aps-agent/server/aps_domain/project_analyze.py

# ============================================================
# 项目深度分析（moduleId: domain-project-analyze, 可重生 ✅）
# Pi 已结构化的工程目录 Excel/SQL → 写入当前项目世界 → 明细表 + 齐备度 + 知识库补洞 plan
# ============================================================
from __future__ import annotations

import copy
import json
import os
import re
from typing import Any

from server.aps_domain.readiness import check_readiness

World = dict[str, Any]

_TABLE_EXT_RE = re.compile(r"\.(xlsx|xlsm|csv|txt)$", re.I)
_INPUT_EXT_RE = re.compile(r"\.(xlsx|xlsm|csv|txt|sql)$", re.I)


def _file_sha256(path: str) -> str:
    """资料文件指纹：采用卡冻结来源文件，批准时校验同一份文件未变化。"""
    from server.agent_core.fallback_highrisk import file_sha256

    return str(file_sha256(path) or "")


def _project_ctx(session_id: str | None) -> tuple[dict | None, str]:
    try:
        from server.state.projects import get_project_store
        snap = get_project_store().snapshot(include_messages=False)
    except Exception:
        return None, ""
    sessions = {s["id"]: s for s in (snap.get("sessions") or []) if isinstance(s, dict)}
    projects = {p["id"]: p for p in (snap.get("projects") or []) if isinstance(p, dict)}
    sess = sessions.get(session_id or "") or {}
    proj = projects.get(sess.get("projectId") or "") if sess.get("projectId") else None
    work = (proj.get("workDir") or "").strip() if proj else ""
    return proj, work


def _project_ctx_readonly(session_id: str | None) -> tuple[dict | None, str]:
    """Resolve session -> project without ProjectStore seeding writes.

    解析顺序：SQLite 只读直查（真实部署主路径）→ 只读 JSON 快照 →
    注入/内存态 ProjectStore（单元测试与桌面内存态）。前两条不写任何东西；
    最后一条只用于直查拿不到项目的场景，避免只读分析在内存态下直接失效。
    """
    sid = str(session_id or "").strip()
    if not sid:
        return None, ""
    try:
        from server.auth.context import get_identity

        identity = get_identity()
    except Exception:
        return None, ""

    try:
        import sqlite3

        from sqlalchemy.engine import make_url

        from server.db.database import _database_url

        url = make_url(_database_url())
        if url.get_backend_name() == "sqlite" and url.database:
            db_path = os.path.abspath(str(url.database))
            if os.path.isfile(db_path):
                uri = "file:" + db_path.replace("\\", "/") + "?mode=ro"
                connection = sqlite3.connect(uri, uri=True)
                connection.row_factory = sqlite3.Row
                try:
                    row = connection.execute(
                        "SELECT p.id, p.name, p.work_dir "
                        "FROM aps_chat_sessions s "
                        "JOIN aps_workspace_projects p "
                        "ON p.id = s.project_id AND p.tenant_uuid = s.tenant_uuid "
                        "JOIN aps_project_members m "
                        "ON m.project_id = p.id AND m.tenant_uuid = p.tenant_uuid "
                        "WHERE s.id = ? AND s.tenant_uuid = ? AND s.scope = 'project' "
                        "AND s.deleted = 0 AND p.deleted = 0 "
                        "AND m.user_id = ? AND m.status = 'active' AND m.deleted = 0 "
                        "LIMIT 1",
                        (sid, identity.tenant_uuid, identity.user_id),
                    ).fetchone()
                finally:
                    connection.close()
                if row is not None:
                    project = {
                        "id": row["id"],
                        "name": row["name"],
                        "workDir": row["work_dir"],
                    }
                    return project, str(row["work_dir"] or "").strip()
    except Exception:
        pass

    legacy_path = os.environ.get("APS_PROJECTS_PATH") or ""
    try:
        if legacy_path and os.path.isfile(legacy_path):
            with open(legacy_path, "r", encoding="utf-8") as source:
                legacy = json.load(source)
            sessions = {
                str(row.get("id")): row
                for row in (legacy.get("sessions") or [])
                if isinstance(row, dict)
            }
            projects = {
                str(row.get("id")): row
                for row in (legacy.get("projects") or [])
                if isinstance(row, dict)
            }
            session = sessions.get(sid) or {}
            project = projects.get(str(session.get("projectId") or "")) or None
            if project:
                return project, str(project.get("workDir") or "").strip()
    except Exception:
        pass
    try:
        from server.state.projects import get_project_store

        snap = get_project_store().snapshot(include_messages=False)
        sessions = {str(row.get("id")): row for row in (snap.get("sessions") or [])
                    if isinstance(row, dict)}
        projects = {str(row.get("id")): row for row in (snap.get("projects") or [])
                    if isinstance(row, dict)}
        session = sessions.get(sid) or {}
        project = projects.get(str(session.get("projectId") or "")) or None
        if project:
            return project, str(project.get("workDir") or "").strip()
    except Exception:
        pass
    return None, ""


def _analysis_error(message: str, *, read_only: bool = False) -> dict[str, Any]:
    return {
        "ok": False,
        "error": message,
        "projectName": None,
        "workDir": "",
        "sources": [],
        "sourcePaths": [],
        "applied": {},
        "owner": None,
        "sqlPreview": None,
        "folder": None,
        "orders": [],
        "materials": [],
        "routings": [],
        "equipment": [],
        "kbActions": [],
        "plan": [],
        "readiness": {"summary": {}},
        "summary": {
            "orders": 0, "materials": 0, "routings": 0,
            "equipment": 0, "bom": 0, "ready": 0, "blocked": 0,
        },
        "canSchedule": False,
        "readOnly": read_only,
        "pendingAdoption": {
            "mode": "none", "filename": "", "sourcePaths": [],
            "batches": [], "counts": {},
        },
        "markdown": message,
        "knowledgeIngest": {},
    }


