agent-lab/tests/test_websocket_api.py

import json
import asyncio
import logging
import subprocess
from collections.abc import AsyncIterator
from html.parser import HTMLParser
from pathlib import Path

import httpx
import pytest
from fastapi.testclient import TestClient
from pydantic import ValidationError

from agent_lab.application.contracts import AgentParams, DebugRunRequest, EventAgentParams
from agent_lab.application.queues import RuntimeQueues
from agent_lab.application.runtime import DebugRuntime
from agent_lab.domain.events import ToolCallEvent
from agent_lab.domain.messages import ChatMessage, StreamItem, TokenUsage
from agent_lab.infrastructure.chat_client import OpenAICompatibleChatClient
from agent_lab.infrastructure.sqlite_store import SQLiteSessionStore
from agent_lab.presentation.web import create_app
from agent_lab.settings import Settings


def _css_rule(css: str, selector: str) -> str:
    start = css.index(f"{selector} {{")
    end = css.index("}", start)
    return css[start:end]


class _HtmlNode:
    def __init__(self, tag: str, attrs: list[tuple[str, str | None]]) -> None:
        self.tag = tag
        self.attrs = dict(attrs)
        self.children: list["_HtmlNode"] = []
        self.text_parts: list[str] = []

    @property
    def classes(self) -> set[str]:
        return set(self.attrs.get("class", "").split())

    def descendants(self) -> list["_HtmlNode"]:
        nodes: list["_HtmlNode"] = []
        for child in self.children:
            nodes.append(child)
            nodes.extend(child.descendants())
        return nodes

    def text_content(self) -> str:
        return "".join(
            [*self.text_parts, *(child.text_content() for child in self.children)]
        )


class _HtmlTreeParser(HTMLParser):
    _void_tags = {
        "area",
        "base",
        "br",
        "col",
        "embed",
        "hr",
        "img",
        "input",
        "link",
        "meta",
        "param",
        "source",
        "track",
        "wbr",
    }

    def __init__(self) -> None:
        super().__init__()
        self.root = _HtmlNode("document", [])
        self.stack = [self.root]

    def handle_starttag(
        self,
        tag: str,
        attrs: list[tuple[str, str | None]],
    ) -> None:
        node = _HtmlNode(tag, attrs)
        self.stack[-1].children.append(node)
        if tag not in self._void_tags:
            self.stack.append(node)

    def handle_startendtag(
        self,
        tag: str,
        attrs: list[tuple[str, str | None]],
    ) -> None:
        self.stack[-1].children.append(_HtmlNode(tag, attrs))

    def handle_endtag(self, tag: str) -> None:
        for index in range(len(self.stack) - 1, 0, -1):
            if self.stack[index].tag == tag:
                del self.stack[index:]
                return

    def handle_data(self, data: str) -> None:
        self.stack[-1].text_parts.append(data)


def _parse_html(html: str) -> _HtmlNode:
    parser = _HtmlTreeParser()
    parser.feed(html)
    return parser.root


def _html_node_by_id(root: _HtmlNode, element_id: str) -> _HtmlNode:
    node = next(
        (
            node
            for node in [root, *root.descendants()]
            if node.attrs.get("id") == element_id
        ),
        None,
    )
    assert node is not None, f"missing HTML element #{element_id}"
    return node


def _run_node_json(source: str) -> object:
    completed = subprocess.run(
        ["node", "--input-type=module", "-e", source],
        capture_output=True,
        text=True,
        check=False,
    )
    assert completed.returncode == 0, completed.stderr
    return json.loads(completed.stdout)


def _console_ui_source(js: str) -> str:
    assert 'const CONSOLE_UI_STORAGE_KEY = "agent-lab-console-ui-v1";' in js
    for function_name in [
        "readConsoleUiState",
        "writeConsoleUiState",
        "setControlSectionExpanded",
        "setDebugDrawerOpen",
        "initializeConsoleUi",
        "updateTopbarStatus",
    ]:
        assert f"function {function_name}(" in js
    return js[
        js.index("function readConsoleUiState(") :
        js.index("function openChatConfig(")
    ]


def _run_load_session_replay_scenario(scenario: str) -> object:
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    load_replay = js[
        js.index("async function loadSessionReplay(") :
        js.index("function setSessionStatus(")
    ]
    harness = """
let currentSessionId = null;
let sessionReplayLoadToken = 0;
const sessionTitle = { value: "" };
const sessionList = { value: "" };
const mutations = {
  workspace: [],
  transcript: [],
  audit: [],
  usage: [],
  status: [],
};
const pending = new Map();

function fetchJson(url) {
  let resolve;
  let reject;
  const promise = new Promise((resolvePromise, rejectPromise) => {
    resolve = resolvePromise;
    reject = rejectPromise;
  });
  pending.set(url, { resolve, reject });
  return promise;
}

function resolveReplay(sessionId) {
  pending.get(`/api/sessions/${sessionId}`).resolve({
    title: `${sessionId} title`,
    config: { marker: sessionId },
  });
  pending.get(`/api/sessions/${sessionId}/messages`).resolve([
    { content: sessionId },
  ]);
  pending.get(`/api/sessions/${sessionId}/audit`).resolve([
    { event: sessionId },
  ]);
  pending.get(`/api/sessions/${sessionId}/usage`).resolve({
    session: { marker: sessionId },
  });
}

function restoreWorkspaceSnapshot(config) {
  mutations.workspace.push(config.marker);
}
function renderReplayMessages(messages) {
  mutations.transcript.push(messages[0].content);
}
function renderAuditReplay(audit) {
  mutations.audit.push(audit[0].event);
}
function renderSessionUsage(usage) {
  mutations.usage.push(usage.session.marker);
}
function setSessionStatus(status) {
  mutations.status.push(status);
}
function isSocketOpen() {
  return false;
}
"""
    return _run_node_json("\n".join([harness, load_replay, scenario]))


class FakeRuntime:
    def __init__(self) -> None:
        self.requests: list[DebugRunRequest] = []

    async def run(self, request: DebugRunRequest) -> AsyncIterator[dict]:
        self.requests.append(request)
        yield {"type": "session_started"}
        yield {"type": "message_delta", "content": "hello"}
        yield {"type": "done"}


class QueueAwareRuntime:
    def __init__(self) -> None:
        self.requests: list[DebugRunRequest] = []
        self.queues: RuntimeQueues | None = None
        self.task: asyncio.Task | None = None

    def start(self, request: DebugRunRequest) -> RuntimeQueues:
        self.requests.append(request)
        self.queues = RuntimeQueues()
        self.task = asyncio.create_task(self._run())
        return self.queues

    async def _run(self) -> None:
        assert self.queues is not None
        await self.queues.output.put({"type": "session_started"})
        message = await self.queues.input.get()
        await self.queues.output.put(
            {"type": "message_delta", "content": message.content}
        )
        await self.queues.output.put({"type": "done"})

    async def aclose(self) -> None:
        if self.task is not None and not self.task.done():
            self.task.cancel()
            await asyncio.gather(self.task, return_exceptions=True)


class PersistentSessionRuntime:
    def __init__(self) -> None:
        self.requests: list[DebugRunRequest] = []
        self.queues: RuntimeQueues | None = None
        self.task: asyncio.Task | None = None

    def start_session(self, request: DebugRunRequest) -> RuntimeQueues:
        self.requests.append(request)
        self.queues = RuntimeQueues()
        self.task = asyncio.create_task(self._run(request))
        return self.queues

    async def _run(self, request: DebugRunRequest) -> None:
        assert self.queues is not None
        await self.queues.output.put({"type": "session_started"})
        await self._emit_turn(1, request.user_message)
        turn_index = 1
        while True:
            message = await self.queues.input.get()
            turn_index += 1
            await self._emit_turn(turn_index, message.content)

    async def _emit_turn(self, turn_index: int, content: str) -> None:
        assert self.queues is not None
        await self.queues.output.put({"type": "turn_started", "turn_index": turn_index})
        await self.queues.output.put({"type": "message_delta", "content": content})
        await self.queues.output.put(
            {"type": "turn_completed", "turn_index": turn_index}
        )

    async def aclose(self) -> None:
        if self.task is not None and not self.task.done():
            self.task.cancel()
            await asyncio.gather(self.task, return_exceptions=True)


def _request_payload() -> dict:
    return {
        "user_message": "debug this",
        "system_prompts": ["You are a debugger."],
        "pre_messages": [{"role": "user", "content": "previous turn"}],
        "chat_agent": {
            "model": "fake-model",
            "temperature": 0.1,
            "max_tokens": 200,
        },
        "event_agent": {
            "enabled_tools": ["handoff_note"],
            "max_event_loops": 2,
        },
    }


def _event_tool_call_from_tools(
    tools: list[dict],
    messages: list[ChatMessage],
) -> StreamItem:
    tool_name = tools[0]["function"]["name"]
    content = ""
    for role in ("assistant", "user"):
        content = next(
            (
                message.content
                for message in reversed(messages)
                if message.role == role and message.content.strip()
            ),
            "",
        )
        if content:
            break
    arguments = {
        "message": content,
        "query": content,
        "title": content,
    }
    return StreamItem.provider_tool_call(
        ToolCallEvent(
            id="event_agent_call_1",
            name=tool_name,
            arguments=arguments,
            raw_arguments=json.dumps(arguments),
        )
    )


def test_agent_params_defaults_include_extra_body_and_event_agent_defaults_to_one_round():
    chat_params = AgentParams()
    event_params = EventAgentParams()

    assert chat_params.extra_body == {
        "thinking": {"type": "disabled"},
        "enable_search": False,
        "search_options": {"forced_search": False},
    }
    assert event_params.extra_body == chat_params.extra_body
    assert event_params.max_event_loops == 1
    assert event_params.max_parallel_events == 4
    assert event_params.batch_timeout_seconds == 15.0
    assert event_params.system_prompt == ""


@pytest.mark.parametrize("value", [0, 17])
def test_event_agent_params_rejects_out_of_range_parallelism(value: int):
    with pytest.raises(ValidationError, match="max_parallel_events"):
        EventAgentParams(max_parallel_events=value)


@pytest.mark.parametrize("value", [0, -1, 121])
def test_event_agent_params_rejects_invalid_batch_timeout(value: float):
    with pytest.raises(ValidationError, match="batch_timeout_seconds"):
        EventAgentParams(batch_timeout_seconds=value)


def test_debug_run_request_defaults_to_dual_agent_tool_invocation_mode():
    request = DebugRunRequest.model_validate(_request_payload())

    assert request.tool_invocation_mode == "dual_agent"


