加快登记预览并结构化补项

app/app/ai_register.py

@@ -10,8 +10,9 @@ import tomllib
 from typing import Any
 from urllib import error, request
 
-from .config import AI_REGISTER_TIMEOUT_SECONDS, CODEX_AUTH_PATH, CODEX_CONFIG_PATH
+from .config import ACCOUNTS_PATH, AI_REGISTER_MODEL, AI_REGISTER_TIMEOUT_SECONDS, CODEX_AUTH_PATH, CODEX_CONFIG_PATH
 from .data import (
+    SUBJECTS,
     class_record_to_line,
     course_summary_to_class_record_line,
     extract_course_summary_from_text,
@@ -19,10 +20,48 @@ from .data import (
     normalize_lines,
     parse_class_record_line,
     parse_payment_line,
+    read_accounts,
 )
 
 
 REGISTER_TYPES = {"class_record", "payment", "course_summary"}
+FIELD_LABELS = {
+    "class_record": {
+        "student": "学生",
+        "date": "日期",
+        "time": "时间",
+        "teacher": "老师",
+        "subject": "科目",
+        "duration": "时长",
+    },
+    "payment": {
+        "student": "学生",
+        "date": "缴费日期",
+        "hours": "课时数",
+    },
+    "course_summary": {
+        "student": "学生",
+        "date": "日期",
+        "time": "时间",
+        "teacher": "老师",
+        "subject": "科目",
+        "body": "小结正文",
+    },
+}
+REQUIRED_FIELDS = {
+    "class_record": ["student", "date", "time", "teacher", "subject"],
+    "payment": ["student", "date", "hours"],
+    "course_summary": ["student", "date", "time", "teacher", "subject", "body"],
+}
+DATE_RE = re.compile(r"(?P<y>\d{4})[./年-]\s*(?P<m>\d{1,2})[./月-]\s*(?P<d>\d{1,2})")
+MONTH_DAY_RE = re.compile(r"(?<!\d)(?P<m>\d{1,2})\s*[月./-]\s*(?P<d>\d{1,2})\s*(?:日|号)?")
+TIME_RANGE_FLEX_RE = re.compile(
+    r"(?P<sh>\d{1,2})\s*(?:[:：点.])\s*(?P<sm>\d{1,2})?\s*"
+    r"(?:-|－|—|–|~|至|到)\s*"
+    r"(?P<eh>\d{1,2})\s*(?:[:：点.])\s*(?P<em>\d{1,2})?"
+)
+TEACHER_RE = re.compile(r"(?P<teacher>[\u4e00-\u9fa5A-Za-z0-9]{1,8}老师)")
+HOURS_RE = re.compile(r"(?P<hours>\d+(?:\.\d+)?)\s*(?:课时|小时)")
 MAX_SESSION_AGE_SECONDS = 30 * 60
 MAX_SESSIONS = 100
 _SESSIONS: dict[str, dict[str, Any]] = {}
@@ -113,7 +152,7 @@ def load_codex_ai_config(
     config = tomllib.loads(config_path.read_text(encoding="utf-8"))
     auth = json.loads(auth_path.read_text(encoding="utf-8"))
     provider_name = str(config.get("model_provider") or "").strip()
-    model = str(config.get("model") or "").strip()
+    model = AI_REGISTER_MODEL or str(config.get("model") or "").strip()
     providers = config.get("model_providers") or {}
     provider = providers.get(provider_name) or {}
     base_url = str(provider.get("base_url") or "").strip().rstrip("/")
@@ -128,6 +167,195 @@ def load_codex_ai_config(
     return CodexAiConfig(provider_name, model, base_url, wire_api, api_key)
 