def _resolve_input_path(input_path: str | None) -> tuple[str, str | None]:
    """确定性校验 Pi 显式传入的文件/目录路径，不解析任何原始问句。"""
    raw = str(input_path or "").strip()
    if not raw:
        return "", None
    candidate = os.path.abspath(raw)
    if not os.path.exists(candidate):
        return "", f"指定的资料路径不存在：{raw}。请核对路径后重新提供 input_path。"
    if os.path.isfile(candidate):
        if not _INPUT_EXT_RE.search(candidate):
            return "", "指定文件类型不支持：仅接受 xlsx、xlsm、csv、txt 或 sql。"
        return candidate, None
    if os.path.isdir(candidate):
        return candidate, None
    return "", "指定路径既不是文件也不是目录，请重新提供有效的 input_path。"


def _direct_table_report(
    world: World,
    paths: list[str],
    *,
    project_name: str,
    work_dir: str,
) -> dict[str, Any]:
    """直接路径旁路：复用 preview_file，但不依赖工作区里的 workDir。"""
    from server.aps_domain.folder_pack import KIND_CN
    from server.aps_domain.importers import preview_file

    required = ("orders", "materials", "routing", "equipment")
    files: list[dict[str, Any]] = []
    batches: list[dict[str, Any]] = []
    kind_counts: dict[str, int] = {}
    for path in paths:
        name = os.path.basename(path)
        try:
            with open(path, "rb") as source:
                preview = preview_file(name, source.read(), world, soft=True)
            file_batches = preview.get("batches") or []
            ok_count = int(preview.get("totalOk") or 0)
            error_count = int(preview.get("totalErrors") or 0)
            primary_kind = next(
                (str(batch.get("kind")) for batch in file_batches if batch.get("okCount")),
                str((file_batches[0] if file_batches else {}).get("kind") or ""),
            )
            field_map = next(
                (list(batch.get("fieldMap") or []) for batch in file_batches if batch.get("fieldMap")),
                [],
            )
            errors: list[str] = []
            warnings: list[str] = []
            headers: list[str] = []
            samples: list[str] = []
            sheets: list[dict[str, Any]] = []
            for batch in file_batches:
                kind = str(batch.get("kind") or primary_kind or "materials")
                count = int(batch.get("okCount") or 0)
                if count:
                    kind_counts[kind] = kind_counts.get(kind, 0) + count
                errors.extend(str(item) for item in (batch.get("errors") or [])[:2])
                warnings.extend(str(item) for item in (batch.get("warnings") or [])[:2])
                if not headers:
                    headers = [str(item) for item in (batch.get("headersRaw") or []) if item][:8]
                for row in (batch.get("okRows") or [])[:1]:
                    if isinstance(row, dict):
                        samples.append("；".join(f"{k}={v}" for k, v in list(row.items())[:6]))
                sheets.append({
                    "sheet": batch.get("sheet"), "kind": kind,
                    "kindCn": KIND_CN.get(kind, kind), "okCount": count,
                    "errorCount": int(batch.get("errorCount") or 0),
                })
                if count:
                    batches.append({
                        "kind": kind, "sheet": batch.get("sheet") or name,
                        "okRows": batch.get("okRows") or [], "okCount": count,
                        "errorCount": int(batch.get("errorCount") or 0),
                        "errors": (batch.get("errors") or [])[:8],
                        "warnings": (batch.get("warnings") or [])[:5],
                        "fieldMap": batch.get("fieldMap") or [], "sourceFile": name,
                    })
            fmap_text = "、".join(
                f"{item.get('source')}→{item.get('target')}" for item in field_map
            )
            issue = (errors or warnings or ([f"字段：{fmap_text}"] if fmap_text else []))
            files.append({
                "name": name, "size": os.path.getsize(path),
                "status": "ok" if ok_count and not error_count else ("warn" if ok_count else "fail"),
                "kind": primary_kind, "kindCn": KIND_CN.get(primary_kind, primary_kind or "未识别"),
                "okCount": ok_count, "errorCount": error_count,
                "errors": errors, "warnings": warnings, "headers": headers,
                "samples": samples, "sheets": sheets, "fieldMap": field_map,
                "fieldMapText": fmap_text, "issue": issue[0] if issue else "",
            })
        except Exception as exc:  # noqa: BLE001
            files.append({
                "name": name, "size": os.path.getsize(path) if os.path.isfile(path) else 0,
                "status": "fail", "kind": "", "kindCn": "读取失败",
                "okCount": 0, "errorCount": 1, "errors": [str(exc)],
                "warnings": [], "headers": [], "samples": [], "sheets": [],
                "fieldMap": [], "fieldMapText": "", "issue": str(exc),
            })
    coverage = {kind: kind_counts.get(kind, 0) > 0 for kind in required}
    coverage_detail = [{
        "key": kind, "label": KIND_CN[kind], "ok": coverage[kind],
        "count": kind_counts.get(kind, 0), "fieldMap": [],
        "note": (f"读到 {kind_counts[kind]} 行" if coverage[kind] else "还没有有效数据"),
    } for kind in required]
    missing = [KIND_CN[kind] for kind in required if not coverage[kind]]
    return {
        "ok": True, "projectName": project_name, "workDir": work_dir,
        "files": files, "coverage": coverage, "coverageDetail": coverage_detail,
        "kindCounts": kind_counts, "missing": missing, "softMissing": [],
        "canSchedule": not missing, "batches": batches,
        "totalOk": sum(int(item.get("okCount") or 0) for item in files),
        "totalErrors": sum(int(item.get("errorCount") or 0) for item in files),
    }


