feat: enable full AI automation and safe startup updates

mcp_server/tools/__init__.py

@@ -26,4 +26,19 @@ __all__ = [
     "mobile_get_device_state",
     "mobile_inspect_trace",
     "mobile_diagnose",
+    "mobile_workspace",
+    "mobile_config",
+    "mobile_task_file",
+    "mobile_batch",
+    "mobile_search_history",
+    "mobile_export_report",
 ]
+
+from mcp_server.tools.automation import (
+    mobile_workspace,
+    mobile_config,
+    mobile_task_file,
+    mobile_batch,
+    mobile_search_history,
+    mobile_export_report,
+)

mcp_server/tools/automation.py

@@ -0,0 +1,279 @@
+"""AI workflow tools backed by the same daemon, queue and persistence as the Web UI."""
+
+import asyncio
+from pathlib import Path
+import re
+from typing import Any, Literal
+from uuid import uuid4
+
+import httpx
+
+from artemis.config.paths import ROOT_DIR
+from artemis.runtime.daemon_client import ensure_daemon_running
+from mcp_server.base import mcp
+
+EXPORT_DIR = ROOT_DIR / "artifacts" / "mcp_exports"
+MODEL_FIELDS = {
+    "base_url",
+    "model",
+    "timeout_seconds",
+    "max_retries",
+    "retry_interval_seconds",
+    "reasoning_effort",
+}
+CAPABILITY_FIELDS = {"video_enabled", "outputter_enabled", "step_summarizer_enabled"}
+
+
+def _id(value: str | None) -> str:
+    if not value or not re.fullmatch(r"[A-Za-z0-9_-]{1,100}", value):
+        raise ValueError("请提供有效的记录编号（字母、数字、下划线或连字符）")
+    return value
+
+
+async def _base_url() -> str:
+    ready, url = await asyncio.to_thread(ensure_daemon_running, timeout=10, wait_ready=True)
+    if not ready or not url:
+        raise ValueError(
+            "无法连接统一任务服务，请先调用 mobile_diagnose；批次功能需要 daemon，不能在独立模式运行"
+        )
+    return url
+
+
+def _check(response: httpx.Response) -> None:
+    if response.is_error:
+        try:
+            detail = response.json().get("detail")
+        except ValueError:
+            detail = None
+        # Validation payloads may include user input; do not echo raw bodies.
+        message = (
+            detail
+            if isinstance(detail, str)
+            else "参数校验失败或服务不可用，请调用 mobile_diagnose 并检查输入"
+        )
+        raise ValueError(f"HTTP {response.status_code}: {message}")
+
+
+async def _request(method: str, route: str, **kwargs) -> Any:
+    url = await _base_url()
+    async with httpx.AsyncClient(base_url=url, timeout=30, trust_env=False) as client:
+        try:
+            response = await client.request(method, route, **kwargs)
+        except httpx.HTTPError as exc:
+            raise ValueError(
+                "任务服务连接中断；写入结果可能已生效，请先查询状态。创建批次重试必须沿用同一 request_id"
+            ) from exc
+        _check(response)
+        return response.json()
+
+
+async def _download(route: str, filename: str) -> dict:
+    url = await _base_url()
+    EXPORT_DIR.mkdir(parents=True, exist_ok=True)
+    path = EXPORT_DIR / f"{uuid4().hex[:12]}-{filename}"
+    partial = path.with_suffix(path.suffix + ".part")
+    try:
+        async with httpx.AsyncClient(base_url=url, timeout=120, trust_env=False) as client:
+            async with client.stream("GET", route) as response:
+                if response.is_error:
+                    await response.aread()
+                    _check(response)
+                with partial.open("wb") as stream:
+                    async for chunk in response.aiter_bytes():
+                        stream.write(chunk)
+        partial.replace(path)
+    finally:
+        partial.unlink(missing_ok=True)
+    return {
+        "path": str(path.resolve()),
+        "filename": filename,
+        "size_bytes": path.stat().st_size,
+        "download_url": url + route,
+    }
+
+
+@mcp.tool()
+async def mobile_workspace() -> dict:
+    """List connected Android devices and the unified running/queued task state.
+
+    Select the user's device; if multiple devices exist and none is chosen, ask first.
+    This does not submit tasks. On connection/device errors use mobile_diagnose.
+    """
+    devices = await _request("GET", "/api/devices")
+    status = await _request("GET", "/api/status")
+    return {"devices": devices.get("devices", []), "queue": status}
+
+
+@mcp.tool()
+async def mobile_config(