 
+def _field_labels(register_type: str) -> dict[str, str]:
+    return FIELD_LABELS.get(register_type, {})
+
+
+def _questions_for_missing(register_type: str, missing_fields: list[str]) -> list[str]:
+    labels = _field_labels(register_type)
+    return [f"请补充{labels.get(field, field)}" for field in missing_fields]
+
+
+def _known_students() -> list[str]:
+    try:
+        return [account.student for account in read_accounts(ACCOUNTS_PATH)]
+    except Exception:
+        return []
+
+
+def _normalize_date(value: object) -> str:
+    text = str(value or "").strip()
+    if not text:
+        return ""
+    match = DATE_RE.search(text)
+    if match:
+        year = int(match.group("y"))
+        month = int(match.group("m"))
+        day = int(match.group("d"))
+        datetime(year, month, day)
+        return f"{year:04d}-{month:02d}-{day:02d}"
+    match = MONTH_DAY_RE.search(text)
+    if match:
+        year = date.today().year
+        month = int(match.group("m"))
+        day = int(match.group("d"))
+        datetime(year, month, day)
+        return f"{year:04d}-{month:02d}-{day:02d}"
+    return ""
+
+
+def _record_date(value: str) -> str:
+    return value.replace("-", ".")
+
+
+def _weekday(value: str) -> str:
+    weekdays = ["星期一", "星期二", "星期三", "星期四", "星期五", "星期六", "星期日"]
+    return weekdays[datetime.strptime(value, "%Y-%m-%d").date().weekday()]
+
+
+def _normalize_time_range(value: object) -> str:
+    text = str(value or "").strip()
+    if not text:
+        return ""
+    standard = re.search(r"\d{1,2}:\d{2}-\d{1,2}:\d{2}", text)
+    if standard:
+        raw = standard.group(0)
+        start, end = raw.split("-", 1)
+        sh, sm = [int(part) for part in start.split(":", 1)]
+        eh, em = [int(part) for part in end.split(":", 1)]
+    else:
+        match = TIME_RANGE_FLEX_RE.search(text)
+        if not match:
+            return ""
+        sh = int(match.group("sh"))
+        sm = int(match.group("sm") or 0)
+        eh = int(match.group("eh"))
+        em = int(match.group("em") or 0)
+    if sh > 23 or eh > 23 or sm > 59 or em > 59:
+        return ""
+    if eh * 60 + em <= sh * 60 + sm:
+        return ""
+    return f"{sh:02d}:{sm:02d}-{eh:02d}:{em:02d}"
+
+
+def _duration_from_time_range(time_range: str) -> str:
+    start, end = time_range.split("-", 1)
+    sh, sm = [int(part) for part in start.split(":", 1)]
+    eh, em = [int(part) for part in end.split(":", 1)]
+    minutes = eh * 60 + em - sh * 60 - sm
+    return f"{minutes // 60}小时{minutes % 60}分"
+
+
+def _duration_from_text(value: object) -> str:
+    text = str(value or "").strip()
+    if not text:
+        return ""
+    if re.fullmatch(r"\d+小时\d+分", text):
+        return text
+    match = re.search(r"(?P<h>\d+)\s*小时\s*(?P<m>\d+)?\s*分?", text)
+    if match:
+        return f"{int(match.group('h'))}小时{int(match.group('m') or 0)}分"
+    match = re.search(r"(?P<hours>\d+(?:\.\d+)?)\s*(?:课时|小时)", text)
+    if match:
+        total = int(round(float(match.group("hours")) * 60))
+        return f"{total // 60}小时{total % 60}分"
+    return ""
+
+
+def _extract_student(text: str) -> str:
+    match = re.search(r"(?:学生|学员)[:：\s]*(?P<student>[\u4e00-\u9fa5A-Za-z0-9]{2,8})", text)
+    if match:
+        return match.group("student")
+    for student in _known_students():
+        if student and student in text:
