feat: 添加 AI 入库预处理功能，支持数据结构化和异常行导出

README.md

@@ -351,11 +351,11 @@ cp /www/wwwroot/inventory/data/inventory.db /www/backup/inventory_$(date +%F).db
 
 ### 第一阶段：先做能直接省时间的功能
 
-- [ ] AI 入库预处理
+- [x] AI 入库预处理
-- [ ] 支持粘贴采购清单、聊天记录、Excel 文本后，由 AI 自动拆分为 `料号 / 名称 / 数量 / 规格 / 备注`
+- [x] 支持粘贴采购清单、聊天记录、Excel 文本后，由 AI 自动拆分为 `料号 / 名称 / 数量 / 规格 / 备注`
-- [ ] 自动识别脏数据，例如缺字段、数量格式异常、字段顺序混乱
+- [x] 自动识别脏数据，例如缺字段、数量格式异常、字段顺序混乱
-- [ ] 输出结构化预览结果，用户确认后再正式写入库存
+- [x] 输出结构化预览结果，用户确认后再正式写入库存
-- [ ] 对接现有快速入库 / 袋装批量新增流程，不直接绕开人工确认
+- [x] 对接现有快速入库 / 袋装批量新增流程，不直接绕开人工确认
 
 ### 第二阶段：提升库存数据质量
 

app.py

@@ -989,6 +989,201 @@ def _parse_bulk_line(line: str):
     }
 
 
+def _split_inbound_line_fields(line: str) -> tuple[list[str], list[str]]:
+    """把原始入库行尽量拆分成 5 列。
+
+    中文说明：用户粘贴的数据分隔符不一定统一，这里优先识别逗号/Tab，
+    其次尝试竖线或连续空格，尽量把一行拆成系统可识别的字段。
+    """
+    warnings = []
+    raw = (line or "").strip()
+    if not raw:
+        return ["", "", "", "", ""], warnings
+
+    normalized = raw.replace("，", ",")
+    if "," in normalized or "\t" in normalized:
+        parts = [p.strip() for p in re.split(r"[,\t]", normalized)]
+    elif "|" in normalized:
+        parts = [p.strip() for p in normalized.split("|")]
+        warnings.append("检测到竖线分隔，已自动转换为标准字段")
+    else:
+        parts = [p.strip() for p in re.split(r"\s{2,}", normalized) if p.strip()]
+        warnings.append("未检测到逗号或Tab，已按连续空格尝试拆分")
+
+    if len(parts) > 5:
+        parts = parts[:4] + [" ".join(parts[4:])]
+        warnings.append("字段超过 5 列，已将多余内容合并到备注")
+
+    while len(parts) < 5:
+        parts.append("")
+
+    return parts, warnings
+
+
+def _format_inbound_line(part_no: str, name: str, quantity: int, specification: str, note: str) -> str:
+    safe_quantity = int(quantity or 0)
+    return f"{part_no}, {name}, {safe_quantity}, {specification}, {note}".strip()
+
+
+def _parse_inbound_preview_rows(raw_lines: list[str]) -> list[dict]:
+    rows = []
+
+    for line_no, line in enumerate(raw_lines, start=1):
+        parts, warnings = _split_inbound_line_fields(line)
+        part_no = (parts[0] or "").strip()
+        name = (parts[1] or "").strip()
+        quantity_raw = (parts[2] or "").strip()
+        specification = (parts[3] or "").strip()
+        note = (parts[4] or "").strip()
+
+        errors = []
+        if not part_no:
+            errors.append("缺少料号")
+        if not name:
+            errors.append("缺少名称")
+
+        quantity = 0
+        try:
+            quantity = _parse_non_negative_int(quantity_raw, 0)
+        except ValueError:
+            errors.append("数量格式错误，必须是大于等于 0 的整数")
+
+        rows.append(
+            {
+                "line_no": line_no,
+                "raw": line,
+                "part_no": part_no,
+                "name": name,
+                "quantity": int(quantity),
+                "quantity_raw": quantity_raw,
+                "specification": specification,
+                "note": note,
+                "errors": errors,
+                "warnings": warnings,
+                "is_valid": len(errors) == 0,
+                "normalized_line": _format_inbound_line(part_no, name, quantity, specification, note),
+            }
+        )
+
+    return rows
+
+
+def _extract_json_object_block(raw_text: str) -> str:
+    text = (raw_text or "").strip()
+    if not text:
+        return ""
+    if text.startswith("```"):
+        text = re.sub(r"^```(?:json)?\\s*", "", text)
+        text = re.sub(r"\\s*```$", "", text)
+
+    first = text.find("{")
+    last = text.rfind("}")
+    if first >= 0 and last > first:
+        return text[first : last + 1]
+    return text
+
+
+def _normalize_ai_inbound_rows(ai_rows: list, fallback_rows: list[dict]) -> list[dict]:
+    by_line = {row["line_no"]: dict(row) for row in fallback_rows}
+