@pytest.mark.parametrize("mode", ["dual_agent", "chat_agent_tools"])
def test_debug_run_request_accepts_supported_tool_invocation_modes(mode: str):
    payload = _request_payload()
    payload["tool_invocation_mode"] = mode

    request = DebugRunRequest.model_validate(payload)

    assert request.tool_invocation_mode == mode


def test_debug_run_request_rejects_unknown_tool_invocation_mode():
    payload = _request_payload()
    payload["tool_invocation_mode"] = "unknown"

    with pytest.raises(ValidationError, match="tool_invocation_mode"):
        DebugRunRequest.model_validate(payload)


def test_health_returns_ok():
    app = create_app(runtime_factory=FakeRuntime)
    client = TestClient(app)

    response = client.get("/health")

    assert response.status_code == 200
    assert response.json() == {"status": "ok"}


def test_api_tools_returns_default_tool_catalog():
    app = create_app(runtime_factory=FakeRuntime)
    client = TestClient(app)

    response = client.get("/api/tools")

    assert response.status_code == 200
    tools = response.json()
    assert [tool["name"] for tool in tools] == [
        "handoff_note",
        "mock_search",
        "mock_ticket",
        "session.terminate",
        "device.volume.adjust",
        "calendar.schedule.create",
        "knowledge.web.search",
    ]
    assert tools[0]["parameters"]["required"] == ["message"]


def test_session_api_creates_lists_and_returns_empty_replay_data(tmp_path):
    settings = Settings(database_path=str(tmp_path / "agent_lab.sqlite3"))
    app = create_app(settings=settings, runtime_factory=FakeRuntime)
    client = TestClient(app)

    created = client.post(
        "/api/sessions",
        json={
            "title": "Debug session",
            "config": {"chat_agent": {"model": "chat-model"}},
        },
    )

    assert created.status_code == 200
    session_id = created.json()["id"]
    assert created.json()["title"] == "Debug session"

    sessions = client.get("/api/sessions")
    assert sessions.status_code == 200
    assert sessions.json()[0]["id"] == session_id
    assert sessions.json()[0]["turn_count"] == 0

    detail = client.get(f"/api/sessions/{session_id}")
    assert detail.status_code == 200
    assert detail.json()["config"]["chat_agent"]["model"] == "chat-model"

    assert client.get(f"/api/sessions/{session_id}/messages").json() == []
    assert client.get(f"/api/sessions/{session_id}/audit").json() == []
    usage = client.get(f"/api/sessions/{session_id}/usage").json()
    assert usage == {
        "calls": [],
        "turns": [],
        "session": {
            "prompt_tokens": 0,
            "completion_tokens": 0,
            "total_tokens": 0,
            "cached_tokens": 0,
            "elapsed_ms": 0,
            "call_count": 0,
            "fallback_count": 0,
            "tool_count": 0,
            "turn_wall_time_ms": None,
        },
    }


def test_session_api_returns_persisted_replay_data(tmp_path):
    database_path = tmp_path / "agent_lab.sqlite3"
    settings = Settings(database_path=str(database_path))
    app = create_app(settings=settings, runtime_factory=FakeRuntime)
    client = TestClient(app)

    session_id = client.post("/api/sessions", json={"title": "Replay"}).json()["id"]
    store = SQLiteSessionStore(database_path)
    store.start_turn(session_id, turn_index=1, user_message="debug this")
    store.append_message(
        session_id,
        turn_index=1,
        message=ChatMessage(role="user", content="debug this"),
    )
    store.append_message(
        session_id,
        turn_index=1,
        message=ChatMessage(
            role="assistant",
            content="answer",
            tool_calls=[
                ToolCallEvent(
                    id="call-1",
                    name="mock_search",
                    arguments={"query": "docs"},
                    raw_arguments='{"query":"docs"}',
                )
            ],
        ),
    )
    store.append_message(
        session_id,
        turn_index=1,
        message=ChatMessage(
            role="tool",
            content='{"results":[]}',
            name="mock_search",
            tool_call_id="call-1",
        ),
    )
    store.append_audit(
        session_id,
        event="chat_agent_request",
        details={"model": "chat-model"},
        turn_index=1,
        round_index=1,
    )
    store.append_usage(
        session_id,
        turn_index=1,
        round_index=1,
        usage=TokenUsage(prompt_tokens=1, completion_tokens=2, total_tokens=3),
        ttft_ms=10,
        elapsed_ms=20,
    )

    messages = client.get(f"/api/sessions/{session_id}/messages").json()
    assert [message["content"] for message in messages] == [
        "debug this",
        "answer",
        '{"results":[]}',
    ]
    assert messages[0]["tool_calls"] == []
    assert messages[1]["tool_calls"] == [
        {
            "id": "call-1",
            "name": "mock_search",
            "arguments": {"query": "docs"},
            "raw_arguments": '{"query":"docs"}',
        }
    ]
    assert messages[2]["tool_call_id"] == "call-1"
    assert messages[2]["tool_calls"] == []
    assert client.get(f"/api/sessions/{session_id}/audit").json()[0]["details"] == {
        "model": "chat-model"
    }
    usage = client.get(f"/api/sessions/{session_id}/usage").json()
    assert usage["calls"][0]["total_tokens"] == 3
    assert usage["turns"][0]["total_tokens"] == 3
    assert usage["session"]["total_tokens"] == 3


def test_default_websocket_runtime_factory_accepts_session_store(tmp_path):
    settings = Settings(database_path=str(tmp_path / "agent_lab.sqlite3"))
    app = create_app(settings=settings)
    client = TestClient(app)

    with client.websocket_connect("/ws/debug") as websocket:
        websocket.send_json({"chat_agent": {"model": "fake-model"}})
        message = websocket.receive_json()

    assert message["type"] == "error"
    assert "user_message" in message["message"]


def test_websocket_debug_streams_runtime_messages():
    runtime = FakeRuntime()
    app = create_app(runtime_factory=lambda: runtime)
    client = TestClient(app)

    with client.websocket_connect("/ws/debug") as websocket:
        websocket.send_json(_request_payload())

        assert websocket.receive_json() == {"type": "session_started"}
        assert websocket.receive_json() == {
            "type": "message_delta",
            "content": "hello",
        }
        assert websocket.receive_json() == {"type": "done"}

    assert runtime.requests[0].user_message == "debug this"
    assert runtime.requests[0].pre_messages[0].content == "previous turn"


def test_websocket_debug_logs_session_lifecycle(caplog):
    runtime = FakeRuntime()
    app = create_app(runtime_factory=lambda: runtime)
    client = TestClient(app)

    with caplog.at_level(logging.INFO, logger="agent_lab.presentation.web"):
        with client.websocket_connect("/ws/debug") as websocket:
            websocket.send_json(_request_payload())
            assert websocket.receive_json() == {"type": "session_started"}
            assert websocket.receive_json() == {
                "type": "message_delta",
                "content": "hello",
            }
            assert websocket.receive_json() == {"type": "done"}

    assert "websocket session accepted" in caplog.text
    assert "websocket request accepted" in caplog.text
    assert "websocket session closed" in caplog.text


def test_websocket_debug_enqueues_user_messages_during_running_session():
    runtime = QueueAwareRuntime()
    app = create_app(runtime_factory=lambda: runtime)
    client = TestClient(app)

    with client.websocket_connect("/ws/debug") as websocket:
        websocket.send_json(_request_payload())

        assert websocket.receive_json() == {"type": "session_started"}
        websocket.send_json({"type": "user_message", "content": "follow-up"})
        assert websocket.receive_json() == {
            "type": "message_delta",
            "content": "follow-up",
        }
        assert websocket.receive_json() == {"type": "done"}

    assert runtime.requests[0].user_message == "debug this"


def test_websocket_debug_keeps_session_open_for_multiple_turns():
    runtime = PersistentSessionRuntime()
    app = create_app(runtime_factory=lambda: runtime)
    client = TestClient(app)

    with client.websocket_connect("/ws/debug") as websocket:
        websocket.send_json(_request_payload())

        assert websocket.receive_json() == {"type": "session_started"}
        assert websocket.receive_json() == {"type": "turn_started", "turn_index": 1}
        assert websocket.receive_json() == {
            "type": "message_delta",
            "content": "debug this",
        }
        assert websocket.receive_json() == {"type": "turn_completed", "turn_index": 1}

        websocket.send_json({"type": "user_message", "content": "follow-up"})
        assert websocket.receive_json() == {"type": "turn_started", "turn_index": 2}
        assert websocket.receive_json() == {
            "type": "message_delta",
            "content": "follow-up",
        }
        assert websocket.receive_json() == {"type": "turn_completed", "turn_index": 2}
        websocket.close()

    assert runtime.requests[0].user_message == "debug this"


def test_websocket_debug_sends_error_for_invalid_request():
    app = create_app(runtime_factory=FakeRuntime)
    client = TestClient(app)

    with client.websocket_connect("/ws/debug") as websocket:
        websocket.send_json({"chat_agent": {"model": "fake-model"}})
        message = websocket.receive_json()

    assert message["type"] == "error"
    assert "user_message" in message["message"]


def test_debug_run_request_rejects_invalid_pre_message_role():
    payload = _request_payload()
    payload["pre_messages"] = [{"role": "developer", "content": "invalid"}]

    with pytest.raises(ValidationError) as exc_info:
        DebugRunRequest.model_validate(payload)

    assert "role" in str(exc_info.value)


def test_debug_run_request_rejects_tool_pre_messages():
    payload = _request_payload()
    payload["pre_messages"] = [
        {
            "role": "tool",
            "content": "orphan result",
            "tool_call_id": "call_1",
        }
    ]

    with pytest.raises(ValidationError, match="pre_messages cannot include tool messages"):
        DebugRunRequest.model_validate(payload)


def test_debug_run_request_rejects_assistant_tool_calls_in_pre_messages():
    payload = _request_payload()
    payload["pre_messages"] = [
        {
            "role": "assistant",
            "content": "I will inspect this.",
            "tool_calls": [
                {
                    "id": "call_1",
                    "name": "mock_search",
                    "arguments": {"query": "latency"},
                    "raw_arguments": '{"query":"latency"}',
                }
            ],
        }
    ]

    with pytest.raises(
        ValidationError,
        match="pre_messages cannot include assistant tool calls",
    ):
        DebugRunRequest.model_validate(payload)


def test_chat_message_allows_internal_tool_replies():
    message = ChatMessage(
        role="tool",
        content='{"message":"handled"}',
        name="handoff_note",
        tool_call_id="call_1",
    )

    assert message.role == "tool"
    assert message.tool_call_id == "call_1"


def test_chat_message_preserves_assistant_tool_calls():
    message = ChatMessage(
        role="assistant",
        content="I will inspect both sources.",
        tool_calls=[
            ToolCallEvent(
                id="call_1",
                name="mock_search",
                arguments={"query": "latency docs"},
                raw_arguments='{"query":"latency docs"}',
            ),
            ToolCallEvent(
                id="call_2",
                name="handoff_note",
                arguments={"message": "inspect provider behavior"},
                raw_arguments='{"message":"inspect provider behavior"}',
            ),
        ],
    )

    assert message.model_dump()["tool_calls"] == [
        {
            "id": "call_1",
            "name": "mock_search",
            "arguments": {"query": "latency docs"},
            "raw_arguments": '{"query":"latency docs"}',
        },
        {
            "id": "call_2",
            "name": "handoff_note",
            "arguments": {"message": "inspect provider behavior"},
            "raw_arguments": '{"message":"inspect provider behavior"}',
        },
    ]


def test_chat_message_rejects_tool_calls_on_non_assistant_role():
    with pytest.raises(ValidationError, match="tool_calls require assistant role"):
        ChatMessage(
            role="user",
            content="invalid",
            tool_calls=[
                ToolCallEvent(
                    id="call_1",
                    name="mock_search",
                    arguments={},
                    raw_arguments="{}",
                )
            ],
        )


def test_chat_message_rejects_tool_role_without_tool_call_id():
    with pytest.raises(ValidationError, match="tool messages require tool_call_id"):
        ChatMessage(role="tool", content="orphan result")


@pytest.mark.parametrize("role", ["system", "user", "assistant"])
def test_chat_message_rejects_tool_call_id_on_non_tool_roles(role: str):
    with pytest.raises(ValidationError, match="tool_call_id requires tool role"):
        ChatMessage(role=role, content="invalid", tool_call_id="call_1")


def test_chat_message_rejects_duplicate_assistant_tool_call_ids():
    with pytest.raises(ValidationError, match="duplicate assistant tool-call ID: call_1"):
        ChatMessage(
            role="assistant",
            content="checking twice",
            tool_calls=[
                ToolCallEvent(
                    id="call_1",
                    name="mock_search",
                    arguments={"query": "first"},
                    raw_arguments='{"query":"first"}',
                ),
                ToolCallEvent(
                    id="call_1",
                    name="handoff_note",
                    arguments={"message": "second"},
                    raw_arguments='{"message":"second"}',
                ),
            ],
        )


class HistoryCapturingChatClient:
    def __init__(self) -> None:
        self.calls = 0
        self.second_call_messages: list[ChatMessage] = []

    async def stream_chat(
        self,
        messages: list[ChatMessage],
        tools: list[dict],
        params: AgentParams,
        tool_choice: dict | None = None,
    ) -> AsyncIterator[StreamItem]:
        if tools:
            yield _event_tool_call_from_tools(tools, messages)
            return
        self.calls += 1
        if self.calls == 1:
            yield StreamItem.message_delta("Need event help.")
            yield StreamItem.text_event(
                ToolCallEvent(
                    id="call_1",
                    name="handoff_note",
                    arguments={},
                    raw_arguments="{}",
                )
            )
            return

        self.second_call_messages = list(messages)
        yield StreamItem.message_delta("Final answer.")