+            return student
+    match = re.search(r"^\s*(?P<student>[\u4e00-\u9fa5]{2,4})(?=\s|[,，。；;:]|\d)", text)
+    return match.group("student") if match else ""
+
+
+def _extract_teacher(text: str) -> str:
+    match = TEACHER_RE.search(text)
+    return match.group("teacher") if match else ""
+
+
+def _extract_subject(text: str) -> str:
+    for subject in SUBJECTS:
+        if subject in text:
+            return subject
+    match = re.search(r"(?:科目|课程)[:：\s]*(?P<subject>[\u4e00-\u9fa5A-Za-z0-9]{1,12})", text)
+    return match.group("subject") if match else ""
+
+
+def _merge_answers(fields: dict[str, str], answers: dict[str, str]) -> dict[str, str]:
+    merged = dict(fields)
+    for key, value in answers.items():
+        text = str(value or "").strip()
+        if not text:
+            continue
+        if key == "followup":
+            continue
+        merged[key] = text
+    return merged
+
+
+def _missing_fields(register_type: str, fields: dict[str, str]) -> list[str]:
+    return [field for field in REQUIRED_FIELDS[register_type] if not str(fields.get(field) or "").strip()]
+
+
+def _needs_info_response(
+    register_type: str,
+    fields: dict[str, str],
+    missing_fields: list[str],
+    *,
+    summary: str = "",
+    ai_used: bool = False,
+    timed_out: bool = False,
+    error: str = "",
+) -> dict[str, Any]:
+    return {
+        "status": "needs_info",
+        "standard_lines": [],
+        "questions": _questions_for_missing(register_type, missing_fields) or [_question_for_error(register_type, error or "信息不完整")],
+        "summary": summary or "请补齐缺失字段后再次生成预览",
+        "ai_used": ai_used,
+        "fields": fields,
+        "missing_fields": missing_fields,
+        "field_labels": _field_labels(register_type),
+        "timed_out": timed_out,
+        "recognition_source": "partial_local" if not ai_used else "model",
+        "error": error,
+    }
+
+
+def _standard_from_fields(register_type: str, fields: dict[str, str]) -> str:
+    if register_type == "class_record":
+        date_value = _normalize_date(fields.get("date"))
+        time_range = _normalize_time_range(fields.get("time"))
+        duration = _duration_from_text(fields.get("duration")) or _duration_from_time_range(time_range)
+        return (
+            f"{_record_date(date_value)}-{_weekday(date_value)}-{time_range}-"
+            f"{fields['student']}-{duration}-{fields['teacher']}-{fields['subject']}"
+        )
+    if register_type == "payment":
+        date_value = _normalize_date(fields.get("date"))
+        hours = str(fields.get("hours") or "").strip()
+        hours = re.sub(r"\s*(?:课时|小时)$", "", hours)
+        return f"{fields['student']}-{date_value}:{hours}"
+    date_value = _normalize_date(fields.get("date"))
+    time_range = _normalize_time_range(fields.get("time"))
+    return "\n".join(
+        [
+            f"学生：{fields['student']}",
+            f"日期：{date_value}",
+            f"时间：{time_range}",
+            f"老师：{fields['teacher']}",
+            f"科目：{fields['subject']}",
+            "小结：",
+            str(fields.get("body") or "").strip(),
+        ]
+    )
+
+
 def _json_from_text(text: str) -> dict[str, Any]:
     try:
         payload = json.loads(text)
@@ -229,6 +457,75 @@ def _question_for_error(register_type: str, message: str) -> str:
     return f"请补齐或修改课程小结信息：{message}"
 