def _preview_apply_template(
    world: World,
    template: dict[str, Any],
    product_code: str,
    product_name: str,
) -> None:
    """Apply a routing template to a detached preview world without persistence."""
    world["flexRoutings"] = [
        row for row in (world.get("flexRoutings") or [])
        if row.get("productCode") != product_code
    ]
    operations = {
        row.get("code"): row for row in (world.get("flexOperations") or [])
        if isinstance(row, dict)
    }
    equipment = world.get("flexEquipment") or []
    for step in template.get("steps") or []:
        code = step.get("operationCode")
        if not code:
            continue
        if code not in operations:
            world.setdefault("flexOperations", []).append({
                "code": code,
                "name": step.get("operationName") or code,
                "isBottleneck": False,
                "changeoverMin": step.get("changeoverMin") or 0,
            })
            operations[code] = world["flexOperations"][-1]
        if not any(code in (row.get("capabilities") or []) for row in equipment):
            for row in equipment:
                if row.get("status") == "RUNNING":
                    row.setdefault("capabilities", []).append(code)
                    row.setdefault("opStdTime", {})[code] = step.get("stdMinDefault") or 1.0
        world.setdefault("flexRoutings", []).append({
            "productCode": product_code,
            "productName": product_name or product_code,
            "seq": step.get("seq"),
            "operationCode": code,
            "requireMold": False,
            "stdTimePerUnit": step.get("stdMinDefault") or 1.0,
            "stdTimeSource": "模板",
        })
    template_ops = [row.get("operationCode") for row in (template.get("steps") or [])]
    teams = world.get("flexTeams") or []
    generic = next((team for team in teams if team.get("code") == "T-TPL"), None)
    if generic is None and teams is not None:
        generic = {
            "code": "T-TPL", "name": "模板工序班组（推断）",
            "memberCount": 4, "supportOps": [], "skillLevel": "L2",
        }
        world.setdefault("flexTeams", []).append(generic)
    if generic is not None:
        for code in template_ops:
            if code and code not in generic.setdefault("supportOps", []):
                generic["supportOps"].append(code)
    if not any(row.get("code") == product_code for row in (world.get("flexMaterials") or [])):
        world.setdefault("flexMaterials", []).append({
            "code": product_code, "name": product_name or product_code,
            "type": "FINISHED_PRODUCT", "unit": "件",
            "stock": 0, "inTransit": 0, "safetyStock": 0, "procurementLeadTime": 0,
        })


def _template_catalog_readonly() -> list[dict[str, Any]]:
    """Load the routing-template catalog without initializing or seeding storage."""
    try:
        from server.knowledge.routing_templates import _BUILTIN
    except Exception:
        return []

    catalog: list[dict[str, Any]] = []
    for item in _BUILTIN:
        steps = []
        for seq, code, name, equipment_type, low, high, setup, changeover in (item.get("steps") or []):
            steps.append({
                "seq": seq,
                "operationCode": code,
                "operationName": name,
                "equipmentType": equipment_type,
                "stdMinLow": low,
                "stdMinHigh": high,
                "stdMinDefault": round((float(low) + float(high)) / 2, 1),
                "setupMin": setup,
                "changeoverMin": changeover,
            })
        catalog.append({
            "code": item.get("code"),
            "name": item.get("name"),
            "industry": "machining",
            "category": item.get("category"),
            "keywords": str(item.get("keywords") or "").split(","),
            "description": item.get("description"),
            "assetId": "",
            "source": "builtin",
            "steps": steps,
        })

    # Overlay tenant/platform templates when a SQLite database already exists.
    # Opening it in mode=ro avoids creating a database or running schema seeding.
    try:
        import sqlite3

        from sqlalchemy.engine import make_url

        from server.auth.context import get_identity
        from server.db.database import _database_url

        url = make_url(_database_url())
        if url.get_backend_name() != "sqlite" or not url.database:
            return catalog
        db_path = os.path.abspath(str(url.database))
        if not os.path.isfile(db_path):
            return catalog

        tenants = ("platform", get_identity().tenant_uuid or "platform")
        placeholders = ",".join("?" for _ in tenants)
        uri = "file:" + db_path.replace("\\", "/") + "?mode=ro"
        connection = sqlite3.connect(uri, uri=True)
        connection.row_factory = sqlite3.Row
        try:
            rows = connection.execute(
                "SELECT id, tenant_uuid, code, name, industry, category, keywords, "
                "description, asset_id, source FROM routing_templates "
                f"WHERE deleted = 0 AND tenant_uuid IN ({placeholders}) ORDER BY id",
                tenants,
            ).fetchall()
            if not rows:
                return catalog

            template_ids = [int(row["id"]) for row in rows]
            id_placeholders = ",".join("?" for _ in template_ids)
            step_rows = connection.execute(
                "SELECT template_id, seq, operation_code, operation_name, equipment_type, "
                "std_min_low, std_min_high, setup_min, changeover_min "
                "FROM routing_template_steps "
                f"WHERE deleted = 0 AND template_id IN ({id_placeholders}) ORDER BY seq",
                template_ids,
            ).fetchall()
            steps_by_template: dict[int, list[dict[str, Any]]] = {}
            for step in step_rows:
                low = float(step["std_min_low"] or 0)
                high = float(step["std_min_high"] or 0)
                steps_by_template.setdefault(int(step["template_id"]), []).append({
                    "seq": step["seq"],
                    "operationCode": step["operation_code"],
                    "operationName": step["operation_name"],
                    "equipmentType": step["equipment_type"],
                    "stdMinLow": low,
                    "stdMinHigh": high,
                    "stdMinDefault": round((low + high) / 2, 1),
                    "setupMin": step["setup_min"],
                    "changeoverMin": step["changeover_min"],
                })

            merged = {str(item.get("code")): item for item in catalog}
            order = [str(item.get("code")) for item in catalog]
            for row in rows:
                code = str(row["code"] or "")
                if not code:
                    continue
                merged[code] = {
                    "code": code,
                    "name": row["name"],
                    "industry": row["industry"],
                    "category": row["category"],
                    "keywords": str(row["keywords"] or "").split(","),
                    "description": row["description"],
                    "assetId": row["asset_id"],
                    "source": row["source"],
                    "steps": steps_by_template.get(int(row["id"]), []),
                }
                if code not in order:
                    order.append(code)
            return [merged[code] for code in order]
        finally:
            connection.close()
    except Exception:
        return catalog