+    section: Literal["model", "capabilities", "reports"] = "model",
+    changes: dict[str, Any] | None = None,
+    persist_defaults: bool = False,
+) -> dict:
+    """Read configuration or explicitly save non-secret defaults for FUTURE tasks.
+
+    Only set persist_defaults=true when the user requests saved/global configuration.
+    model changes: model (preset or custom), base_url, reasoning_effort
+    (default/none/minimal/low/medium/high/xhigh), timeout_seconds, max_retries,
+    retry_interval_seconds. 'default' omits reasoning; unsupported values fail visibly.
+    capabilities: video_enabled (recording + analysis), outputter_enabled,
+    step_summarizer_enabled. reports: formats=['docx','xlsx'] (at least one).
+    API keys are never accepted or returned: use the Web UI or .env for credentials.
+    Saving settings does not change an already running task. Flash/Pro are execution
+    profiles, not model names. Model retries do not repeat device test cases.
+    """
+    route, allowed = {
+        "model": ("/api/system/model-settings", MODEL_FIELDS),
+        "capabilities": ("/api/system/capability-settings", CAPABILITY_FIELDS),
+        "reports": ("/api/automation/preferences", {"formats"}),
+    }[section]
+    if changes is not None:
+        if not persist_defaults:
+            raise ValueError("保存全局配置需要 persist_defaults=true，且应来自用户的明确要求")
+        if set(changes) - allowed:
+            raise ValueError("包含不支持的配置字段；密钥请通过网页或 .env 设置")
+    current = await _request("GET", route)
+    if changes is None:
+        return current
+    payload = {key: value for key, value in current.items() if key in allowed}
+    payload.update(changes)
+    return await _request("PUT" if section == "reports" else "POST", route, json=payload)
+
+
+@mcp.tool()
+async def mobile_task_file(
+    action: Literal["template", "preview"],
+    path: str | None = None,
+) -> dict:
+    """Download the standard XLSX task template, or validate a local XLSX without executing it.
+
+    preview requires an absolute .xlsx path on the MCP host. Returns tasks, row errors,
+    and valid. Correct all errors, strip preview-only 'row', then submit tasks via
+    mobile_batch(create). Creating XLSX is optional: AI can submit JSON cases directly.
+    """
+    if action == "template":
+        return await _download("/api/automation/template", "artemis-tasks.xlsx")
+    source = Path(path or "")
+    if not source.is_absolute() or source.suffix.lower() != ".xlsx" or not source.is_file():
+        raise ValueError("请提供本机现有 .xlsx 文件的绝对路径")
+    from apps.admin_console.services.task_import import MAX_BYTES
+
+    if source.stat().st_size > MAX_BYTES:
+        raise ValueError("文件超过 Excel 导入大小限制")
+    with source.open("rb") as stream:
+        return await _request(
+            "POST",
+            "/api/automation/import/preview",
+            files={
+                "file": (
+                    source.name,
+                    stream,
+                    "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+                )
+            },
+        )
+
+
+@mcp.tool()
+async def mobile_batch(
+    action: Literal["create", "list", "get", "lookup", "start", "pause", "resume", "retry_failed"],
+    batch_id: str | None = None,
+    request_id: str | None = None,
+    batch: dict[str, Any] | None = None,
+) -> Any:
+    """Create, execute and manage durable batches on the unified device queue.
+
+    create requires a unique request_id and batch={name,device_serial,cases:[{name,goal,
+    expected?,app?,preconditions?,parameters?:{key:primitive},tags?}],profile?:flash|pro,
+    formats?:[docx,xlsx],stop_on_failure?:false,timeout_seconds?:1800}. Required device
+    must be user-selected. Creation validates and saves a DRAFT; start executes it.
+    Reuse the same request_id and exact payload after a connection failure; never invent
+    a new ID for the same submission. lookup(request_id) resolves an uncertain outcome.
+    get returns per-case status. pause stops future dispatch, not the active task;
+    use mobile_manage_task(stop) to stop that task. resume never repeats finished cases.
+    retry_failed creates a new draft only after an explicit rerun request; it is not