 
+def _extract_class_record_fields(text: str) -> dict[str, str]:
+    return {
+        "student": _extract_student(text),
+        "date": _normalize_date(text),
+        "time": _normalize_time_range(text),
+        "teacher": _extract_teacher(text),
+        "subject": _extract_subject(text),
+        "duration": _duration_from_text(text),
+    }
+
+
+def _extract_payment_fields(text: str) -> dict[str, str]:
+    hours_match = HOURS_RE.search(text)
+    return {
+        "student": _extract_student(text),
+        "date": _normalize_date(text),
+        "hours": hours_match.group("hours") if hours_match else "",
+    }
+
+
+def _extract_course_summary_fields(text: str) -> dict[str, str]:
+    try:
+        raw = extract_course_summary_from_text(text, known_students=_known_students())
+    except ValueError:
+        raw = {}
+    body = str(raw.get("body") or "").strip()
+    if body == text.strip() and any(label in text for label in ("小结", "正文", "内容")):
+        body = re.split(r"(?:小结|正文|内容)[:：]?", text, maxsplit=1)[-1].strip()
+    return {
+        "student": str(raw.get("student") or _extract_student(text)).strip(),
+        "date": _normalize_date(raw.get("date_iso") or text),
+        "time": _normalize_time_range(raw.get("time_range") or text),
+        "teacher": str(raw.get("teacher") or _extract_teacher(text)).strip(),
+        "subject": str(raw.get("subject") or _extract_subject(text)).strip(),
+        "body": body or text.strip(),
+    }
+
+
+def extract_local_fields(register_type: str, lines: list[str]) -> dict[str, str]:
+    text = "\n".join(lines).strip()
+    if register_type == "class_record":
+        return _extract_class_record_fields(text)
+    if register_type == "payment":
+        return _extract_payment_fields(text)
+    return _extract_course_summary_fields(text)
+
+
+def fields_preview(register_type: str, fields: dict[str, str]) -> dict[str, Any] | None:
+    missing = _missing_fields(register_type, fields)
+    if missing:
+        return _needs_info_response(register_type, fields, missing)
+    try:
+        standard_lines = validate_standard_lines(register_type, [_standard_from_fields(register_type, fields)])
+    except (ValueError, KeyError):
+        return None
+    return {
+        "status": "ready",
+        "standard_lines": standard_lines,
+        "questions": [],
+        "summary": "已通过本地规则生成预览",
+        "ai_used": False,
+        "fields": fields,
+        "missing_fields": [],
+        "field_labels": _field_labels(register_type),
+        "timed_out": False,
+        "recognition_source": "local",
+    }
+
+
 def validate_standard_lines(register_type: str, lines: list[str]) -> list[str]:
     if not lines:
         raise ValueError("模型未返回标准行")
@@ -257,17 +554,23 @@ def validate_standard_lines(register_type: str, lines: list[str]) -> list[str]:
     return standard_lines
 