def _recommend_templates_readonly(text: str, top_k: int = 3) -> list[dict[str, Any]]:
    """Match templates against an immutable preview catalog without seeding."""
    text = (text or "").strip()
    scored: list[tuple[float, dict[str, Any]]] = []
    for template in _template_catalog_readonly():
        score = 0.0
        for keyword in template.get("keywords") or []:
            if keyword and keyword in text:
                score += 2.0
        if template.get("category") and template["category"] in text:
            score += 1.5
        for step in template.get("steps") or []:
            if step.get("operationName") and step["operationName"] in text:
                score += 0.5
        if score > 0:
            scored.append((score, template))
    scored.sort(key=lambda item: -item[0])
    return [dict(template, matchScore=score) for score, template in scored[:top_k]]


def _fill_from_knowledge(
    world: World,
    product_codes: list[str],
    *,
    apply: bool = True,
) -> list[dict[str, Any]]:
    """缺工艺时用行业模板/知识库补齐；只读模式仅预演模板效果。"""
    actions: list[dict[str, Any]] = []
    try:
        from server.knowledge.routing_templates import recommend_templates, apply_template_to_product
    except Exception:
        return actions
    routes = {r.get("productCode") for r in (world.get("flexRoutings") or [])}
    mats = {m.get("code") for m in (world.get("flexMaterials") or [])}
    for pc in product_codes:
        if not pc:
            continue
        if pc not in mats:
            world.setdefault("flexMaterials", []).append({
                "id": len(world.get("flexMaterials") or []) + 1,
                "code": pc, "name": pc, "type": "FINISHED_PRODUCT", "unit": "件",
                "stock": 0, "inTransit": 0, "safetyStock": 0, "procurementLeadTime": 0,
            })
            actions.append({"type": "material_stub", "productCode": pc,
                            "detail": f"物料主数据缺失，已建产品档案占位 {pc}"})
        if pc in routes:
            continue
        name = next((m.get("name") for m in (world.get("flexMaterials") or [])
                     if m.get("code") == pc), pc)
        try:
            if apply:
                recs = recommend_templates(f"{pc} {name}") or []
            else:
                recs = _recommend_templates_readonly(f"{pc} {name}") or []
        except Exception:
            recs = []
        if not recs:
            actions.append({"type": "need_routing", "productCode": pc,
                            "detail": f"产品 {pc} 无工艺，知识库也无匹配模板，需客户提供工艺路线"})
            continue
        tpl = recs[0]
        code = tpl.get("code") or tpl.get("templateCode")
        try:
            if apply:
                apply_template_to_product(world, code, pc)
            else:
                _preview_apply_template(world, tpl, pc, name)
            actions.append({
                "type": "template_applied", "productCode": pc,
                "template": code, "templateName": tpl.get("name"),
                "detail": (
                    f"产品 {pc} 缺工艺 → 已用知识库模板「{tpl.get('name') or code}」生成"
                    if apply else
                    f"产品 {pc} 缺工艺 → 只读预演知识库模板「{tpl.get('name') or code}」"
                ),
            })
            routes.add(pc)
        except Exception as exc:
            actions.append({"type": "template_failed", "productCode": pc,
                            "detail": f"模板 {code} 应用失败：{exc}"})
    return actions


def _preview_replace_flex(
    world: World,
    values: dict[str, Any],
    *,
    ensure_calendar: bool = False,
) -> dict[str, int]:
    """Load an already parsed flex snapshot into a detached preview world."""
    fields = (
        ("flexOrders", ("flexOrders", "orders")),
        ("flexMaterials", ("flexMaterials", "materials")),
        ("flexEquipment", ("flexEquipment", "equipment")),
        ("flexRoutings", ("flexRoutings", "routing")),
        ("flexBom", ("flexBom", "bom")),
        ("flexOperations", ("flexOperations", "operations")),
        ("flexZones", ("flexZones", "zones")),
        ("flexTeams", ("flexTeams", "teams")),
        ("flexPersonnel", ("flexPersonnel", "personnel")),
        ("flexFactoryResources", ("flexFactoryResources", "factoryResources")),
        ("flexPartners", ("flexPartners", "partners")),
        ("flexWip", ("flexWip", "wip")),
    )
    summary: dict[str, int] = {}
    for target, sources in fields:
        source = next((name for name in sources if name in values), None)
        if source is None:
            continue
        rows = copy.deepcopy(values.get(source) or [])
        world[target] = rows
        if rows:
            summary[target] = len(rows)
    if ensure_calendar and not world.get("flexCalendar"):
        world["flexCalendar"] = [{
            "shiftCode": "D", "startTime": "08:00", "endTime": "17:00",
            "breaks": [{"start": "12:00", "end": "13:00"}],
            "workdays": [1, 2, 3, 4, 5],
        }]
        summary["calendar"] = 1
    return summary


def _preview_apply_batches(world: World, batches: list[dict[str, Any]]) -> dict[str, Any]:
    """Apply reviewed import rows to a detached preview world for read-only analysis."""
    profile_batches = [batch for batch in batches if batch.get("sourceProfile")]
    if profile_batches:
        if len(profile_batches) != len(batches):
            raise ValueError("完整工作簿与其他导入批次不能混合确认，请分别导入")
        from server.importers.planning_workbook import apply_planning_batches

        return apply_planning_batches(world, profile_batches)

    summary: dict[str, int] = {}

    def _next_id(table: str) -> int:
        rows = world.get(table) or []
        return max(
            (row.get("id", 0) for row in rows if isinstance(row.get("id"), int)),
            default=0,
        ) + 1

    def _upsert(table: str, row: dict[str, Any], keys: tuple[str, ...]) -> None:
        rows = world.setdefault(table, [])
        existing = next(
            (
                item for item in rows
                if all(item.get(key) == row.get(key) for key in keys)
            ),
            None,
        )
        if existing is not None:
            existing.update(row)
        else:
            rows.append({"id": _next_id(table), **row})