+    for raw_row in ai_rows or []:
+        if not isinstance(raw_row, dict):
+            continue
+        try:
+            line_no = int(raw_row.get("line_no"))
+        except (TypeError, ValueError):
+            continue
+        if line_no not in by_line:
+            continue
+
+        current = by_line[line_no]
+        part_no = str(raw_row.get("part_no", current["part_no"]) or "").strip()
+        name = str(raw_row.get("name", current["name"]) or "").strip()
+        specification = str(raw_row.get("specification", current["specification"]) or "").strip()
+        note = str(raw_row.get("note", current["note"]) or "").strip()
+
+        quantity = current["quantity"]
+        quantity_candidate = raw_row.get("quantity", current["quantity"])
+        try:
+            quantity = _parse_non_negative_int(str(quantity_candidate), 0)
+        except (TypeError, ValueError):
+            # AI 数量不可信时保留规则解析值，不覆盖。
+            pass
+
+        errors = []
+        if not part_no:
+            errors.append("缺少料号")
+        if not name:
+            errors.append("缺少名称")
+
+        warnings = list(current.get("warnings", []))
+        for w in raw_row.get("warnings", []) if isinstance(raw_row.get("warnings", []), list) else []:
+            text = str(w or "").strip()
+            if text:
+                warnings.append(text)
+
+        current.update(
+            {
+                "part_no": part_no,
+                "name": name,
+                "quantity": int(quantity),
+                "specification": specification,
+                "note": note,
+                "errors": errors,
+                "warnings": warnings,
+                "is_valid": len(errors) == 0,
+                "normalized_line": _format_inbound_line(part_no, name, quantity, specification, note),
+            }
+        )
+
+    return [by_line[idx] for idx in sorted(by_line.keys())]
+
+
+def _ai_enhance_inbound_preview(raw_lines: list[str], mode: str, fallback_rows: list[dict], settings: dict) -> tuple[list[dict], str]:
+    """使用 AI 对规则解析结果做二次修正。
+
+    中文说明：AI 负责“更聪明地拆分和纠错”，但最终仍会做字段约束；
+    如果 AI 不可用或返回异常，自动退回规则解析，不影响使用。
+    """
+    api_key = (settings.get("api_key") or "").strip()
+    api_url = (settings.get("api_url") or "").strip()
+    model = (settings.get("model") or "").strip()
+
+    if not api_key or not api_url or not model:
+        return fallback_rows, "AI 参数未完整配置，已使用规则解析结果"
+
+    numbered_lines = [{"line_no": idx, "raw": line} for idx, line in enumerate(raw_lines, start=1)]
+    system_prompt = (
+        "你是电子元件入库清洗助手。"
+        "必须只输出 JSON，不要 Markdown，不要解释文字。"
+        "请输出对象: {\"rows\":[{\"line_no\":number,\"part_no\":string,\"name\":string,\"quantity\":number,\"specification\":string,\"note\":string,\"warnings\":string[]}]}。"
+        "不要新增或删除行号；每个 line_no 仅返回一条。"
+        "quantity 必须是 >=0 的整数；无法确定时返回 0 并在 warnings 中说明。"
+    )
+    user_prompt = (
+        f"导入模式: {mode}\n"
+        "原始行(JSON):\n"
+        + json.dumps(numbered_lines, ensure_ascii=False)
+        + "\n规则解析参考(JSON):\n"
+        + json.dumps(fallback_rows, ensure_ascii=False)
+    )
+
+    try:
+        suggestion = _call_siliconflow_chat(
+            system_prompt,
+            user_prompt,
+            api_url=api_url,
+            model=model,
+            api_key=api_key,
+            timeout=int(settings.get("timeout", 30)),
+        )
+        parsed = json.loads(_extract_json_object_block(suggestion))
+        ai_rows = parsed.get("rows", []) if isinstance(parsed, dict) else []
+        return _normalize_ai_inbound_rows(ai_rows, fallback_rows), ""
+    except Exception:
+        return fallback_rows, "AI 解析失败，已自动回退到规则解析结果"
+
+
 def log_inventory_event(
     *,
     event_type: str,
@@ -2468,6 +2663,44 @@ def search_page():
     )
 