 
-def local_preview(register_type: str, lines: list[str]) -> dict[str, Any] | None:
+def local_preview(register_type: str, lines: list[str], answers: dict[str, str] | None = None) -> dict[str, Any] | None:
     try:
         standard_lines = validate_standard_lines(register_type, lines)
     except ValueError:
-        return None
+        fields = _merge_answers(extract_local_fields(register_type, lines), answers or {})
+        return fields_preview(register_type, fields)
     return {
         "status": "ready",
         "standard_lines": standard_lines,
         "questions": [],
         "summary": "已按标准格式通过本地校验",
         "ai_used": False,
+        "fields": {},
+        "missing_fields": [],
+        "field_labels": _field_labels(register_type),
+        "timed_out": False,
+        "recognition_source": "local",
     }
 
 
@@ -286,35 +589,35 @@ def preview_register(
     merged_answers = {**dict(session.get("answers") or {}), **(answers or {})}
     session["answers"] = merged_answers
 
-    local = local_preview(normalized_type, source_lines)
-    if local is not None and not merged_answers:
+    local = local_preview(normalized_type, source_lines, merged_answers)
+    if local is not None:
         return {"conversation_id": conversation_id, "type": normalized_type, **local}
 
     try:
         model_payload = call_model_for_standardization(normalized_type, source_lines, merged_answers)
     except ValueError as exc:
-        if local is not None:
-            return {
-                "conversation_id": conversation_id,
-                "type": normalized_type,
-                **local,
-                "warning": "模型不可用，已使用本地标准格式校验预览",
-            }
+        fields = _merge_answers(extract_local_fields(normalized_type, source_lines), merged_answers)
+        missing = _missing_fields(normalized_type, fields)
+        timed_out = "timed out" in str(exc).lower() or "timeout" in str(exc).lower()
         return {
             "conversation_id": conversation_id,
             "type": normalized_type,
-            "status": "error",
-            "standard_lines": [],
-            "questions": [_question_for_error(normalized_type, str(exc))],
-            "summary": "",
-            "ai_used": True,
-            "error": str(exc),
+            **_needs_info_response(
+                normalized_type,
+                fields,
+                missing,
+                ai_used=True,
+                timed_out=timed_out,
+                error=str(exc),
+                summary="模型未及时生成可靠预览，请补齐字段后再生成",
+            ),
         }
 
     status = str(model_payload.get("status") or "").strip()
     questions = [str(item).strip() for item in model_payload.get("questions") or [] if str(item).strip()]
     if status == "needs_info" or questions:
         session["questions"] = questions
+        fields = _merge_answers(extract_local_fields(normalized_type, source_lines), merged_answers)
         return {
             "conversation_id": conversation_id,
             "type": normalized_type,
@@ -323,11 +626,17 @@ def preview_register(
             "questions": questions or ["请补齐缺失信息"],
             "summary": str(model_payload.get("summary") or ""),
             "ai_used": True,
+            "fields": fields,
+            "missing_fields": _missing_fields(normalized_type, fields),
+            "field_labels": _field_labels(normalized_type),
+            "timed_out": False,
+            "recognition_source": "model",
         }
 
     try:
         standard_lines = validate_standard_lines(normalized_type, model_payload.get("standard_lines") or [])
     except ValueError as exc:
+        fields = _merge_answers(extract_local_fields(normalized_type, source_lines), merged_answers)
         return {
             "conversation_id": conversation_id,
             "type": normalized_type,
@@ -337,6 +646,11 @@ def preview_register(
             "summary": str(model_payload.get("summary") or ""),
             "ai_used": True,
             "error": str(exc),
+            "fields": fields,
+            "missing_fields": _missing_fields(normalized_type, fields),
+            "field_labels": _field_labels(normalized_type),
+            "timed_out": False,
+            "recognition_source": "model",
         }
     return {
         "conversation_id": conversation_id,
@@ -346,4 +660,9 @@ def preview_register(
         "questions": [],
         "summary": str(model_payload.get("summary") or "已生成标准登记内容"),
         "ai_used": True,
+        "fields": {},
+        "missing_fields": [],
+        "field_labels": _field_labels(normalized_type),
+        "timed_out": False,
+        "recognition_source": "model",
     }

app/app/config.py

@@ -17,7 +17,8 @@ COURSE_SUMMARY_STATE_PATH = Path(os.getenv("COURSE_SUMMARY_STATE_PATH", "/data/c
 OPERATION_LOGS_PATH = Path(os.getenv("OPERATION_LOGS_PATH", "/data/operation_logs.jsonl"))
 CODEX_CONFIG_PATH = Path(os.getenv("CODEX_CONFIG_PATH", "/run/codex/config.toml"))
 CODEX_AUTH_PATH = Path(os.getenv("CODEX_AUTH_PATH", "/run/codex/auth.json"))
-AI_REGISTER_TIMEOUT_SECONDS = float(os.getenv("AI_REGISTER_TIMEOUT_SECONDS", "90"))
+AI_REGISTER_TIMEOUT_SECONDS = float(os.getenv("AI_REGISTER_TIMEOUT_SECONDS", "8"))
+AI_REGISTER_MODEL = os.getenv("AI_REGISTER_MODEL", "").strip()
 