    for batch in batches:
        kind = batch.get("kind")
        for row in batch.get("okRows") or []:
            if kind == "orders":
                _upsert("flexOrders", {
                    "orderNo": row.get("orderNo"),
                    "productCode": row.get("productCode"),
                    "quantity": row.get("quantity"),
                    "dueDate": row.get("deliveryDate"),
                    "priority": row.get("priority", 5),
                    "customerName": row.get("customerName") or "现场客户",
                    "customerLevel": row.get("customerLevel") or "B",
                    "isRush": bool(row.get("isRush")),
                    "status": "RELEASED",
                }, ("orderNo",))
                summary["flexOrders"] = summary.get("flexOrders", 0) + 1
            elif kind == "materials":
                _upsert("flexMaterials", dict(row), ("code",))
                summary["flexMaterials"] = summary.get("flexMaterials", 0) + 1
            elif kind == "calendar":
                _upsert("flexCalendar", dict(row), ("shiftCode",))
                summary["flexCalendar"] = summary.get("flexCalendar", 0) + 1
            elif kind == "equipment":
                _upsert("flexEquipment", dict(row), ("code",))
                summary["flexEquipment"] = summary.get("flexEquipment", 0) + 1
            elif kind == "molds":
                _upsert("flexMolds", dict(row), ("code",))
                summary["flexMolds"] = summary.get("flexMolds", 0) + 1
            elif kind == "operations":
                _upsert("flexOperations", dict(row), ("code",))
                summary["flexOperations"] = summary.get("flexOperations", 0) + 1
            elif kind == "zones":
                _upsert("flexZones", dict(row), ("code",))
                summary["flexZones"] = summary.get("flexZones", 0) + 1
            elif kind == "routing":
                _upsert("flexRoutings", dict(row), ("productCode", "seq"))
                summary["flexRoutings"] = summary.get("flexRoutings", 0) + 1
            elif kind == "bom":
                _upsert("flexBom", dict(row), ("productCode", "materialCode"))
                summary["flexBom"] = summary.get("flexBom", 0) + 1
    return {"summary": summary, "total": sum(summary.values())}


def analyze_project_deep(
    world: World,
    session_id: str | None = None,
    *,
    apply_sql: bool = True,
    query: str = "",
    input_path: str | None = None,
    next_id=None,
    apply: bool = True,
) -> dict[str, Any]:
    """深度分析显式资料路径或当前项目目录；query 仅为兼容旧调用的弃用参数。"""
    del query  # 原始自然语言不参与路径、实体或数据语义推断。
    try:
        from server.agent_core.progress import emit_thinking
    except Exception:
        def emit_thinking(*_a, **_k):  # type: ignore
            return None

    explicit_input, input_error = _resolve_input_path(input_path)
    if input_error:
        return _analysis_error(input_error, read_only=not apply)
    proj, work = _project_ctx(session_id) if apply else _project_ctx_readonly(session_id)
    if explicit_input:
        work = explicit_input if os.path.isdir(explicit_input) else os.path.dirname(explicit_input)
        if proj is None:
            proj = {
                "name": os.path.basename(work.rstrip("\\/")) or "当前项目",
                "workDir": work,
            }
    if not work or not os.path.isdir(work):
        return _analysis_error(
            "请明确要分析的资料路径：提供 input_path，或先选择带有工程目录的项目。"
            "不能从原始问句推断路径。",
            read_only=not apply,
        )
    analysis_world = world if apply else copy.deepcopy(world)
    sources: list[str] = []
    source_paths: list[str] = []
    sql_preview: dict[str, Any] | None = None
    applied: dict[str, int] = {}
    kb_actions: list[dict[str, Any]] = []
    mom_applied = False
    mom_path = ""

    def apply_folder_batches(report: dict[str, Any] | None) -> None:
        batches = (report or {}).get("batches") or []
        if not batches or next_id is None:
            return
        if apply:
            from server.aps_domain.importers import apply_import_commit

            result = apply_import_commit(analysis_world, next_id, batches)
        else:
            result = _preview_apply_batches(analysis_world, batches)
        # apply_import_commit returns {summary: {...}, total: N}; keep accepting
        # the historical flat counter shape for callers that still provide it.
        counts = (
            result.get("summary")
            if isinstance(result, dict) and isinstance(result.get("summary"), dict)
            else result
        ) or {}
        for key, value in counts.items():
            if isinstance(value, (int, float)) and not isinstance(value, bool):
                applied[key] = applied.get(key, 0) + int(value)