@pytest.mark.asyncio
async def test_runtime_appends_event_summary_without_tool_reply_history():
    request = DebugRunRequest(
        user_message="debug this",
        system_prompts=["You are a debugger."],
        pre_messages=[],
        chat_agent=AgentParams(model="fake-model", temperature=0.1, max_tokens=200),
        event_agent=EventAgentParams(enabled_tools=["handoff_note"], max_event_loops=2),
    )
    client = HistoryCapturingChatClient()
    runtime = DebugRuntime(client)

    outputs = [message async for message in runtime.run(request)]

    assert client.calls == 2
    assert [message.role for message in client.second_call_messages] == [
        "system",
        "system",
        "user",
        "assistant",
        "user",
    ]
    assert "Available events:" in client.second_call_messages[1].content
    assert client.second_call_messages[3].content == "Need event help."
    assert not any(message.role == "tool" for message in client.second_call_messages)
    assert client.second_call_messages[4].content.startswith("EventAgent results:\n")
    assert client.second_call_messages[4].name == "event_agent"
    assert outputs[-1] == {"type": "done"}


@pytest.mark.asyncio
async def test_openai_chat_client_streams_sse_chunks_through_parser():
    requests: list[httpx.Request] = []

    def handler(request: httpx.Request) -> httpx.Response:
        requests.append(request)
        payload = json.loads(request.content)
        assert payload["stream"] is True
        assert payload["model"] == "model-x"
        assert payload["messages"] == [{"role": "user", "content": "hi"}]
        assert payload["tools"][0]["function"]["name"] == "handoff_note"
        assert "tool_choice" not in payload
        assert payload["temperature"] == 0.3
        assert payload["max_tokens"] == 50
        assert request.headers["authorization"] == "Bearer test-key"
        return httpx.Response(
            200,
            content=(
                b'data: {"choices":[{"delta":{"content":"hi"},"finish_reason":null}]}\n\n'
                b"data: [DONE]\n\n"
            ),
        )

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="test-key",
            base_url="https://llm.test/v1",
            default_model="default-model",
            request_timeout_seconds=5,
            http_client=http_client,
        )
        items = [
            item
            async for item in client.stream_chat(
                messages=[ChatMessage(role="user", content="hi")],
                tools=[
                    {
                        "type": "function",
                        "function": {"name": "handoff_note", "parameters": {}},
                    }
                ],
                params=AgentParams(model="model-x", temperature=0.3, max_tokens=50),
            )
        ]

    assert requests[0].url.path == "/v1/chat/completions"
    assert [item.content for item in items if item.kind == "message_delta"] == ["hi"]
    assert [item.raw_chunk for item in items if item.kind == "raw_chunk"] == [
        {"choices": [{"delta": {"content": "hi"}, "finish_reason": None}]}
    ]


@pytest.mark.asyncio
async def test_openai_chat_client_serializes_explicit_tool_choice():
    captured_payloads: list[dict] = []
    forced_choice = {
        "type": "function",
        "function": {"name": "handoff_note"},
    }

    def handler(request: httpx.Request) -> httpx.Response:
        captured_payloads.append(json.loads(request.content))
        return httpx.Response(200, content=b"data: [DONE]\n\n")

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            http_client=http_client,
        )
        [
            item
            async for item in client.stream_chat(
                messages=[ChatMessage(role="user", content="prepare handoff")],
                tools=[
                    {
                        "type": "function",
                        "function": {"name": "handoff_note", "parameters": {}},
                    }
                ],
                params=AgentParams(model="provider-default"),
                tool_choice=forced_choice,
            )
        ]

    assert captured_payloads[0]["tool_choice"] == forced_choice


@pytest.mark.asyncio
async def test_openai_chat_client_flushes_one_provider_tool_call_at_done():
    def handler(request: httpx.Request) -> httpx.Response:
        return httpx.Response(
            200,
            content=(
                b'data: {"choices":[{"delta":{"tool_calls":[{"index":0,'
                b'"id":"call_1","function":{"name":"mock_search",'
                b'"arguments":"{\\"query\\":\\""}}]},"finish_reason":null}]}\n\n'
                b'data: {"choices":[{"delta":{"tool_calls":[{"index":0,'
                b'"function":{"arguments":"latency docs\\"}"}}]},'
                b'"finish_reason":null}]}\n\n'
                b"data: [DONE]\n\n"
            ),
        )

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            http_client=http_client,
        )
        items = [
            item
            async for item in client.stream_chat(
                messages=[ChatMessage(role="user", content="find docs")],
                tools=[
                    {
                        "type": "function",
                        "function": {"name": "mock_search", "parameters": {}},
                    }
                ],
                params=AgentParams(model="provider-default"),
            )
        ]

    provider_calls = [
        item.event for item in items if item.kind == "provider_tool_call"
    ]
    assert provider_calls == [
        ToolCallEvent(
            id="call_1",
            name="mock_search",
            arguments={"query": "latency docs"},
            raw_arguments='{"query":"latency docs"}',
        )
    ]


@pytest.mark.asyncio
async def test_openai_chat_client_uses_default_model_when_request_model_is_blank():
    captured_payloads: list[dict] = []

    def handler(request: httpx.Request) -> httpx.Response:
        captured_payloads.append(json.loads(request.content))
        return httpx.Response(
            200,
            content=b'data: {"choices":[{"delta":{"content":"hi"},"finish_reason":null}]}\n\n',
        )

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            http_client=http_client,
        )
        [
            item
            async for item in client.stream_chat(
                messages=[ChatMessage(role="user", content="hi")],
                tools=[],
                params=AgentParams(model="   ", temperature=0.3, max_tokens=50),
            )
        ]

    assert captured_payloads[0]["model"] == "provider-default"


@pytest.mark.asyncio
async def test_openai_chat_client_serializes_provider_tool_transcript():
    captured_payloads: list[dict] = []

    def handler(request: httpx.Request) -> httpx.Response:
        captured_payloads.append(json.loads(request.content))
        return httpx.Response(
            200,
            content=b'data: {"choices":[{"delta":{"content":"done"},"finish_reason":null}]}\n\n',
        )

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            http_client=http_client,
        )
        [
            item
            async for item in client.stream_chat(
                messages=[
                    ChatMessage(role="user", content="inspect both"),
                    ChatMessage(
                        role="assistant",
                        content="I will inspect both sources.",
                        tool_calls=[
                            ToolCallEvent(
                                id="call_1",
                                name="mock_search",
                                arguments={"query": "latency docs"},
                                raw_arguments='{"query":"latency docs"}',
                            ),
                            ToolCallEvent(
                                id="call_2",
                                name="handoff_note",
                                arguments={"message": "inspect provider behavior"},
                                raw_arguments='{"message":"inspect provider behavior"}',
                            ),
                        ],
                    ),
                    ChatMessage(
                        role="tool",
                        content='{"results":[]}',
                        name="mock_search",
                        tool_call_id="call_1",
                    ),
                    ChatMessage(
                        role="tool",
                        content='{"message":"handled"}',
                        name="handoff_note",
                        tool_call_id="call_2",
                    ),
                    ChatMessage(role="user", content="continue"),
                ],
                tools=[],
                params=AgentParams(
                    model="provider-default",
                    temperature=0.3,
                    max_tokens=50,
                ),
            )
        ]

    assert captured_payloads[0]["messages"] == [
        {"role": "user", "content": "inspect both"},
        {
            "role": "assistant",
            "content": "I will inspect both sources.",
            "tool_calls": [
                {
                    "id": "call_1",
                    "type": "function",
                    "function": {
                        "name": "mock_search",
                        "arguments": '{"query":"latency docs"}',
                    },
                },
                {
                    "id": "call_2",
                    "type": "function",
                    "function": {
                        "name": "handoff_note",
                        "arguments": '{"message":"inspect provider behavior"}',
                    },
                },
            ],
        },
        {
            "role": "tool",
            "content": '{"results":[]}',
            "tool_call_id": "call_1",
        },
        {
            "role": "tool",
            "content": '{"message":"handled"}',
            "tool_call_id": "call_2",
        },
        {"role": "user", "content": "continue"},
    ]


async def _assert_tool_transcript_rejected(
    messages: list[ChatMessage],
    error_match: str,
) -> None:
    network_called = False

    def handler(request: httpx.Request) -> httpx.Response:
        nonlocal network_called
        network_called = True
        return httpx.Response(200)

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            http_client=http_client,
        )

        with pytest.raises(ValueError, match=error_match):
            [
                item
                async for item in client.stream_chat(
                    messages=messages,
                    tools=[],
                    params=AgentParams(model="provider-default"),
                )
            ]

    assert network_called is False


@pytest.mark.asyncio
async def test_openai_chat_client_rejects_orphan_tool_reply_before_network():
    await _assert_tool_transcript_rejected(
        [
            ChatMessage(
                role="tool",
                content="orphan",
                tool_call_id="call_1",
            )
        ],
        "orphan tool reply: call_1",
    )


@pytest.mark.asyncio
async def test_openai_chat_client_rejects_mismatched_tool_call_id_before_network():
    await _assert_tool_transcript_rejected(
        [
            ChatMessage(
                role="assistant",
                content="",
                tool_calls=[
                    ToolCallEvent(
                        id="call_1",
                        name="mock_search",
                        arguments={},
                        raw_arguments="{}",
                    )
                ],
            ),
            ChatMessage(role="tool", content="wrong", tool_call_id="call_2"),
        ],
        "mismatched tool_call_id: call_2",
    )


@pytest.mark.asyncio
async def test_openai_chat_client_rejects_duplicate_tool_reply_before_network():
    await _assert_tool_transcript_rejected(
        [
            ChatMessage(
                role="assistant",
                content="",
                tool_calls=[
                    ToolCallEvent(
                        id="call_1",
                        name="mock_search",
                        arguments={},
                        raw_arguments="{}",
                    )
                ],
            ),
            ChatMessage(role="tool", content="first", tool_call_id="call_1"),
            ChatMessage(role="tool", content="duplicate", tool_call_id="call_1"),
        ],
        "duplicate tool reply: call_1",
    )