 
+@app.route("/ai/inbound-parse", methods=["POST"])
+def ai_inbound_parse():
+    """AI 入库预处理接口。
+
+    中文说明：
+    1. 输入原始多行文本；
+    2. 返回结构化行数据 + 脏数据识别；
+    3. 仅提供“预处理与预览”，最终入库仍由用户手动确认提交。
+    """
+    raw_text = request.form.get("lines", "")
+    mode = (request.form.get("mode", "box") or "box").strip().lower()
+    if mode not in {"box", "bag"}:
+        mode = "box"
+
+    lines = [line.strip() for line in (raw_text or "").splitlines() if line.strip()]
+    if not lines:
+        return {"ok": False, "message": "请至少输入一行待处理文本"}, 400
+
+    fallback_rows = _parse_inbound_preview_rows(lines)
+    settings = _get_ai_settings()
+    rows, parse_notice = _ai_enhance_inbound_preview(lines, mode, fallback_rows, settings)
+
+    valid_rows = [row for row in rows if row.get("is_valid")]
+    invalid_rows = [row for row in rows if not row.get("is_valid")]
+    normalized_lines = "\n".join([row.get("normalized_line", "") for row in valid_rows if row.get("normalized_line")])
+
+    return {
+        "ok": True,
+        "mode": mode,
+        "total_lines": len(rows),
+        "valid_count": len(valid_rows),
+        "invalid_count": len(invalid_rows),
+        "parse_notice": parse_notice,
+        "rows": rows,
+        "normalized_lines": normalized_lines,
+    }
+
+
 @app.route("/ai/restock-plan", methods=["POST"])
 def ai_restock_plan():
     """生成 AI 补货建议。

static/css/style.css

@@ -575,6 +575,8 @@ body {
 .muted {
     color: var(--muted);
     font-size: 14px;
+    overflow-wrap: anywhere;
+    word-break: break-word;
 }
 
 .new-box-form {
@@ -856,6 +858,7 @@ body.modal-open {
     display: grid;
     grid-template-columns: 1fr 1fr;
     gap: var(--space-2);
+    min-width: 0;
 }
 
 .form-grid label {
@@ -863,6 +866,13 @@ body.modal-open {
     flex-direction: column;
     gap: 6px;
     font-size: 14px;
+    min-width: 0;
+    overflow-wrap: anywhere;
+    word-break: break-word;
+}
+
+.form-grid > * {
+    min-width: 0;
 }
 
 .form-grid .full {
@@ -880,6 +890,8 @@ input[type="search"] {
     padding: 8px 2px;
     font: inherit;
     transition: border-color 140ms ease;
+    width: 100%;
+    min-width: 0;
 }
 
 input[type="text"]:focus,
@@ -898,6 +910,8 @@ textarea {
     background: transparent;
     color: var(--text);
     font: inherit;
+    width: 100%;
+    min-width: 0;
 }
 
 input[type="checkbox"] {
@@ -968,6 +982,22 @@ th {
     margin-bottom: var(--space-1);
 }
 
+.ai-preview {
+    margin-top: var(--space-1);
+    border: 1px solid var(--line);
+    border-radius: var(--radius);
+    background: color-mix(in srgb, var(--card) 90%, var(--card-alt));
+    padding: 8px;
+}
+
+.ai-preview .row-invalid {
+    background: color-mix(in srgb, var(--danger) 12%, var(--card));
+}
+
+#ai-inbound-status.ok {
+    color: color-mix(in srgb, var(--accent-press) 75%, var(--text));
+}
+
 .entry-shell {
     display: grid;
     grid-template-columns: minmax(0, 1fr) 320px;