+    transport-idempotent, so after an uncertain response use list instead of retrying.
+    Restarted batches remain paused. No automatic failed-case retries or action replay.
+    """
+    if action == "create":
+        rid = _id(request_id)
+        if not batch:
+            raise ValueError("create 需要 batch 内容")
+        return await _request("POST", "/api/automation/batches", json={**batch, "request_id": rid})
+    if action == "list":
+        return await _request("GET", "/api/automation/batches")
+    if action == "lookup":
+        return await _request("GET", f"/api/automation/batches/by-request/{_id(request_id)}")
+    route = f"/api/automation/batches/{_id(batch_id)}"
+    if action == "get":
+        return await _request("GET", route)
+    if action in {"start", "resume"}:
+        current = await _request("GET", route)
+        if current["status"] == "running":
+            return current
+    return await _request(
+        "POST", route + "/" + ("resume" if action == "start" else action), json={}
+    )
+
+
+@mcp.tool()
+async def mobile_search_history(
+    query: str = "",
+    status: str = "",
+    start: float | None = None,
+    end: float | None = None,
+    page: int = 1,
+    size: int = 30,
+) -> dict:
+    """Search ALL persisted terminal tasks by name, goal, ID or tags, with pagination.
+
+    start/end are Unix seconds, inclusive/exclusive; convert the user's local dates
+    before calling. Status: completed/success/failed/cancelled/interrupted/skipped.
+    Return items,total,page,size; inspect evidence/report before claiming a test passed.
+    """
+    params = {"q": query, "status": status, "page": page, "size": size}
+    if start is not None:
+        params["start"] = start
+    if end is not None:
+        params["end"] = end
+    return await _request("GET", "/api/automation/history", params=params)
+
+
+@mcp.tool()
+async def mobile_export_report(
+    session_id: str | None = None,
+    batch_id: str | None = None,
+    formats: list[Literal["docx", "xlsx"]] | None = None,
+) -> dict:
+    """Export a task's Word/Excel reports or a batch ZIP to local files on the MCP host.
+
+    Provide exactly one of session_id/batch_id. Task formats omitted uses saved defaults;
+    batch ZIP uses its frozen formats and includes summary + individual case reports.
+    Returns absolute paths, download URLs and sizes; task results also include structured
+    verdict/evidence. Running exports are snapshots. 'completed' is not proof of PASS;
+    report 'unverified' when checks/evidence are missing. Never fabricate test results.
+    """
+    if bool(session_id) == bool(batch_id):
+        raise ValueError("请且仅请提供 session_id 或 batch_id")
+    if batch_id:
+        bid = _id(batch_id)
+        return {
+            "batch_id": bid,
+            "files": [
+                await _download(f"/api/automation/batches/{bid}/archive", f"batch-{bid}.zip")
+            ],
+        }
+    sid = _id(session_id)
+    chosen = (
+        formats
+        if formats is not None
+        else (await _request("GET", "/api/automation/preferences"))["formats"]
+    )
+    if not chosen or len(chosen) > 2 or any(fmt not in {"docx", "xlsx"} for fmt in chosen):
+        raise ValueError("formats 至少选择 docx 或 xlsx，最多两项")
+    report = await _request("GET", f"/api/automation/reports/{sid}")
+    files = [
+        await _download(f"/api/automation/reports/{sid}/{fmt}", f"task-{sid}.{fmt}")
+        for fmt in dict.fromkeys(chosen)
+    ]
+    return {"report": report, "files": files}

mcp_server/tools/inspect_trace.py

@@ -78,7 +78,7 @@ async def mobile_inspect_trace(
       when the action failed/was intercepted, else null), and
       `action_overlay_screenshot` (the action visually marked — e.g. red circle
       for taps — key for verifying the agent tapped the right element).
-    - **'view_step_details'**: Full replay of one step exactly as the agent's
+    - **'view_step_details'**: Stored details of one step exactly as the agent's
       own context showed it: what the screen showed, its reasoning, every tool
       call it made (name, arguments, result — e.g. what the explorer or OCR
       reported), the planned action, any safety-net interception and the