@pytest.mark.asyncio
async def test_openai_chat_client_rejects_non_tool_message_with_unresolved_calls():
    await _assert_tool_transcript_rejected(
        [
            ChatMessage(
                role="assistant",
                content="checking",
                tool_calls=[
                    ToolCallEvent(
                        id="call_1",
                        name="mock_search",
                        arguments={},
                        raw_arguments="{}",
                    ),
                    ToolCallEvent(
                        id="call_2",
                        name="handoff_note",
                        arguments={},
                        raw_arguments="{}",
                    ),
                ],
            ),
            ChatMessage(role="tool", content="done", tool_call_id="call_1"),
            ChatMessage(role="user", content="continue too early"),
        ],
        "unresolved tool calls before user message: call_2",
    )


@pytest.mark.asyncio
async def test_openai_chat_client_rejects_unfinished_tool_calls_at_end():
    await _assert_tool_transcript_rejected(
        [
            ChatMessage(
                role="assistant",
                content="checking",
                tool_calls=[
                    ToolCallEvent(
                        id="call_1",
                        name="mock_search",
                        arguments={},
                        raw_arguments="{}",
                    )
                ],
            )
        ],
        "unresolved tool calls at end: call_1",
    )


@pytest.mark.asyncio
async def test_openai_chat_client_rejects_reused_tool_call_id_across_transcript():
    await _assert_tool_transcript_rejected(
        [
            ChatMessage(
                role="assistant",
                content="first round",
                tool_calls=[
                    ToolCallEvent(
                        id="call_1",
                        name="mock_search",
                        arguments={},
                        raw_arguments="{}",
                    )
                ],
            ),
            ChatMessage(role="tool", content="done", tool_call_id="call_1"),
            ChatMessage(
                role="assistant",
                content="second round",
                tool_calls=[
                    ToolCallEvent(
                        id="call_1",
                        name="handoff_note",
                        arguments={},
                        raw_arguments="{}",
                    )
                ],
            ),
        ],
        "duplicate assistant tool-call ID across transcript: call_1",
    )


async def _assert_tool_choice_rejected(
    tools: list[dict],
    tool_choice: dict,
    error_match: str,
) -> None:
    network_called = False

    def handler(request: httpx.Request) -> httpx.Response:
        nonlocal network_called
        network_called = True
        return httpx.Response(200)

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            http_client=http_client,
        )
        with pytest.raises(ValueError, match=error_match):
            [
                item
                async for item in client.stream_chat(
                    messages=[ChatMessage(role="user", content="use a tool")],
                    tools=tools,
                    params=AgentParams(model="provider-default"),
                    tool_choice=tool_choice,
                )
            ]

    assert network_called is False


@pytest.mark.asyncio
async def test_openai_chat_client_rejects_forced_choice_without_tools():
    await _assert_tool_choice_rejected(
        tools=[],
        tool_choice={
            "type": "function",
            "function": {"name": "handoff_note"},
        },
        error_match="forced tool choice requires tools",
    )


@pytest.mark.asyncio
async def test_openai_chat_client_rejects_forced_choice_for_unknown_tool():
    await _assert_tool_choice_rejected(
        tools=[
            {
                "type": "function",
                "function": {"name": "mock_search", "parameters": {}},
            }
        ],
        tool_choice={
            "type": "function",
            "function": {"name": "handoff_note"},
        },
        error_match="forced tool choice references unknown tool: handoff_note",
    )


def test_static_pre_message_role_selector_does_not_offer_tool():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()

    assert '<option value="tool">tool</option>' not in html


def test_static_tools_ui_is_dynamic_and_not_hardcoded_to_handoff_note_checkbox():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert 'id="tool-handoff-note"' not in html
    assert "#tool-handoff-note" not in js
    assert 'id="tool-list"' in html
    assert '"/api/tools"' in js
    assert "fetch" in js


def test_static_workspace_snapshot_controls_are_available_outside_agent_config():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    chat_dialog_start = html.index('<dialog id="chat-config-dialog"')
    event_dialog_start = html.index('<dialog id="event-config-dialog"')
    chat_dialog_html = html[chat_dialog_start:event_dialog_start]
    sidebar_html = html[:chat_dialog_start]

    assert 'id="open-prompt-config"' not in html
    assert '<dialog id="prompt-dialog"' not in html
    assert 'id="workspace-snapshot-name"' in sidebar_html
    assert 'id="saved-workspace-snapshots"' in sidebar_html
    assert 'id="save-workspace-snapshot"' in sidebar_html
    assert 'id="load-workspace-snapshot"' in sidebar_html
    assert 'id="delete-workspace-snapshot"' in sidebar_html
    assert 'id="workspace-snapshot-name"' not in chat_dialog_html
    assert 'id="saved-workspace-snapshots"' not in chat_dialog_html
    assert 'id="save-workspace-snapshot"' not in chat_dialog_html
    assert 'id="system-prompts"' in chat_dialog_html
    assert 'id="pre-messages"' in chat_dialog_html


def test_static_session_ui_controls_and_replay_panels_are_available():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    chat_dialog_start = html.index('<dialog id="chat-config-dialog"')
    event_dialog_start = html.index('<dialog id="event-config-dialog"')
    chat_dialog_html = html[chat_dialog_start:event_dialog_start]
    event_dialog_html = html[event_dialog_start:]
    sidebar_html = html[:chat_dialog_start]

    for control_id in [
        "session-title",
        "session-list",
        "new-session",
        "refresh-sessions",
        "load-session",
        "session-status",
    ]:
        assert f'id="{control_id}"' in sidebar_html
        assert f'id="{control_id}"' not in chat_dialog_html
        assert f'id="{control_id}"' not in event_dialog_html

    for replay_id in [
        "audit-replay",
        "audit-count",
        "session-usage-summary",
        "session-usage-turns",
        "session-usage-calls",
    ]:
        assert f'id="{replay_id}"' in html


def test_static_professional_console_has_semantic_three_region_shell():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    root = _parse_html(html)
    layout = _html_node_by_id(root, "console-layout")
    direct_regions = layout.children

    assert layout.tag == "main"
    assert "layout" in layout.classes
    assert [(node.tag, node.attrs.get("aria-label")) for node in direct_regions] == [
        ("aside", "Runtime controls"),
        ("section", "Conversation workspace"),
        ("aside", "Audit and usage"),
    ]
    assert "controls" in direct_regions[0].classes
    assert "workspace" in direct_regions[1].classes
    assert "debug-drawer" in direct_regions[2].classes
    assert direct_regions[2].attrs["id"] == "debug-drawer"

    assert "Agent Lab" in html
    assert "Chat and event runtime console" in html
    for status_id in [
        "current-session-label",
        "current-mode-label",
        "connection-status",
    ]:
        assert "status-badge" in _html_node_by_id(root, status_id).classes

    drawer_toggle = _html_node_by_id(root, "toggle-debug-drawer")
    assert drawer_toggle.attrs["aria-expanded"] == "true"
    assert drawer_toggle.attrs["aria-controls"] == "debug-drawer"

    workspace = direct_regions[1]
    assert [child.attrs.get("class") for child in workspace.children] == [
        "workspace-body",
        "stats",
        "composer",
    ]
    assert _html_node_by_id(workspace, "messages").attrs["class"] == "messages"
    assert _html_node_by_id(workspace, "chat-form").tag == "form"

    drawer_ids = {
        node.attrs["id"]
        for node in [direct_regions[2], *direct_regions[2].descendants()]
        if "id" in node.attrs
    }
    workspace_ids = {
        node.attrs["id"]
        for node in [workspace, *workspace.descendants()]
        if "id" in node.attrs
    }
    for replay_id in [
        "audit-count",
        "audit-summary",
        "open-audit-dialog",
        "session-usage-summary",
        "session-usage-turns",
        "session-usage-calls",
    ]:
        assert replay_id in drawer_ids
        assert replay_id not in workspace_ids


def test_static_professional_console_control_sections_have_accessible_defaults():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    root = _parse_html(html)
    controls = next(
        node
        for node in _html_node_by_id(root, "console-layout").children
        if "controls" in node.classes
    )
    sections = [child for child in controls.children if "control-section" in child.classes]
    toggles = [
        node
        for node in root.descendants()
        if "control-section-toggle" in node.classes
    ]

    assert len(sections) == 4
    assert len(toggles) == 4

    expected_defaults = {
        "session-controls": ("Session", True),
        "tool-controls": ("Tools", True),
        "agent-config-controls": ("Agent Config", False),
        "snapshot-controls": ("Snapshot", False),
    }
    for section, (target_id, (title, expanded)) in zip(
        sections,
        expected_defaults.items(),
        strict=True,
    ):
        section_toggles = [
            node
            for node in section.children
            if "control-section-toggle" in node.classes
        ]
        assert len(section_toggles) == 1
        toggle = section_toggles[0]
        body = _html_node_by_id(section, target_id)

        assert toggle.tag == "button"
        assert toggle.attrs["type"] == "button"
        assert toggle.attrs["data-collapse-target"] == target_id
        assert toggle.attrs["aria-controls"] == target_id
        assert toggle.attrs["aria-expanded"] == str(expanded).lower()
        assert title in toggle.text_content()
        assert ("hidden" in body.attrs) is not expanded


def test_static_console_ui_section_click_opens_hidden_content_and_persists():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    console_ui = _console_ui_source(js)
    result = _run_node_json(
        "\n".join(
            [
                'const CONSOLE_UI_STORAGE_KEY = "agent-lab-console-ui-v1";',
                """
function makeElement({ dataset = {}, expanded = "false", hidden = false } = {}) {
  const listeners = {};
  return {
    dataset,
    hidden,
    textContent: "",
    attributes: { "aria-expanded": expanded },
    classList: { toggle() {} },
    addEventListener(type, listener) { listeners[type] = listener; },
    click() { listeners.click(); },
    getAttribute(name) { return this.attributes[name]; },
    setAttribute(name, value) { this.attributes[name] = String(value); },
  };
}
const sectionButton = makeElement({
  dataset: { collapseTarget: "agent-config-controls" },
  expanded: "false",
});
const sectionBody = makeElement({ hidden: true });
const drawerButton = makeElement({ expanded: "true" });
drawerButton.textContent = "Hide Debug";
const drawer = makeElement();
const layout = makeElement();
const elements = {
  "#agent-config-controls": sectionBody,
  "#toggle-debug-drawer": drawerButton,
  "#debug-drawer": drawer,
  "#console-layout": layout,
};
const document = {
  querySelector(selector) { return elements[selector] || null; },
  querySelectorAll(selector) {
    return selector === ".control-section-toggle" ? [sectionButton] : [];
  },
};
let stored = JSON.stringify({
  debug_drawer_open: false,
  sections: { "agent-config-controls": false },
});
globalThis.localStorage = {
  getItem() { return stored; },
  setItem(key, value) { stored = value; },
};
""",
                console_ui,
                """
initializeConsoleUi();
sectionButton.click();
console.log(JSON.stringify({
  expanded: sectionButton.getAttribute("aria-expanded"),
  hidden: sectionBody.hidden,
  stored: JSON.parse(stored),
}));
""",
            ]
        )
    )