templates/box.html

@@ -76,12 +76,34 @@
 					<h2 id="quick-inbound-title">快速入库</h2>
 					<button class="btn btn-light" type="button" id="close-quick-inbound">关闭</button>
 				</div>
-				<p class="hint">每行一条: 料号, 名称, 数量, 规格, 备注。支持英文逗号或Tab分隔；检测到同料号或同参数时不会自动合并，需要人工确认。</p>
+				<p class="hint">每行一条: 料号, 名称, 数量, 规格, 备注。支持英文逗号或Tab分隔；先点“AI预处理”查看结构化结果，再确认导入。</p>
-				<form method="post" action="{{ url_for('quick_inbound', box_id=box.id) }}">
+				<form method="post" id="quick-inbound-form" action="{% if box.box_type == 'bag' %}{{ url_for('add_bag_items_batch', box_id=box.id) }}{% else %}{{ url_for('quick_inbound', box_id=box.id) }}{% endif %}">
-					<textarea class="batch-input" name="lines" rows="8" placeholder="10K-0603, 电阻10K 0603, 500, 1%, 常用\n100nF-0603, 电容100nF 0603, 300, 50V X7R, 去耦"></textarea>
+					<input type="hidden" id="ai-inbound-mode" value="{% if box.box_type == 'bag' %}bag{% else %}box{% endif %}">
+					<textarea class="batch-input" id="quick-inbound-lines" name="lines" rows="8" placeholder="10K-0603, 电阻10K 0603, 500, 1%, 常用\n100nF-0603, 电容100nF 0603, 300, 50V X7R, 去耦"></textarea>
 					<p class="hint">建议: part_no 用厂家型号，name 用品类+型号，specification 只写关键参数。</p>
+					<p class="hint" id="ai-inbound-status" aria-live="polite"></p>
+					<section class="ai-preview" id="ai-inbound-preview" hidden>
+						<div class="table-wrap">
+							<table>
+								<thead>
+									<tr>
+										<th>行</th>
+										<th>料号</th>
+										<th>名称</th>
+										<th>数量</th>
+										<th>规格</th>
+										<th>备注</th>
+										<th>识别提示</th>
+									</tr>
+								</thead>
+								<tbody id="ai-inbound-preview-body"></tbody>
+							</table>
+						</div>
+					</section>
 					<div class="actions">
-						<button class="btn" type="submit">批量快速入库</button>
+						<button class="btn btn-light" type="button" id="ai-inbound-parse-btn">AI预处理并预览</button>
+						<button class="btn btn-light" type="button" id="ai-inbound-export-invalid-btn" disabled>导出异常行</button>
+						<button class="btn" type="submit">确认导入</button>
 					</div>
 				</form>
 			</div>
@@ -217,6 +239,184 @@
 				});
 			});
 		})();
+
+		(function () {
+			var parseBtn = document.getElementById('ai-inbound-parse-btn');
+			var exportInvalidBtn = document.getElementById('ai-inbound-export-invalid-btn');
+			var textarea = document.getElementById('quick-inbound-lines');
+			var modeInput = document.getElementById('ai-inbound-mode');
+			var status = document.getElementById('ai-inbound-status');
+			var preview = document.getElementById('ai-inbound-preview');
+			var previewBody = document.getElementById('ai-inbound-preview-body');
+			var latestRows = [];
+
+			if (!parseBtn || !textarea || !status || !preview || !previewBody || !exportInvalidBtn) {
+				return;
+			}
+
+			function escapeHtml(text) {
+				return String(text || '')
+					.replace(/&/g, '&amp;')
+					.replace(/</g, '&lt;')
+					.replace(/>/g, '&gt;')
+					.replace(/"/g, '&quot;')
+					.replace(/'/g, '&#39;');
+			}
+
+			function renderRows(rows) {
+				latestRows = Array.isArray(rows) ? rows : [];
+				var invalidCount = latestRows.filter(function (row) {
+					return !row.is_valid;
+				}).length;
+				exportInvalidBtn.disabled = invalidCount <= 0;
+
+				if (!Array.isArray(rows) || !rows.length) {
+					preview.hidden = true;
+					previewBody.innerHTML = '';
+					return;
+				}
+
+				var html = rows.map(function (row) {
+					var messages = [];
+					(row.errors || []).forEach(function (msg) {
+						messages.push('错误: ' + msg);
+					});
+					(row.warnings || []).forEach(function (msg) {
+						messages.push('提示: ' + msg);
+					});
+					var tips = messages.length ? messages.join('<br>') : '正常';
+					var rowClass = row.is_valid ? '' : ' class="row-invalid"';
+
+					return '<tr' + rowClass + '>'
+						+ '<td>' + escapeHtml(row.line_no) + '</td>'
+						+ '<td>' + escapeHtml(row.part_no) + '</td>'