    emit_thinking("定位工程目录", (proj or {}).get("name") or "当前项目", pct=2)
    if work and os.path.isdir(work):
        names = ([os.path.basename(explicit_input)] if os.path.isfile(explicit_input)
                 else sorted(os.listdir(work)))
        sqls = [n for n in names if n.lower().endswith(".sql") and not n.startswith("~$")]
        xls = [n for n in names if _TABLE_EXT_RE.search(n) and not n.startswith("~$")]
        source_paths = [
            os.path.join(work, name) for name in (*sqls, *xls)
            if os.path.isfile(os.path.join(work, name))
        ]
        folder = None
        emit_thinking(
            "清点目录文件",
            f"SQL {len(sqls)} · Excel/CSV {len(xls)}",
            pct=5,
        )
        if sqls:
            from server.importers.sql_pack import preview_sql_file, apply_sql_pack_to_world
            # 优先最大的 sql（完整库）
            sqls_sorted = sorted(sqls, key=lambda n: os.path.getsize(os.path.join(work, n)), reverse=True)
            path = os.path.join(work, sqls_sorted[0])
            sources.append(f"SQL:{sqls_sorted[0]}")
            try:
                sql_preview = preview_sql_file(path)
                # Read-only analysis must still materialize SQL into the detached
                # preview world when callers disable both write switches.
                if sql_preview.get("flex") and (apply_sql or not apply):
                    emit_thinking(
                        "写入项目世界" if apply else "预演项目世界",
                        "订单 / 物料 / 工艺 / 设备",
                        pct=93,
                    )
                    # SQL 包是工程目录真相源，始终整包覆盖，避免旧错误工艺残留
                    if apply:
                        applied = apply_sql_pack_to_world(
                            analysis_world, sql_preview["flex"], replace=True)
                    else:
                        applied = _preview_replace_flex(
                            analysis_world,
                            sql_preview["flex"],
                            ensure_calendar=True,
                        )
                    emit_thinking(
                        "入库完成",
                        "、".join(f"{k}{v}" for k, v in applied.items()) or "无新增",
                        status="done", pct=95,
                    )
            except Exception as exc:
                sql_preview = {"error": str(exc), "filename": sqls_sorted[0]}
                emit_thinking("SQL 解析失败", str(exc), status="warn")
        elif xls:
            # 无 SQL 时：识别 MOM 主数据收集表并写入世界（否则深度分析永远是 0）
            try:
                from server.importers.mom_pack import (
                    apply_mom_pack_to_world,
                    is_mom_workbook,
                    mom_pack_is_adoptable,
                    mom_pack_row_count,
                    parse_mom_workbook,
                )
                mom_files = [
                    n for n in xls
                    if is_mom_workbook(os.path.join(work, n))
                ]
                # 文件名含 MOM/主数据 优先；否则对全部 xlsx 试探
                candidates = mom_files or xls
                for n in candidates:
                    path = os.path.join(work, n)
                    if not is_mom_workbook(path):
                        continue
                    pack = parse_mom_workbook(path)
                    if not mom_pack_is_adoptable(pack):
                        # 结构证据（≥2 张模型表）或主干行（物料/BOM/订单/工艺路线）
                        # 不足：采用即整表替换，这里不作为 MOM 来源，避免清空既有主数据。
                        emit_thinking("MOM 主数据表不足以采用", n, status="warn")
                        continue
                    sources.append(f"MOM:{n}")
                    emit_thinking("识别 MOM 主数据表", n, pct=90)
                    if apply:
                        applied = apply_mom_pack_to_world(analysis_world, pack, replace=True)
                    else:
                        applied = _preview_replace_flex(analysis_world, pack)
                    result = {"stats": pack.get("stats") or {}}
                    mom_applied = True
                    mom_path = path
                    emit_thinking(
                        "MOM 入库完成",
                        "、".join(f"{k}{v}" for k, v in applied.items()) or "无新增",
                        status="done", pct=95,
                    )
                    stats = result.get("stats") or {}
                    if stats.get("note"):
                        sources.append(str(stats["note"]))
                    break
            except Exception as exc:
                emit_thinking("MOM Excel 解析失败", str(exc), status="warn")
            if not mom_applied:
                sources.append(f"Excel×{len(xls)}")
                try:
                    emit_thinking("解析并写入 Excel/CSV", f"{len(xls)} 个表格", pct=96)
                    if explicit_input:
                        folder = _direct_table_report(
                            analysis_world,
                            [os.path.join(work, name) for name in xls],
                            project_name=(proj or {}).get("name") or "当前项目",
                            work_dir=work,
                        )
                    else:
                        from server.aps_domain.folder_pack import analyze_work_dir
                        folder = analyze_work_dir(analysis_world, session_id, apply_sql=apply)
                    apply_folder_batches(folder)
                except Exception as exc:
                    if folder is None:
                        folder = {
                            "ok": False,
                            "projectName": (proj or {}).get("name"),
                            "workDir": work,
                            "files": [], "coverage": {}, "missing": [],
                            "canSchedule": False, "batches": [],
                            "totalOk": 0, "totalErrors": 1,
                            "error": str(exc),
                        }
                    else:
                        folder["error"] = str(exc)
            else:
                try:
                    if explicit_input:
                        folder = _direct_table_report(
                            analysis_world,
                            [os.path.join(work, name) for name in xls],
                            project_name=(proj or {}).get("name") or "当前项目",
                            work_dir=work,
                        )
                    else:
                        from server.aps_domain.folder_pack import analyze_work_dir
                        folder = analyze_work_dir(analysis_world, session_id, apply_sql=apply)
                except Exception as exc:
                    if folder is None:
                        folder = {
                            "ok": False,
                            "projectName": (proj or {}).get("name"),
                            "workDir": work,
                            "files": [], "coverage": {}, "missing": [],
                            "canSchedule": False, "batches": [],
                            "totalOk": 0, "totalErrors": 1,
                            "error": str(exc),
                        }
                    else:
                        folder["error"] = str(exc)
        else:
            folder = None

        if sqls and xls:
            sources.append(f"Excel×{len(xls)}")
            try:
                emit_thinking("核对 Excel/CSV", f"{len(xls)} 个表格", pct=96)
                if explicit_input:
                    folder = _direct_table_report(
                        analysis_world,
                        [os.path.join(work, name) for name in xls],
                        project_name=(proj or {}).get("name") or "当前项目",
                        work_dir=work,
                    )
                else:
                    from server.aps_domain.folder_pack import analyze_work_dir
                    folder = analyze_work_dir(analysis_world, session_id, apply_sql=apply)
                apply_folder_batches(folder)
            except Exception as exc:
                if folder is None:
                    folder = {
                        "ok": False,
                        "projectName": (proj or {}).get("name"),
                        "workDir": work,
                        "files": [], "coverage": {}, "missing": [],
                        "canSchedule": False, "batches": [],
                        "totalOk": 0, "totalErrors": 1,
                        "error": str(exc),
                    }
                else:
                    folder["error"] = str(exc)
    else:
        folder = None
        emit_thinking("未挂工程目录", "仅分析当前项目世界里的数据", status="warn", pct=10)