    assert result == {
        "expanded": "true",
        "hidden": False,
        "stored": {
            "debug_drawer_open": False,
            "sections": {"agent-config-controls": True},
        },
    }


def test_static_console_ui_drawer_click_updates_layout_visibility_and_control():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    console_ui = _console_ui_source(js)
    result = _run_node_json(
        "\n".join(
            [
                'const CONSOLE_UI_STORAGE_KEY = "agent-lab-console-ui-v1";',
                """
function makeElement({ expanded = "true" } = {}) {
  const listeners = {};
  const classes = new Set();
  return {
    hidden: false,
    textContent: "Hide Debug",
    attributes: { "aria-expanded": expanded },
    classList: {
      toggle(name, force) {
        if (force) classes.add(name); else classes.delete(name);
      },
      contains(name) { return classes.has(name); },
    },
    addEventListener(type, listener) { listeners[type] = listener; },
    click() { listeners.click(); },
    getAttribute(name) { return this.attributes[name]; },
    setAttribute(name, value) { this.attributes[name] = String(value); },
  };
}
const drawerButton = makeElement();
const drawer = makeElement();
const layout = makeElement();
const elements = {
  "#toggle-debug-drawer": drawerButton,
  "#debug-drawer": drawer,
  "#console-layout": layout,
};
const document = {
  querySelector(selector) { return elements[selector] || null; },
  querySelectorAll() { return []; },
};
let stored = JSON.stringify({ sections: { "snapshot-controls": true } });
globalThis.localStorage = {
  getItem() { return stored; },
  setItem(key, value) { stored = value; },
};
function snapshot() {
  return {
    closed: layout.classList.contains("debug-drawer-closed"),
    hidden: drawer.hidden,
    text: drawerButton.textContent,
    expanded: drawerButton.getAttribute("aria-expanded"),
    stored: JSON.parse(stored),
  };
}
""",
                console_ui,
                """
initializeConsoleUi();
drawerButton.click();
const closed = snapshot();
drawerButton.click();
const opened = snapshot();
console.log(JSON.stringify({ closed, opened }));
""",
            ]
        )
    )

    assert result["closed"] == {
        "closed": True,
        "hidden": True,
        "text": "Show Debug",
        "expanded": "false",
        "stored": {
            "sections": {"snapshot-controls": True},
            "debug_drawer_open": False,
        },
    }
    assert result["opened"] == {
        "closed": False,
        "hidden": False,
        "text": "Hide Debug",
        "expanded": "true",
        "stored": {
            "sections": {"snapshot-controls": True},
            "debug_drawer_open": True,
        },
    }


def test_static_console_ui_restores_valid_state_without_writing_on_initialize():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    console_ui = _console_ui_source(js)
    result = _run_node_json(
        "\n".join(
            [
                'const CONSOLE_UI_STORAGE_KEY = "agent-lab-console-ui-v1";',
                """
function makeElement({ dataset = {}, expanded = "false", hidden = false } = {}) {
  const listeners = {};
  const classes = new Set();
  return {
    dataset,
    hidden,
    textContent: "",
    attributes: { "aria-expanded": expanded },
    classList: {
      toggle(name, force) {
        if (force) classes.add(name); else classes.delete(name);
      },
      contains(name) { return classes.has(name); },
    },
    addEventListener(type, listener) { listeners[type] = listener; },
    getAttribute(name) { return this.attributes[name]; },
    setAttribute(name, value) { this.attributes[name] = String(value); },
  };
}
const agentButton = makeElement({
  dataset: { collapseTarget: "agent-config-controls" },
  expanded: "false",
});
const snapshotButton = makeElement({
  dataset: { collapseTarget: "snapshot-controls" },
  expanded: "false",
});
const agentBody = makeElement({ hidden: true });
const snapshotBody = makeElement({ hidden: true });
const drawerButton = makeElement({ expanded: "true" });
drawerButton.textContent = "Hide Debug";
const drawer = makeElement();
const layout = makeElement();
const elements = {
  "#agent-config-controls": agentBody,
  "#snapshot-controls": snapshotBody,
  "#toggle-debug-drawer": drawerButton,
  "#debug-drawer": drawer,
  "#console-layout": layout,
};
const document = {
  querySelector(selector) { return elements[selector] || null; },
  querySelectorAll(selector) {
    return selector === ".control-section-toggle"
      ? [agentButton, snapshotButton]
      : [];
  },
};
let writes = 0;
globalThis.localStorage = {
  getItem() {
    return JSON.stringify({
      debug_drawer_open: false,
      sections: { "agent-config-controls": true },
    });
  },
  setItem() { writes += 1; },
};
""",
                console_ui,
                """
initializeConsoleUi();
console.log(JSON.stringify({
  agentExpanded: agentButton.getAttribute("aria-expanded"),
  agentHidden: agentBody.hidden,
  snapshotExpanded: snapshotButton.getAttribute("aria-expanded"),
  snapshotHidden: snapshotBody.hidden,
  drawerExpanded: drawerButton.getAttribute("aria-expanded"),
  drawerHidden: drawer.hidden,
  drawerClosed: layout.classList.contains("debug-drawer-closed"),
  writes,
}));
""",
            ]
        )
    )

    assert result == {
        "agentExpanded": "true",
        "agentHidden": False,
        "snapshotExpanded": "false",
        "snapshotHidden": True,
        "drawerExpanded": "false",
        "drawerHidden": True,
        "drawerClosed": True,
        "writes": 0,
    }


def test_static_console_ui_storage_failures_fall_back_without_throwing():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    console_ui = _console_ui_source(js)
    result = _run_node_json(
        "\n".join(
            [
                'const CONSOLE_UI_STORAGE_KEY = "agent-lab-console-ui-v1";',
                """
function makeElement({ dataset = {}, expanded = "false", hidden = false } = {}) {
  return {
    dataset,
    hidden,
    textContent: "",
    attributes: { "aria-expanded": expanded },
    classList: { toggle() {} },
    addEventListener() {},
    getAttribute(name) { return this.attributes[name]; },
    setAttribute(name, value) { this.attributes[name] = String(value); },
  };
}
const sectionButton = makeElement({
  dataset: { collapseTarget: "snapshot-controls" },
  expanded: "false",
});
const sectionBody = makeElement({ hidden: true });
const drawerButton = makeElement({ expanded: "true" });
const drawer = makeElement();
const layout = makeElement();
const elements = {
  "#snapshot-controls": sectionBody,
  "#toggle-debug-drawer": drawerButton,
  "#debug-drawer": drawer,
  "#console-layout": layout,
};
const document = {
  querySelector(selector) { return elements[selector] || null; },
  querySelectorAll(selector) {
    return selector === ".control-section-toggle" ? [sectionButton] : [];
  },
};
globalThis.localStorage = {
  getItem() { return "{malformed"; },
  setItem() { throw new Error("storage unavailable"); },
};
""",
                console_ui,
                """
const malformed = readConsoleUiState();
writeConsoleUiState({ debug_drawer_open: true });
initializeConsoleUi();
const malformedDefaults = {
  sectionExpanded: sectionButton.getAttribute("aria-expanded"),
  sectionHidden: sectionBody.hidden,
  drawerExpanded: drawerButton.getAttribute("aria-expanded"),
  drawerHidden: drawer.hidden,
};
Object.defineProperty(globalThis, "localStorage", {
  configurable: true,
  get() { throw new Error("storage unavailable"); },
});
const unavailable = readConsoleUiState();
writeConsoleUiState({ sections: {} });
initializeConsoleUi();
console.log(JSON.stringify({ malformed, malformedDefaults, unavailable }));
""",
            ]
        )
    )

    assert result == {
        "malformed": {},
        "malformedDefaults": {
            "sectionExpanded": "false",
            "sectionHidden": True,
            "drawerExpanded": "true",
            "drawerHidden": False,
        },
        "unavailable": {},
    }


def test_static_topbar_status_updates_for_session_and_tool_mode():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    console_ui = _console_ui_source(js)
    result = _run_node_json(
        "\n".join(
            [
                'const CONSOLE_UI_STORAGE_KEY = "agent-lab-console-ui-v1";',
                """
const sessionLabel = { textContent: "" };
const modeLabel = { textContent: "" };
const toolInvocationMode = { value: "dual_agent" };
const document = {
  querySelector(selector) {
    return {
      "#current-session-label": sessionLabel,
      "#current-mode-label": modeLabel,
    }[selector] || null;
  },
};
let currentSessionId = null;
""",
                console_ui,
                """
updateTopbarStatus();
const initial = { session: sessionLabel.textContent, mode: modeLabel.textContent };
currentSessionId = "1234567890abcdef";
toolInvocationMode.value = "chat_agent_tools";
updateTopbarStatus();
console.log(JSON.stringify({
  initial,
  updated: { session: sessionLabel.textContent, mode: modeLabel.textContent },
}));
""",
            ]
        )
    )

    assert result == {
        "initial": {"session": "No session", "mode": "Dual Agent"},
        "updated": {"session": "Session 12345678", "mode": "ChatAgent Tools"},
    }


def test_static_console_ui_bootstrap_and_session_paths_refresh_topbar_status():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    refresh_tool_mode = js[
        js.index("function refreshToolModeState(") :
        js.index("function updateToolModeHelp(")
    ]

    assert "initializeConsoleUi();" in js[: js.index("function openChatConfig(")]
    assert "updateTopbarStatus();" in refresh_tool_mode
    assert "currentSessionId = session.id;\n    updateTopbarStatus();" in js
    assert "currentSessionId = sessionId;\n  updateTopbarStatus();" in js
    assert "currentSessionId = message.session_id;\n      updateTopbarStatus();" in js


def test_static_presentation_state_is_excluded_from_request_and_snapshot():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    build_snapshot = js[
        js.index("function buildWorkspaceSnapshot()") :
        js.index("function restoreWorkspaceSnapshot(")
    ]
    build_request = js[
        js.index("function buildRequest(") :
        js.index("function selectedTools(")
    ]

    for source in [build_snapshot, build_request]:
        assert "CONSOLE_UI_STORAGE_KEY" not in source
        assert "readConsoleUiState" not in source
        assert "writeConsoleUiState" not in source
        assert "debug_drawer_open" not in source


def test_static_professional_console_defines_visual_and_responsive_contracts():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    css = Path("src/agent_lab/presentation/static/styles.css").read_text()

    assert '/static/styles.css?v=20260714-professional-console' in html
    assert '/static/app.js?v=20260714-professional-console' in html
    for token in [
        "--canvas:",
        "--surface:",
        "--surface-subtle:",
        "--border:",
        "--text:",
        "--muted:",
        "--accent:",
        "--accent-strong:",
        "--danger:",
        "--radius:",
        "--shadow:",
    ]:
        assert token in _css_rule(css, ":root")

    assert "button-primary" in html
    assert "button-secondary" in html
    assert "button-danger" in html
    assert ".button-primary" in css
    assert ".button-secondary" in css
    assert ".button-danger" in css
    assert ".status-badge" in css
    assert ":focus-visible" in css
    assert ".tool-card:has(input:checked)" in css
    assert "display: none" in _css_rule(css, "[hidden]")
    assert "@media (max-width: 900px)" in css
    assert "\nsection + section {" not in css


def test_static_chat_agent_prompt_is_a_single_sortable_list_with_event_instructions():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    css = Path("src/agent_lab/presentation/static/styles.css").read_text()