+						+ '<td>' + escapeHtml(row.name) + '</td>'
+						+ '<td>' + escapeHtml(row.quantity) + '</td>'
+						+ '<td>' + escapeHtml(row.specification) + '</td>'
+						+ '<td>' + escapeHtml(row.note) + '</td>'
+						+ '<td>' + tips + '</td>'
+						+ '</tr>';
+				}).join('');
+
+				previewBody.innerHTML = html;
+				preview.hidden = false;
+			}
+
+			parseBtn.addEventListener('click', function () {
+				var lines = (textarea.value || '').trim();
+				if (!lines) {
+					status.textContent = '请先粘贴至少一行内容';
+					status.classList.remove('ok');
+					return;
+				}
+
+				parseBtn.disabled = true;
+				exportInvalidBtn.disabled = true;
+				status.textContent = '正在进行 AI 预处理...';
+				status.classList.remove('ok');
+
+				var payload = new URLSearchParams();
+				payload.set('lines', lines);
+				payload.set('mode', modeInput ? modeInput.value : 'box');
+
+				fetch('{{ url_for('ai_inbound_parse') }}', {
+					method: 'POST',
+					headers: {
+						'Content-Type': 'application/x-www-form-urlencoded;charset=UTF-8'
+					},
+					body: payload.toString()
+				}).then(function (resp) {
+					return resp.json().then(function (data) {
+						if (!resp.ok || !data.ok) {
+							throw new Error(data.message || '预处理失败');
+						}
+						return data;
+					});
+				}).then(function (data) {
+					renderRows(data.rows || []);
+					if (data.normalized_lines) {
+						textarea.value = data.normalized_lines;
+					}
+
+					var baseText = '预处理完成: 有效 ' + (data.valid_count || 0) + ' 行';
+					if (data.invalid_count) {
+						baseText += '，异常 ' + data.invalid_count + ' 行';
+					}
+					if (data.parse_notice) {
+						baseText += '。' + data.parse_notice;
+					}
+
+					status.textContent = baseText;
+					status.classList.add('ok');
+				}).catch(function (error) {
+					status.textContent = '预处理失败: ' + error.message;
+					status.classList.remove('ok');
+				}).finally(function () {
+					parseBtn.disabled = false;
+				});
+			});
+
+			exportInvalidBtn.addEventListener('click', function () {
+				var invalidRows = latestRows.filter(function (row) {
+					return !row.is_valid;
+				});
+
+				if (!invalidRows.length) {
+					status.textContent = '当前没有异常行可导出';
+					status.classList.remove('ok');
+					return;
+				}
+
+				var headers = ['line_no', 'raw', 'part_no', 'name', 'quantity', 'specification', 'note', 'errors', 'warnings'];
+				var rows = [headers];
+
+				invalidRows.forEach(function (row) {
+					rows.push([
+						row.line_no || '',
+						row.raw || '',
+						row.part_no || '',
+						row.name || '',
+						row.quantity || 0,
+						row.specification || '',
+						row.note || '',
+						(row.errors || []).join(' | '),
+						(row.warnings || []).join(' | ')
+					]);
+				});
+
+				function csvCell(value) {
+					var text = String(value == null ? '' : value);
+					if (/[",\n]/.test(text)) {
+						return '"' + text.replace(/"/g, '""') + '"';
+					}
+					return text;
+				}
+
+				var csv = rows.map(function (cols) {
+					return cols.map(csvCell).join(',');
+				}).join('\n');
+				var blob = new Blob(['\ufeff' + csv], { type: 'text/csv;charset=utf-8;' });
+				var url = URL.createObjectURL(blob);
+				var now = new Date();
+				var stamp = now.getFullYear()
+					+ String(now.getMonth() + 1).padStart(2, '0')
+					+ String(now.getDate()).padStart(2, '0')
+					+ '_'
+					+ String(now.getHours()).padStart(2, '0')
+					+ String(now.getMinutes()).padStart(2, '0')
+					+ String(now.getSeconds()).padStart(2, '0');
+				var link = document.createElement('a');
+				link.href = url;
+				link.download = 'inbound_invalid_rows_' + stamp + '.csv';
+				document.body.appendChild(link);
+				link.click();
+				document.body.removeChild(link);
+				URL.revokeObjectURL(url);
+
+				status.textContent = '已导出异常行 ' + invalidRows.length + ' 条';
+				status.classList.add('ok');
+			});
+		})();
 	</script>
 </body>
 </html>