    # 知识库补洞：缺工艺的产品
    emit_thinking("检查齐备度与知识库", "缺工艺时自动匹配行业模板", pct=97)
    active = [o for o in (analysis_world.get("flexOrders") or [])
              if (o.get("status") or "") not in ("DONE", "CANCELLED")]
    pcs = list({str(o.get("productCode") or "") for o in active})
    kb_actions = _fill_from_knowledge(analysis_world, pcs, apply=apply)
    if kb_actions:
        emit_thinking(
            "知识库补洞",
            f"{len(kb_actions)} 项（模板/占位/待客户补）",
            status="done", pct=98,
        )

    ready = check_readiness(analysis_world)
    emit_thinking("整理分析报告", f"订单 {len(active)} 张", status="done", pct=99)
    mats = analysis_world.get("flexMaterials") or []
    routes = analysis_world.get("flexRoutings") or []
    equip = [e for e in (analysis_world.get("flexEquipment") or []) if e.get("status") == "RUNNING"]
    bom = analysis_world.get("flexBom") or []

    order_rows = [{
        "orderNo": o.get("orderNo"),
        "productCode": o.get("productCode"),
        "productName": o.get("productName") or next(
            (m.get("name") for m in mats if m.get("code") == o.get("productCode")), ""),
        "quantity": o.get("quantity"),
        "dueDate": o.get("dueDate"),
        "customerName": o.get("customerName") or "",
        "ready": next((r["ready"] for r in ready["orders"] if r["orderNo"] == o.get("orderNo")), False),
    } for o in active[:80]]

    # 主数据抽样：订单涉及的产品优先
    need_codes = {r["productCode"] for r in order_rows if r.get("productCode")}
    mat_rows = []
    for m in mats:
        if need_codes and m.get("code") not in need_codes and len(mat_rows) >= 40:
            continue
        if m.get("code") in need_codes or len(mat_rows) < 40:
            mat_rows.append({
                "code": m.get("code"), "name": m.get("name"),
                "type": m.get("type"), "unit": m.get("unit"), "stock": m.get("stock"),
            })
        if len(mat_rows) >= 60:
            break

    route_rows = []
    for r in routes:
        if need_codes and r.get("productCode") not in need_codes and len(route_rows) >= 40:
            continue
        route_rows.append({
            "productCode": r.get("productCode"),
            "seq": r.get("seq"),
            "operationCode": r.get("operationCode"),
            "operationName": r.get("operationName") or r.get("operationCode"),
            "stdMin": r.get("stdTimePerUnit"),
        })
        if len(route_rows) >= 80:
            break

    equip_rows = [{
        "code": e.get("code"), "name": e.get("name"),
        "capabilities": "、".join((e.get("capabilities") or [])[:6]),
    } for e in equip[:40]]

    # 缺口 plan：先库后客户（同类缺口合并，避免刷屏）
    plan: list[str] = []
    need_rt = [a for a in kb_actions if a.get("type") == "need_routing"]
    applied_tpl = [a for a in kb_actions if a.get("type") == "template_applied"]
    other_kb = [a for a in kb_actions if a.get("type") not in ("need_routing", "template_applied")]
    if applied_tpl:
        plan.append(f"✓ 已用知识库模板补工艺 {len(applied_tpl)} 个产品")
        for a in applied_tpl[:5]:
            plan.append(f"  · {a.get('productCode')} ← {a.get('templateName') or a.get('template')}")
        if len(applied_tpl) > 5:
            plan.append(f"  · …另有 {len(applied_tpl) - 5} 个")
    if need_rt:
        samples = "、".join(str(a.get("productCode") or "") for a in need_rt[:8])
        more = f" 等共 {len(need_rt)} 个" if len(need_rt) > 8 else f"（{len(need_rt)} 个）"
        plan.append(
            f"△ SQL/知识库仍未挂上工艺的产品：{samples}{more}。"
            f"若现场库本有工艺，请重新「分析项目」（工艺按订单 craftl 关联，不是中间件料号）"
        )
    for a in other_kb[:6]:
        plan.append(f"· {a['detail']}")
    s = ready["summary"]
    if s.get("blocked"):
        plan.append(f"△ 仍有 {s['blocked']} 张单齐备度未过，见订单明细「是否可排」列")
    if not active:
        plan.append("△ 没有订单：请在工程目录放订单表/SQL，或说「新建订单」")
    if not equip:
        plan.append("△ 没有可用设备：请提供设备台账，或从 SQL 的 md_equipment 导入")
    if not plan:
        plan.append("✓ 主数据与知识库已齐，可直接说「给我排产」")

    proj_name = (proj or {}).get("name") or "当前项目"
    md_lines = [
        f"**项目深度分析 · {proj_name}**",
        "",
        f"数据来源：{'、'.join(sources) if sources else '当前项目世界（未挂工程目录文件）'}",
    ]
    # 绑定当前登录用户
    try:
        from server.knowledge.ingest import bind_world_to_identity, ingest_project_analyze_report
        owner_meta = bind_world_to_identity(analysis_world, sources=sources)
        md_lines.append(
            f"归属：{owner_meta.get('ownerFullname') or owner_meta.get('ownerUsername')} "
            f"（用户 {owner_meta.get('ownerUserId')} · 租户 {str(owner_meta.get('tenantUuid') or '')[:8]}…）"
        )
    except Exception:
        owner_meta = {}
        ingest_project_analyze_report = None  # type: ignore

    if apply and applied:
        md_lines.append(
            "已入库："
            + "、".join(f"{k} {v}" for k, v in applied.items())
        )
    if sql_preview and sql_preview.get("fieldMapText"):
        md_lines.append(f"SQL 包识别：{sql_preview['fieldMapText']}")
    md_lines += [
        "",
        f"订单 **{len(active)}** · 物料 **{len(mats)}** · 工艺步 **{len(routes)}** · "
        f"设备 **{len(equip)}** · BOM **{len(bom)}**",
        "",
        "### 订单明细（本项目）",
        "",
        "| 订单号 | 产品 | 名称 | 数量 | 交期 | 客户 | 可排 |",
        "| --- | --- | --- | ---: | --- | --- | --- |",
    ]
    for r in order_rows[:30]:
        md_lines.append(
            f"| {r['orderNo']} | {r['productCode']} | {str(r.get('productName') or '')[:16]} "
            f"| {r['quantity']} | {r['dueDate'] or '—'} | {str(r.get('customerName') or '')[:12]} "
            f"| {'✓' if r['ready'] else '×'} |"
        )
    if len(order_rows) > 30:
        md_lines.append(f"| … | 另有 {len(order_rows) - 30} 张 | | | | | |")