    assert 'id="chat-system-prompt"' not in html
    assert 'id="chat-event-prompt-preview"' not in html
    assert 'class="prompt-list stack"' in html
    assert 'class="prompt-item"' in html
    assert 'data-prompt-role="system"' in html
    assert "prompt-type" in html
    assert "function buildAvailableEventsPrompt(" in js
    assert "Available events:" in js
    assert "function updateEventPromptItem(" in js
    assert "function createSystemPrompt(" in js
    assert "function movePromptItem(" in js
    assert 'draggable = true' in js
    assert "dataset.promptRole" in js
    assert "prompt-title" in js
    assert "delete-system-prompt" in js
    assert 'prompt_kind: "event_agent_system"' in js
    assert "system_prompts: collectSystemPrompts()" in js
    assert 'system_prompt: ""' in js
    assert ".prompt-item" in css
    assert ".prompt-item.dragging" in css
    assert ".prompt-meta" in css
    assert ".prompt-type" in css
    assert ".prompt-actions button" in css
    assert "#open-prompt-config" not in js
    assert "promptDialog" not in js


def test_static_event_prompt_refresh_uses_unified_tool_mode_state_path():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    render_tools = js[
        js.index("function renderTools(") :
        js.index("function formatRequiredParameters(")
    ]
    set_all_tools = js[
        js.index("function setAllTools(") :
        js.index("function updateToolCount(")
    ]
    restore_snapshot = js[
        js.index("function restoreWorkspaceSnapshot(") :
        js.index("function updateToolModeHelp(")
    ]
    apply_selected_tools = js[
        js.index("function applySelectedTools(") :
        js.index("function initializePromptList(")
    ]
    initialize_prompt_list = js[
        js.index("function initializePromptList(") :
        js.index("function buildPromptItems(")
    ]
    refresh_tool_mode = js[
        js.index("function refreshToolModeState(") :
        js.index("function updateToolModeHelp(")
    ]
    remove_generated_prompts = js[
        js.index("function removeGeneratedEventPromptItems(") :
        js.index("function ensureEventPromptItem(")
    ]

    assert (
        'toolInvocationMode.addEventListener("change", refreshToolModeState);'
        in js
    )
    assert "refreshToolModeState();" in render_tools
    assert "updateEventPromptItem(" not in render_tools
    assert "refreshToolModeState();" in set_all_tools
    assert "updateEventPromptItem(" not in set_all_tools
    assert "refreshToolModeState();" in restore_snapshot
    assert "refreshToolModeState();" in apply_selected_tools
    assert "refreshToolModeState();" in initialize_prompt_list
    assert "removeGeneratedEventPromptItems();" in refresh_tool_mode
    assert "if (!isChatAgentTools)" in refresh_tool_mode
    assert "updateEventPromptItem(availableTools);" in refresh_tool_mode
    assert 'querySelectorAll(".prompt-item")' in remove_generated_prompts
    assert "isGeneratedEventPromptItem(item)" in remove_generated_prompts
    assert "item.remove();" in remove_generated_prompts
    assert 'textarea.readOnly = promptItem.prompt_kind === "event_agent_system"' in js

    assert "let eventPromptInsertionIndex" in js
    capture_position = js[
        js.index("function captureEventPromptInsertionIndex(") :
        js.index("function removeGeneratedEventPromptItems(")
    ]
    ensure_event_prompt = js[
        js.index("function ensureEventPromptItem(") :
        js.index("function bindPromptItemControls(")
    ]
    restore_prompts = js[
        js.index("function restoreSystemPrompts(") :
        js.index("function restorePreMessages(")
    ]

    assert "return retainedPromptCount;" in capture_position
    assert "captureEventPromptInsertionIndex();" in refresh_tool_mode
    assert "capturedEventPromptIndex !== null" in refresh_tool_mode
    assert "eventPromptInsertionIndex = capturedEventPromptIndex;" in refresh_tool_mode
    assert "deriveEventPromptInsertionIndex(prompts);" in restore_prompts
    assert "eventPromptInsertionIndex = restoredEventPromptIndex;" in restore_prompts
    assert "const insertionTarget" in ensure_event_prompt
    assert "systemPrompts.insertBefore(item, insertionTarget);" in ensure_event_prompt
    assert "systemPrompts.append(item);" not in ensure_event_prompt


def test_static_event_prompt_order_survives_refresh_and_tool_mode_roundtrip():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    generated = "\n".join(
        [
            "You may request EventAgent work using this text protocol.",
            "Available events:",
            "- handoff_note: Handoff a note",
            "<agent_events>",
            "handoff_note",
            "</agent_events>",
            "Use exact event names only, one per line. Do not include parameters.",
        ]
    )

    assert "function captureEventPromptInsertionIndex(" in js
    refresh_tool_mode = js[
        js.index("function refreshToolModeState(") :
        js.index("function updateToolModeHelp(")
    ]
    generated_matcher = js[
        js.index("function isGeneratedEventPromptContent(") :
        js.index("function buildAvailableEventsPrompt(")
    ]
    event_prompt_functions = js[
        js.index("function updateEventPromptItem(") :
        js.index("function bindPromptItemControls(")
    ]
    result = _run_node_json(
        "\n".join(
            [
                refresh_tool_mode,
                generated_matcher,
                event_prompt_functions,
                """
const generated = [
  "You may request EventAgent work using this text protocol.",
  "Available events:",
  "- handoff_note: Handoff a note",
  "<agent_events>",
  "handoff_note",
  "</agent_events>",
  "Use exact event names only, one per line. Do not include parameters.",
].join("\\n");
const items = [];
function makePromptItem(promptKind, content) {
  const item = {
    dataset: { promptKind, promptRole: "system" },
    content,
    querySelector() {
      return {
        get value() { return item.content; },
        set value(value) { item.content = value; },
      };
    },
    remove() {
      const index = items.indexOf(item);
      if (index >= 0) items.splice(index, 1);
    },
  };
  return item;
}
items.push(
  makePromptItem("custom", "alpha"),
  makePromptItem("event_agent_system", "stale generated prompt"),
  makePromptItem("custom", "beta"),
  makePromptItem("custom", generated),
  makePromptItem("custom", "gamma"),
);
const systemPrompts = {
  querySelectorAll() { return [...items]; },
  querySelector() {
    return items.find((item) => item.dataset.promptKind === "event_agent_system") || null;
  },
  insertBefore(item, target) {
    const currentIndex = items.indexOf(item);
    if (currentIndex >= 0) items.splice(currentIndex, 1);
    const targetIndex = target ? items.indexOf(target) : items.length;
    items.splice(targetIndex < 0 ? items.length : targetIndex, 0, item);
  },
};
const toolInvocationMode = { value: "dual_agent" };
const eventAgentModelSettings = { querySelectorAll() { return []; } };
const availableTools = [{ name: "handoff_note" }];
let eventPromptInsertionIndex = 0;
function updateTopbarStatus() {}
function updateToolModeHelp() {}
function selectedTools() { return ["handoff_note"]; }
function buildAvailableEventsPrompt() { return generated; }
function createSystemPrompt() {
  return makePromptItem("event_agent_system", "");
}
function snapshot() {
  return items.map((item) => [item.dataset.promptKind, item.content]);
}

refreshToolModeState();
const afterRefresh = snapshot();
toolInvocationMode.value = "chat_agent_tools";
refreshToolModeState();
refreshToolModeState();
const whileHidden = snapshot();
const hiddenIndex = eventPromptInsertionIndex;
toolInvocationMode.value = "dual_agent";
refreshToolModeState();
const afterRoundtrip = snapshot();
console.log(JSON.stringify({ afterRefresh, whileHidden, hiddenIndex, afterRoundtrip }));
""",
            ]
        )
    )

    expected_visible = [
        ["custom", "alpha"],
        ["event_agent_system", generated],
        ["custom", "beta"],
        ["custom", "gamma"],
    ]
    assert result["afterRefresh"] == expected_visible
    assert result["whileHidden"] == [
        ["custom", "alpha"],
        ["custom", "beta"],
        ["custom", "gamma"],
    ]
    assert result["hiddenIndex"] == 1
    assert result["afterRoundtrip"] == expected_visible


def test_static_prompt_restore_derives_first_generated_event_prompt_position():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert "function deriveEventPromptInsertionIndex(" in js
    normalization = js[
        js.index("function normalizePromptItems(") :
        js.index("function buildAvailableEventsPrompt(")
    ]
    result = _run_node_json(
        "\n".join(
            [
                normalization,
                """
const generated = [
  "You may request EventAgent work using this text protocol.",
  "Available events:",
  "<agent_events>",
  "handoff_note",
  "</agent_events>",
  "Use exact event names only, one per line. Do not include parameters.",
].join("\\n");
const prompts = [
  { prompt_kind: "custom", content: "alpha" },
  generated,
  { prompt_kind: "custom", content: "beta" },
  { prompt_kind: "event_agent_system", content: generated },
  { prompt_kind: "custom", content: "gamma" },
];
console.log(JSON.stringify({
  insertionIndex: deriveEventPromptInsertionIndex(prompts),
  normalized: normalizePromptItems(prompts),
}));
""",
            ]
        )
    )

    assert result["insertionIndex"] == 1
    assert result["normalized"] == [
        {"prompt_kind": "custom", "role": "system", "content": "alpha"},
        {"prompt_kind": "custom", "role": "system", "content": "beta"},
        {"prompt_kind": "custom", "role": "system", "content": "gamma"},
    ]


def test_static_prompt_restore_discards_generated_event_prompt_derived_state():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    ordinary = "Keep this custom prompt with an <agent_events> example."
    exact_name_docs = (
        "Use exact event names only, one per line. Do not include parameters."
    )

    assert "function isGeneratedEventPromptContent(" in js
    normalization = js[
        js.index("function normalizePromptItems(") :
        js.index("function buildAvailableEventsPrompt(")
    ]
    result = _run_node_json(
        "\n".join(
            [
                normalization,
                """
const generated = [
  "You may request EventAgent work using this text protocol.",
  "Available events:",
  "- handoff_note: Handoff a note",
  "First write the user-facing reply normally. If events are needed, append this block after the visible reply:",
  "<agent_events>",
  "handoff_note",
  "</agent_events>",
  "Use exact event names only, one per line. Do not include parameters.",
].join("\\n");
const ordinary = "Keep this custom prompt with an <agent_events> example.";
const exactNameDocs = "Use exact event names only, one per line. Do not include parameters.";
console.log(JSON.stringify({
  matches: [
    isGeneratedEventPromptContent(generated),
    isGeneratedEventPromptContent(ordinary),
    isGeneratedEventPromptContent(exactNameDocs),
  ],
  normalized: normalizePromptItems([
    { prompt_kind: "event_agent_system", role: "system", content: generated },
    generated,
    ordinary,
    { prompt_kind: "custom", role: "system", content: exactNameDocs },
  ]),
  empty: normalizePromptItems([]),
}));
""",
            ]
        )
    )

    assert result["matches"] == [True, False, False]
    assert result["normalized"] == [
        {"prompt_kind": "custom", "role": "system", "content": ordinary},
        {"prompt_kind": "custom", "role": "system", "content": exact_name_docs},
    ]
    assert result["empty"] == [
        {"prompt_kind": "custom", "role": "system", "content": ""}
    ]