 BASIC_AUTH_PASSWORD = os.getenv("BASIC_AUTH_PASSWORD", "")
 ACCOUNTS_AUTH_PASSWORD = os.getenv("ACCOUNTS_AUTH_PASSWORD") or os.getenv("ACCOUNT_AUTH_PASSWORD", "")

app/app/static/admin.js

@@ -993,7 +993,23 @@ function renderRegisterPreview(type, data, statusNode) {
   }
   if (data.status === "needs_info") {
     const questions = (data.questions || []).map((item) => `<li>${escapeHtml(item)}</li>`).join("");
-    previewNode.innerHTML = `<strong>需要补充信息</strong><ul>${questions}</ul><textarea data-ai-answer rows="3" placeholder="在这里补充缺失信息后再次生成预览"></textarea>`;
+    const fields = data.fields || {};
+    const labels = data.field_labels || {};
+    const missing = new Set(data.missing_fields || []);
+    const fieldNames = Object.keys(labels);
+    const fieldInputs = fieldNames
+      .map((name) => {
+        const label = labels[name] || name;
+        const value = fields[name] || "";
+        const requiredMark = missing.has(name) ? " *" : "";
+        const input =
+          name === "body"
+            ? `<textarea data-ai-field="${escapeHtml(name)}" rows="4">${escapeHtml(value)}</textarea>`
+            : `<input data-ai-field="${escapeHtml(name)}" value="${escapeHtml(value)}">`;
+        return `<label><span>${escapeHtml(label)}${requiredMark}</span>${input}</label>`;
+      })
+      .join("");
+    previewNode.innerHTML = `<strong>需要补充信息</strong><ul>${questions}</ul><div class="register-field-grid">${fieldInputs}</div>`;
     confirmButton.hidden = true;
     registerPreviewState[type] = {
       conversationId: data.conversation_id,
@@ -1025,8 +1041,13 @@ async function previewRegister(event, type, lines, statusNode) {
     course_summary: summaryRegisterPreview,
   }[type];
   const answerNode = previewNode ? previewNode.querySelector("[data-ai-answer]") : null;
+  const fieldNodes = previewNode ? previewNode.querySelectorAll("[data-ai-field]") : [];
   const state = registerPreviewState[type] || {};
-  const answers = answerNode && answerNode.value.trim() ? { followup: answerNode.value.trim() } : {};
+  const answers = {};
+  if (answerNode && answerNode.value.trim()) answers.followup = answerNode.value.trim();
+  fieldNodes.forEach((node) => {
+    if (node.value.trim()) answers[node.dataset.aiField] = node.value.trim();
+  });
   clearRegisterPreview(type, !state.conversationId);
   statusNode.textContent = "正在生成预览";
   try {

app/app/static/styles.css

@@ -431,6 +431,29 @@ textarea:focus {
   min-height: 72px;
 }
 
+.register-field-grid {
+  display: grid;
+  grid-template-columns: repeat(auto-fit, minmax(160px, 1fr));
+  gap: 10px;
+}
+
+.register-field-grid label {
+  display: grid;
+  gap: 4px;
+  color: #52606d;
+  font-size: 13px;
+}
+
+.register-field-grid input,
+.register-field-grid textarea {
+  width: 100%;
+  box-sizing: border-box;
+}
+
+.register-field-grid label:has(textarea) {
+  grid-column: 1 / -1;
+}
+
 .line-code {
   display: block;
   max-width: 420px;

app/docker-compose.yml

@@ -19,7 +19,8 @@ services:
       OPERATION_LOGS_PATH: ${OPERATION_LOGS_PATH:-/data/operation_logs.jsonl}
       CODEX_CONFIG_PATH: ${CODEX_CONFIG_PATH:-/run/codex/config.toml}
       CODEX_AUTH_PATH: ${CODEX_AUTH_PATH:-/run/codex/auth.json}
-      AI_REGISTER_TIMEOUT_SECONDS: ${AI_REGISTER_TIMEOUT_SECONDS:-90}
+      AI_REGISTER_TIMEOUT_SECONDS: ${AI_REGISTER_TIMEOUT_SECONDS:-8}
+      AI_REGISTER_MODEL: ${AI_REGISTER_MODEL:-}
       INGEST_AUTH_TOKEN: ${INGEST_AUTH_TOKEN:-}
     ports:
       - "${APP_PORT:-18080}:8000"