    md_lines += [
        "",
        "### 主数据（产品/物料抽样）",
        "",
        "| 编码 | 名称 | 类型 | 单位 | 库存 |",
        "| --- | --- | --- | --- | ---: |",
    ]
    for m in mat_rows[:25]:
        md_lines.append(
            f"| {m['code']} | {str(m.get('name') or '')[:20]} | {m.get('type')} "
            f"| {m.get('unit')} | {m.get('stock') or 0} |"
        )

    md_lines += ["", "### 补数 Plan（先知识库/数据库，再找客户）", ""]
    for i, p in enumerate(plan[:12], 1):
        md_lines.append(f"{i}. {p}")

    def _adoption_counts(values: dict[str, Any] | None) -> dict[str, int]:
        return {
            str(key): int(value)
            for key, value in (values or {}).items()
            if isinstance(value, (int, float))
            and not isinstance(value, bool)
            and int(value) > 0
        }

    folder_batches = [
        copy.deepcopy(batch)
        for batch in ((folder or {}).get("batches") or [])
        if batch.get("okRows")
    ]
    sql_flex = (sql_preview or {}).get("flex") or {}
    sql_available = bool(sql_flex.get("stats"))
    if mom_applied and mom_path:
        pending_adoption = {
            "mode": "mom",
            "filename": os.path.basename(mom_path),
            "sourcePaths": [mom_path],
            "batches": [],
            "counts": _adoption_counts(applied),
            # 出卡/执行两处都按这个口径复核，避免零业务行的 MOM 表升级成整表替换。
            "businessRows": mom_pack_row_count(pack),
            "adoptable": mom_pack_is_adoptable(pack),
        }
        if not apply:
            # 只读分析不出卡不入库：把来源路径与指纹冻结给采用卡，批准后按同一份文件采用。
            pending_adoption["applyPath"] = mom_path
            pending_adoption["applySha256"] = _file_sha256(mom_path)
    elif folder_batches:
        batch_files = list(dict.fromkeys(
            str(batch.get("sourceFile") or "")
            for batch in folder_batches
            if batch.get("sourceFile")
        ))
        filename = batch_files[0] if len(batch_files) == 1 else ""
        batch_file_set = set(batch_files)
        selected_paths = [
            path for path in source_paths
            if os.path.basename(path) in batch_file_set
        ]
        pending_adoption = {
            "mode": "folder",
            "filename": filename,
            "sourcePaths": selected_paths or list(source_paths),
            "batches": folder_batches,
            "counts": _adoption_counts(applied) or {
                kind: sum(
                    len(batch.get("okRows") or [])
                    for batch in folder_batches
                    if batch.get("kind") == kind
                )
                for kind in {str(batch.get("kind") or "") for batch in folder_batches}
                if kind
            },
        }
    elif sql_available:
        sql_name = str((sql_preview or {}).get("filename") or "")
        pending_adoption = {
            "mode": "sql",
            "filename": sql_name,
            "sourcePaths": [
                path for path in source_paths
                if sql_name and os.path.basename(path) == sql_name
            ] or list(source_paths),
            "batches": [],
            "counts": _adoption_counts(applied),
        }
        if not apply:
            # SQL 没有 okRows 批次：冻结已解析载荷，采用时整包写入而不是重读文件。
            payload = copy.deepcopy(sql_flex)
            if payload.get("stats"):
                pending_adoption["sqlPayload"] = payload
    else:
        pending_adoption = {
            "mode": "none", "filename": "", "sourcePaths": [],
            "batches": [], "counts": {},
        }

    if not apply and pending_adoption["mode"] != "none":
        md_lines.append("只读分析：资料尚未采用，确认后再写入项目数据。")

    result = {
        "ok": True,
        "projectName": proj_name,
        "workDir": work,
        "sources": sources,
        "sourcePaths": source_paths,
        "applied": applied if apply else {},
        "owner": {
            "userId": owner_meta.get("ownerUserId"),
            "username": owner_meta.get("ownerUsername"),
            "fullname": owner_meta.get("ownerFullname"),
            "tenantUuid": owner_meta.get("tenantUuid"),
        } if owner_meta else None,
        "sqlPreview": {
            "filename": (sql_preview or {}).get("filename"),
            "fieldMapText": (sql_preview or {}).get("fieldMapText"),
            "tables": (sql_preview or {}).get("tables") or [],
            "error": (sql_preview or {}).get("error"),
        } if sql_preview else None,
        "folder": folder,
        "orders": order_rows,
        "materials": mat_rows,
        "routings": route_rows,
        "equipment": equip_rows,
        "kbActions": kb_actions,
        "plan": plan,
        "readiness": ready,
        "summary": {
            "orders": len(active), "materials": len(mats),
            "routings": len(routes), "equipment": len(equip), "bom": len(bom),
            "ready": s.get("ready", 0), "blocked": s.get("blocked", 0),
        },
        "canSchedule": s.get("ready", 0) > 0 and not ready.get("globalIssues"),
        "readOnly": not apply,
        "pendingAdoption": pending_adoption,
        "markdown": "\n".join(md_lines),
    }

    # 分析结果写入当前租户知识库（平台知识模块，按登录租户隔离）
    knowledge_ingest: dict[str, Any] = {}
    if apply:
        try:
            has_data = bool(
                result["summary"]["materials"]
                or result["summary"]["orders"]
                or result["summary"]["equipment"]
                or result["summary"]["bom"]
                or result.get("sourcePaths")
            )
            if ingest_project_analyze_report is not None and has_data:
                knowledge_ingest = ingest_project_analyze_report(result, world=analysis_world)
                if knowledge_ingest.get("assetId"):
                    md_lines.append("")
                    md_lines.append(
                        f"已写入知识库：《{knowledge_ingest.get('title')}》"
                        f"（{knowledge_ingest.get('chunkCount')} 段，{knowledge_ingest.get('version')}）"
                    )
                    result["markdown"] = "\n".join(md_lines)
        except Exception as exc:
            knowledge_ingest = {"error": str(exc)}
    result["knowledgeIngest"] = knowledge_ingest
    return result