def test_static_event_prompt_payload_singleton_depends_on_tool_mode():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    managed_generated = "\n".join(
        [
            "You may request EventAgent work using this text protocol.",
            "Available events:",
            "- managed_event: Managed event",
            "<agent_events>",
            "managed_event",
            "</agent_events>",
            "Use exact event names only, one per line. Do not include parameters.",
        ]
    )
    prompt_collection = js[
        js.index("function buildPromptItems(") :
        js.index("function normalizePromptItems(")
    ]
    generated_matcher = js[
        js.index("function isGeneratedEventPromptContent(") :
        js.index("function buildAvailableEventsPrompt(")
    ]
    result = _run_node_json(
        "\n".join(
            [
                prompt_collection,
                generated_matcher,
                """
const impostorGenerated = [
  "You may request EventAgent work using this text protocol.",
  "Available events:",
  "- impostor_event: Impostor custom event",
  "<agent_events>",
  "impostor_event",
  "</agent_events>",
  "Use exact event names only, one per line. Do not include parameters.",
].join("\\n");
const managedGenerated = [
  "You may request EventAgent work using this text protocol.",
  "Available events:",
  "- managed_event: Managed event",
  "<agent_events>",
  "managed_event",
  "</agent_events>",
  "Use exact event names only, one per line. Do not include parameters.",
].join("\\n");
function promptItem(promptKind, content) {
  return {
    dataset: { promptKind, promptRole: "system" },
    querySelector() {
      return { value: content };
    },
  };
}
const systemPrompts = {
  querySelectorAll() {
    return [
      promptItem("custom", "ordinary custom"),
      promptItem("custom", impostorGenerated),
      promptItem("event_agent_system", managedGenerated),
    ];
  },
};
const toolInvocationMode = { value: "chat_agent_tools" };
const chatAgentTools = collectSystemPrompts();
toolInvocationMode.value = "dual_agent";
const dualAgent = collectSystemPrompts();
const generatedMatches = [
  isGeneratedEventPromptContent(impostorGenerated),
  isGeneratedEventPromptContent(managedGenerated),
];
console.log(JSON.stringify({ chatAgentTools, dualAgent, generatedMatches }));
""",
            ]
        )
    )

    assert result["generatedMatches"] == [True, True]
    assert result["chatAgentTools"] == ["ordinary custom"]
    assert result["dualAgent"] == ["ordinary custom", managed_generated]


def test_static_tool_mode_disables_only_event_agent_model_settings_group():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    model_settings_start = html.index('id="event-agent-model-settings"')
    shared_settings_start = html.index('id="max-event-loops"')
    model_settings_html = html[model_settings_start:shared_settings_start]
    refresh_tool_mode = js[
        js.index("function refreshToolModeState(") :
        js.index("function updateToolModeHelp(")
    ]

    for control_id in [
        "event-model",
        "event-temperature",
        "event-max-tokens",
        "event-system-prompt",
        "event-thinking-disabled",
        "event-enable-search",
        "event-forced-search",
    ]:
        assert f'id="{control_id}"' in model_settings_html

    for control_id in [
        "max-event-loops",
        "max-parallel-events",
        "batch-timeout-seconds",
    ]:
        assert f'id="{control_id}"' not in model_settings_html

    assert 'document.querySelector("#event-agent-model-settings")' in js
    assert '.querySelectorAll("input, textarea, select, button")' in refresh_tool_mode
    assert "control.disabled = isChatAgentTools;" in refresh_tool_mode


def test_static_app_script_is_versioned_and_click_binding_is_guarded():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert '<script src="/static/app.js?v=' in html
    assert "function bindClick(" in js
    assert 'bindClick("#open-chat-config-panel", openChatConfig)' in js
    assert 'bindClick("#open-event-config-panel", openEventConfig)' in js


def test_static_agent_config_controls_are_aligned_for_chat_and_event_agents():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert '<dialog id="chat-config-dialog"' in html
    assert '<dialog id="event-config-dialog"' in html
    assert html.count('id="open-chat-config-panel"') == 1
    assert html.count('id="open-event-config-panel"') == 1
    assert 'id="open-chat-config"' not in html
    assert 'id="open-event-config"' not in html
    assert '#open-chat-config"' not in js
    assert '#open-event-config"' not in js
    assert '#open-chat-config-panel"' in js
    assert '#open-event-config-panel"' in js
    assert 'id="close-chat-config"' in html
    assert 'id="close-event-config"' in html
    assert "chatConfigDialog.showModal()" in js
    assert "eventConfigDialog.showModal()" in js
    assert "chatConfigDialog.close()" in js
    assert "eventConfigDialog.close()" in js

    for control_id in [
        "model",
        "temperature",
        "max-tokens",
        "chat-thinking-disabled",
        "chat-enable-search",
        "chat-forced-search",
        "event-model",
        "event-temperature",
        "event-max-tokens",
        "event-system-prompt",
        "event-thinking-disabled",
        "event-enable-search",
        "event-forced-search",
    ]:
        assert f'id="{control_id}"' in html

    assert 'extra_body: buildExtraBody("chat")' in js
    assert 'extra_body: buildExtraBody("event")' in js
    assert 'system_prompt: ""' in js
    assert 'system_prompt: document.querySelector("#event-system-prompt")' in js


def test_static_tool_mode_and_batch_controls_explain_active_configuration():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert 'id="tool-invocation-mode"' in html
    assert '<option value="dual_agent" selected>' in html
    assert '<option value="chat_agent_tools">' in html
    assert 'id="tool-mode-help"' in html
    assert 'id="max-parallel-events"' in html
    assert 'value="4"' in html
    assert 'id="batch-timeout-seconds"' in html
    assert 'value="15"' in html
    assert "missing-argument fallback" in js
    assert "EventAgent model, temperature, max tokens, system prompt, and extra body are inactive" in js
    assert "enabled tools, loop budget, batch concurrency, and timeout remain active" in js
    assert "function updateToolModeHelp()" in js


def test_static_tool_mode_config_roundtrips_request_snapshot_and_session_restore():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    build_snapshot = js[
        js.index("function buildWorkspaceSnapshot()") :
        js.index("function restoreWorkspaceSnapshot(")
    ]
    restore_snapshot = js[
        js.index("function restoreWorkspaceSnapshot(") :
        js.index("function buildExtraBody(")
    ]
    build_request = js[
        js.index("function buildRequest(") :
        js.index("function selectedTools(")
    ]
    load_replay = js[
        js.index("async function loadSessionReplay(") :
        js.index("function setSessionStatus(")
    ]
    create_session = js[
        js.index("async function createSession()") :
        js.index("async function loadSelectedSession()")
    ]

    for source in [build_snapshot, build_request]:
        assert 'tool_invocation_mode: document.querySelector("#tool-invocation-mode").value' in source
        assert 'max_parallel_events: Number(document.querySelector("#max-parallel-events").value)' in source
        assert 'batch_timeout_seconds: Number(document.querySelector("#batch-timeout-seconds").value)' in source

    assert 'snapshot.tool_invocation_mode || "dual_agent"' in restore_snapshot
    assert 'eventAgent.max_parallel_events ?? 4' in restore_snapshot
    assert 'eventAgent.batch_timeout_seconds ?? 15' in restore_snapshot
    assert "refreshToolModeState();" in restore_snapshot
    assert '`/api/sessions/${sessionId}`' in load_replay
    assert "const [session, messages, audit, usage] = await Promise.all([" in load_replay
    assert "restoreWorkspaceSnapshot(session.config || {});" in load_replay
    assert "config: buildWorkspaceSnapshot()" in create_session


def test_static_tools_ui_shows_tool_metadata_and_selection_actions():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert 'id="tool-count"' in html
    assert 'id="select-all-tools"' in html
    assert 'id="clear-tools"' in html
    assert "tool-card" in js
    assert "tool-description" in js
    assert "tool-required" in js
    assert "function updateToolCount()" in js


def test_static_app_handles_audit_messages():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    css = Path("src/agent_lab/presentation/static/styles.css").read_text()

    assert 'message.type === "audit"' in js
    assert "appendAuditEntry(message)" in js
    assert 'appendLog("audit"' not in js
    assert "function renderAuditDetail(" in js
    assert "chat_agent_request" in js
    assert "event_agent_response" in js
    assert "chat_message_stream_started" in js
    assert "chat_message_stream_finished" in js
    assert ".audit-event" in css
    assert ".audit-detail" in css


def test_static_audit_replay_uses_event_stream_and_detail_panel():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    css = Path("src/agent_lab/presentation/static/styles.css").read_text()

    assert '<link rel="stylesheet" href="/static/styles.css?v=' in html
    assert "20260714-professional-console" in html
    assert 'id="audit-dialog"' in html
    assert 'id="open-audit-dialog"' in html
    assert 'id="close-audit-dialog"' in html
    assert 'id="audit-summary"' in html
    assert 'id="audit-replay" class="audit-stream"' in html
    assert 'id="audit-detail"' in html
    assert "auditDialog.showModal()" in js
    assert "auditDialog.close()" in js
    assert "auditEntryKey(entry)" in js
    assert "selectAuditEntry(key)" in js
    assert 'meta.className = "audit-detail-meta"' in js
    assert 'appendMetaField(meta, "Since User Message"' in js
    assert "auditRelativeTime(entry)" in js
    assert 'const list = document.createElement("ul")' in js
    assert "appendKeyValueSection(auditDetail, \"Model Parameters\"" in js
    assert "appendRepliesSection(auditDetail, details.replies)" in js
    assert "appendRawChunksSection(auditDetail, \"Raw LLM Chunks\"" in js
    assert "appendRawModelChunksSection(auditDetail, details.raw_model_chunks)" in js
    assert "Enabled This Round" in js
    assert "Configured Events" in js
    assert "details.configured_events" in js
    assert "enabled this round" in js
    assert ".audit-modal" in css
    assert ".audit-modal-body" in css
    assert ".audit-modal-body section + section" in css
    assert ".audit-detail-title" in css
    assert ".detail-tags code" in css
    assert ".audit-detail-meta span" not in css
    assert ".detail-tags span" not in css
    assert "display: block" in _css_rule(css, ".audit-detail")
    assert "display: grid" not in _css_rule(css, ".audit-detail")
    assert "display: block" in _css_rule(css, ".audit-detail-header")
    assert "display: grid" not in _css_rule(css, ".audit-detail-header")
    assert "display: inline-flex" in _css_rule(css, ".audit-detail-meta div")
    assert ".audit-marker::before" in css
    assert ".audit-event.is-selected" in css


def test_static_layout_keeps_controls_and_composer_fixed_with_scrollable_panes():
    css = Path("src/agent_lab/presentation/static/styles.css").read_text()

    layout_rule = _css_rule(css, ".layout")
    closed_rule = _css_rule(css, ".layout.debug-drawer-closed")
    workspace_rule = _css_rule(css, ".workspace")

    assert "height: calc(100vh - 64px)" in layout_rule
    assert (
        "grid-template-columns: minmax(260px, 320px) minmax(0, 1fr) "
        "minmax(300px, 380px)"
    ) in layout_rule
    assert "overflow: hidden" in layout_rule
    assert "grid-template-columns: minmax(260px, 320px) minmax(0, 1fr) 0" in closed_rule
    assert "overflow: auto" in _css_rule(css, ".controls")
    assert "grid-template-rows: minmax(0, 1fr) auto auto" in workspace_rule
    assert "overflow: hidden" in _css_rule(css, ".workspace-body")
    assert "overflow: auto" in _css_rule(css, ".messages")
    assert "overflow: auto" in _css_rule(css, ".debug-drawer")

