{"name": "team", "api_base": "https://llm.example/v1"},
{"name": ["team"], "api_base": ["https://llm.example/v1"]},
),
(
"/api/settings/provider/update",
"update_provider_settings",
{"provider": "team", "api_base": "https://llm.example/v2"},
{"provider": ["team"], "api_base": ["https://llm.example/v2"]},
),
],
)
@pytest.mark.asyncio
async def test_model_preset_mutation_routes(
async def test_runtime_config_mutation_routes_refresh_live_runtime(
monkeypatch,
route_path: str,
function_name: str,
@@ -336,6 +360,7 @@ async def test_model_preset_mutation_routes(
expected_query: dict[str, list[str]],
) -> None:
captured: dict[str, object] = {}
refresh_runtime_config = MagicMock()
def mutate(query, *, config_path=None):
captured["query"] = query
@@ -344,12 +369,15 @@ async def test_model_preset_mutation_routes(
monkeypatch.setattr(f"nanobot.webui.settings_routes.{function_name}", mutate)
request = _mutation_request(route_path, payload)
response = await _router().dispatch(None, request, route_path)
response = await _router(
refresh_runtime_config=refresh_runtime_config,
).dispatch(None, request, route_path)
assert response is not None
assert response.status_code == 200
assert json.loads(response.body)["routed"] == function_name
assert captured["query"] == expected_query
refresh_runtime_config.assert_called_once_with()
@pytest.mark.asyncio
+1
View File
@@ -1271,6 +1271,7 @@ export type InboundEvent =
event: "turn_model_updated";
chat_id: string;
model_name: string;
model_preset?: string | null;
}
| ({
event: "turn_end";
+2
View File
@@ -1636,12 +1636,14 @@ describe("NanobotClient", () => {
event: "turn_model_updated",
chat_id: "chat-a",
model_name: "deepseek/deepseek-chat",
model_preset: "Deep Research",
});
expect(chatHandler).toHaveBeenCalledWith({
event: "turn_model_updated",
chat_id: "chat-a",
model_name: "deepseek/deepseek-chat",
model_preset: "Deep Research",
});
});