    responsive_css = css[css.index("@media (max-width: 900px)") :]
    assert "height: auto" in responsive_css
    assert "overflow-x: hidden" in responsive_css
    assert ".debug-drawer[hidden]" in responsive_css


def test_static_app_shows_wait_state_and_immediate_user_echo():
    html = Path("src/agent_lab/presentation/static/index.html").read_text()
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    css = Path("src/agent_lab/presentation/static/styles.css").read_text()

    assert 'id="run-button"' in html
    assert 'const runButton = document.querySelector("#run-button")' in js
    assert 'appendLog("user", submittedMessage)' in js
    assert 'statusEl.textContent = "Waiting for first token"' in js
    assert "function startElapsedTimer()" in js
    assert "function stopElapsedTimer()" in js
    assert "function setRunState(" in js
    assert "if (firstTokenAt) {\n    hasBackendRoundStats = true;" in js
    assert "hasBackendRoundStats && firstTokenAt" in js
    assert ".user" in css


def test_static_app_reuses_websocket_for_session_turns():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert "function isSocketOpen()" in js
    assert "socket.send(JSON.stringify({ type: \"user_message\"" in js
    assert "session_id: currentSessionId || undefined" in js
    assert "message.session_id" in js
    assert 'message.type === "turn_completed"' in js
    assert 'message.type === "done"' in js


def test_static_app_loads_session_replay_and_usage():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    create_session = js[
        js.index("async function createSession()") :
        js.index("async function loadSelectedSession()")
    ]
    load_session = js[
        js.index("async function loadSelectedSession()") :
        js.index("async function loadSessionReplay(")
    ]

    assert "function loadSessions(" in js
    assert "function createSession(" in js
    assert "function setSessionStatus(" in js
    assert '"Refreshing sessions"' in js
    assert '"Creating session"' in js
    assert '"Session created"' in js
    assert 'sessionStatus.textContent = message' in js
    assert "function loadSessionReplay(" in js
    assert 'fetchJson("/api/sessions")' in js
    assert 'method: "POST"' in js
    assert '`/api/sessions/${sessionId}/messages`' in js
    assert '`/api/sessions/${sessionId}/audit`' in js
    assert '`/api/sessions/${sessionId}/usage`' in js
    assert "function renderAuditReplay(" in js
    assert "function renderSessionUsage(" in js
    assert "session.turn_wall_time_ms ?? session.elapsed_ms" in js
    assert "turn.turn_wall_time_ms ?? turn.elapsed_ms" in js
    assert "call.mode" in js
    assert "call.agent" in js
    assert "call.call_kind" in js
    assert "call.used_fallback" in js
    assert "call.tool_latency_ms" in js
    assert "sessionUsageSummary" in js
    assert "Finish current turn before switching sessions" in js
    assert "await loadSessionReplay(session.id, { preserveStatus: true });" in create_session
    assert "await loadSessionReplay(sessionId);" in load_session


def test_static_session_replay_ignores_superseded_success():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    result = _run_load_session_replay_scenario(
        """
currentSessionId = "old";
const oldLoad = loadSessionReplay("old");
currentSessionId = "new";
const newLoad = loadSessionReplay("new");
resolveReplay("new");
await newLoad;
resolveReplay("old");
await oldLoad;
console.log(JSON.stringify({
  ...mutations,
  title: sessionTitle.value,
  selected: sessionList.value,
}));
"""
    )

    assert result == {
        "workspace": ["new"],
        "transcript": ["new"],
        "audit": ["new"],
        "usage": ["new"],
        "status": ["Session loaded"],
        "title": "new title",
        "selected": "new",
    }
    assert "let sessionReplayLoadToken = 0;" in js


def test_static_session_replay_ignores_superseded_failure_status():
    result = _run_load_session_replay_scenario(
        """
currentSessionId = "old";
const oldLoad = loadSessionReplay("old");
currentSessionId = "new";
const newLoad = loadSessionReplay("new");
resolveReplay("new");
await newLoad;
pending.get("/api/sessions/old").reject(new Error("stale failure"));
await oldLoad;
console.log(JSON.stringify(mutations.status));
"""
    )

    assert result == ["Session loaded"]


def test_static_session_replay_rejects_stale_invocation_before_token_advance():
    result = _run_load_session_replay_scenario(
        """
currentSessionId = "new";
const newLoad = loadSessionReplay("new");
const tokenAfterNew = sessionReplayLoadToken;
const oldLoad = loadSessionReplay("old");
const oldResult = await Promise.race([
  oldLoad.then(() => "returned"),
  new Promise((resolve) => setTimeout(() => resolve("pending"), 0)),
]);
const tokenAfterOld = sessionReplayLoadToken;
const oldFetchStarted = pending.has("/api/sessions/old");
resolveReplay("new");
await newLoad;
console.log(JSON.stringify({
  oldResult,
  tokenAfterNew,
  tokenAfterOld,
  oldFetchStarted,
  mutations,
  title: sessionTitle.value,
  selected: sessionList.value,
}));
"""
    )

    assert result == {
        "oldResult": "returned",
        "tokenAfterNew": 1,
        "tokenAfterOld": 1,
        "oldFetchStarted": False,
        "mutations": {
            "workspace": ["new"],
            "transcript": ["new"],
            "audit": ["new"],
            "usage": ["new"],
            "status": ["Session loaded"],
        },
        "title": "new title",
        "selected": "new",
    }


def test_static_replay_hides_whitespace_assistant_tool_protocol_records():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    render_replay = js[
        js.index("function renderReplayMessages(") :
        js.index("function appendAuditEntry(")
    ]
    result = _run_node_json(
        "\n".join(
            [
                """
let activeAssistant = null;
const messagesEl = {
  children: [],
  scrollHeight: 0,
  scrollTop: 0,
  set textContent(value) {
    if (value === "") {
      this.children = [];
    }
  },
  append(item) {
    this.children.push(item);
  },
};
const document = {
  createElement() {
    return { className: "", textContent: "" };
  },
};
function appendLog(kind, content) {
  messagesEl.append({ className: `message ${kind}`, textContent: content });
}
""",
                render_replay,
                """
renderReplayMessages([
  { role: "assistant", content: "  \\n\\t", tool_calls: [{ id: "call-1" }] },
  { role: "assistant", content: "visible", tool_calls: [] },
]);
console.log(JSON.stringify(messagesEl.children.map((item) => item.textContent)));
""",
            ]
        )
    )

    assert result == ["visible"]


def test_static_session_switch_guard_uses_active_turn_not_open_socket():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()
    create_session = js[js.index("async function createSession()") : js.index("async function loadSelectedSession()")]
    load_session = js[js.index("async function loadSelectedSession()") : js.index("async function loadSessionReplay(")]

    assert "function hasActiveTurn()" in js
    assert "turnActive = isRunning" in js
    assert "if (hasActiveTurn())" in create_session
    assert "if (hasActiveTurn())" in load_session
    assert "if (isSocketOpen())" not in create_session
    assert "if (isSocketOpen())" not in load_session
    assert "closeSocket();" in create_session
    assert "closeSocket();" in load_session
    assert "const activeSocket = socket" in js
    assert "if (socket !== activeSocket)" in js


def test_static_workspace_snapshot_uses_stable_storage_hooks():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert "agent-lab.workspace-snapshots.v1" in js
    assert "agent-lab.prompt-sets.v1" in js
    assert "function saveWorkspaceSnapshot()" in js
    assert "function loadWorkspaceSnapshot()" in js
    assert "function deleteWorkspaceSnapshot()" in js
    assert "function buildWorkspaceSnapshot()" in js
    assert "prompt_items: buildPromptItems()" in js
    assert "enabled_tools: selectedTools()" in js
    assert "extra_body: buildExtraBody(\"chat\")" in js
    assert "extra_body: buildExtraBody(\"event\")" in js


def test_static_app_handles_backend_round_stats_messages():
    js = Path("src/agent_lab/presentation/static/app.js").read_text()

    assert 'message.type === "round_stats"' in js
    assert "function updateRoundStats(" in js
    assert "#stat-ttft" in js
    assert "#stat-elapsed" in js


@pytest.mark.asyncio
async def test_openai_chat_client_omits_stream_options_when_usage_disabled():
    captured_payloads: list[dict] = []

    def handler(request: httpx.Request) -> httpx.Response:
        captured_payloads.append(json.loads(request.content))
        return httpx.Response(
            200,
            content=b'data: {"choices":[{"delta":{"content":"hi"},"finish_reason":null}]}\n\n',
        )

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            include_usage=False,
            http_client=http_client,
        )
        [
            item
            async for item in client.stream_chat(
                messages=[ChatMessage(role="user", content="hi")],
                tools=[],
                params=AgentParams(model="provider-default", temperature=0.3, max_tokens=50),
            )
        ]

    assert "stream_options" not in captured_payloads[0]


@pytest.mark.asyncio
async def test_openai_chat_client_merges_agent_extra_body_into_payload():
    captured_payloads: list[dict] = []

    def handler(request: httpx.Request) -> httpx.Response:
        captured_payloads.append(json.loads(request.content))
        return httpx.Response(
            200,
            content=b'data: {"choices":[{"delta":{"content":"hi"},"finish_reason":null}]}\n\n',
        )

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            http_client=http_client,
        )
        [
            item
            async for item in client.stream_chat(
                messages=[ChatMessage(role="user", content="hi")],
                tools=[],
                params=AgentParams(
                    model="provider-default",
                    temperature=0.3,
                    max_tokens=50,
                    extra_body={
                        "thinking": {"type": "disabled"},
                        "enable_search": False,
                        "search_options": {"forced_search": False},
                    },
                ),
            )
        ]

    assert captured_payloads[0]["thinking"] == {"type": "disabled"}
    assert captured_payloads[0]["enable_search"] is False
    assert captured_payloads[0]["search_options"] == {"forced_search": False}


@pytest.mark.asyncio
@pytest.mark.parametrize(
    ("reserved_key", "override_value"),
    [
        ("model", "overridden-model"),
        ("messages", []),
        ("tools", []),
        (
            "tool_choice",
            {"type": "function", "function": {"name": "other_tool"}},
        ),
        ("stream", False),
        ("stream_options", {"include_usage": False}),
        ("temperature", 1.0),
        ("max_tokens", 1),
    ],
)
async def test_openai_chat_client_rejects_reserved_extra_body_fields_before_network(
    reserved_key: str,
    override_value: object,
):
    network_called = False

    def handler(request: httpx.Request) -> httpx.Response:
        nonlocal network_called
        network_called = True
        return httpx.Response(200)

    async with httpx.AsyncClient(
        transport=httpx.MockTransport(handler),
        base_url="https://llm.test/v1",
    ) as http_client:
        client = OpenAICompatibleChatClient(
            api_key="",
            base_url="https://llm.test/v1",
            default_model="provider-default",
            request_timeout_seconds=5,
            http_client=http_client,
        )
        with pytest.raises(
            ValueError,
            match=f"extra_body contains reserved field: {reserved_key}",
        ):
            [
                item
                async for item in client.stream_chat(
                    messages=[ChatMessage(role="user", content="hi")],
                    tools=[],
                    params=AgentParams(extra_body={reserved_key: override_value}),
                )
            ]

    assert network_called is False