feat(agent): LLM-powered collection node with welcome stage and session-scoped delete (assignment 4)

- prompts/collection.md: versioned collection system prompt extracted from workflow node yQaBxfrWjbRr with source/version/porting notes - llm.py: LLMService protocol; FakeLLMService (scripted, records calls); OpenAILLMService (async, wait_for timeout with cancellation, retry only on timeout/429/connection, errors normalized) - collect.py: run_collect reference impl - one LLM call produces state prefix + body, controlled repair (max 1 retry) for missing prefix or illegal stage proposal, whitelist guard, CollectReplyError without raw model content; dialogue history + confirmed fields injected into prompt so the model skips known info (作业4-5) - fields.py / clock.py: three-state field candidates (unknown/ candidate/confirmed), per-type validation (count not inferred from description, future time rejected), injectable Clock - graph.py: welcome node - first message returns fixed 1001 greeting without LLM (mirrors workflow is_first_message -> Assigned Reply#3), route empty stage to welcome - endpoints/delete_session: backend switch - langgraph clears graph checkpoint, fastgpt path unchanged - config: AGENT_LLM_PROVIDER fake|openai with validation - untrack .env (secrets), move template to root .env.example Verified: - pytest 103 passed; smoke turns 1/2/1 - Real LLM e2e: POST /chat multi-turn context works - Golden eval collection_skip_known_info: 12/12 cases pass - Golden eval no_false: 8/10 cases pass; 4 confirm/photo cases fail at final step - model routes on short reply '对' (prompt-adherence gap, photo/confirm stages are assignment 5 scope); recorded in evals/results/20261020-144856 Known limits: field extraction from model reply not wired (stub); photo/confirm nodes still scaffold; clarify counting is approximate.
2026-10-20 14:53:46 +08:00
parent 8323d53a70
commit 1c6b1a513d
17 changed files with 1596 additions and 34 deletions
@@ -0,0 +1,136 @@
+"""事故字段候选与校验模型（作业 4 关键代码）。
+
+来源字段清单对照 ``src/api/endpoints.py`` 的 get_info 分组：
+- 事故字段（acd_keys）与两方人员字段（human1/human2_keys）；
+- 每个字段区分 未知 / 候选 / 已确认 三态（计划作业 4-5）：
+  已明确提供的有效信息直接记为已确认并跳过已知项；
+  候选值需经核实节点确认后才覆盖正式字段。
+"""
+
+from enum import StrEnum
+from typing import Any
+
+from pydantic import BaseModel, Field
+
+from .clock import Clock, SystemClock, validate_accident_time
+
+ACCIDENT_FIELD_KEYS = (
+    "ywrysw", "ywfjdc", "ywmtc", "bjrjs",
+    "sgfssj", "sfsgxc", "jdcsl", "sgyy",
+)
+PERSON1_FIELD_KEYS = (
+    "xm1", "hpzl1", "hphm1", "sfzmhm1", "sfzmwh1", "sjhm1", "sjwh1", "csbw1",
+)
+PERSON2_FIELD_KEYS = (
+    "xm2", "hpzl2", "hphm2", "sfzmhm2", "sfzmwh2", "sjhm2", "sjwh2", "csbw2",
+)
+KNOWN_FIELD_KEYS = frozenset(
+    ACCIDENT_FIELD_KEYS + PERSON1_FIELD_KEYS + PERSON2_FIELD_KEYS
+)
+
+TIME_FIELDS = frozenset({"sgfssj"})  # 事故发生时间：需时间校验
+
+# 布尔语义字段：明确肯定/否定之外（模糊、ASR 破碎）不得猜测。
+_BOOL_FIELD_KEYS = frozenset({"ywrysw", "ywfjdc", "ywmtc", "sfsgxc", "bjrjs"})
+_BOOL_TRUE_WORDS = frozenset({"是", "是的", "有", "有人受伤", "true", "1"})
+_BOOL_FALSE_WORDS = frozenset({"否", "不是", "没有", "没人受伤", "false", "0"})
+
+
+class FieldStatus(StrEnum):
+    UNKNOWN = "unknown"
+    CANDIDATE = "candidate"
+    CONFIRMED = "confirmed"
+
+
+class FieldCandidate(BaseModel):
+    """一个待提交/核实的字段值。"""
+
+    key: str
+    value: Any
+    status: FieldStatus = FieldStatus.CANDIDATE
+    source_turn: int = 0  # 产生候选的对话轮数
+
+
+def _parse_count(value) -> int | None:
+    """把数量解析为正整数；无法解析或小于 1 返回 None。
+
+    注意：不得从“追尾”“剐蹭”等描述性词语推断数量——
+    那些输入走 NLP 理解后才可能成为候选，本函数只做字面解析。
+    """
+    if isinstance(value, bool):
+        return None
+    if isinstance(value, int):
+        return value if value >= 1 else None
+    if isinstance(value, str) and value.strip().isdigit():
+        n = int(value.strip())
+        return n if n >= 1 else None
+    return None
+
+
+def _parse_bool(value) -> bool | None:
+    """解析明确肯定/否定；模糊或无法判定返回 None（保持候选，不猜）。"""
+    if isinstance(value, bool):
+        return value
+    if isinstance(value, str):
+        v = value.strip().lower()
+        if v in _BOOL_TRUE_WORDS:
+            return True
+        if v in _BOOL_FALSE_WORDS:
+            return False
+    return None
+
+
+def validate_candidate(
+    candidate: FieldCandidate,
+    clock: Clock | None = None,
+) -> FieldCandidate:
+    """校验候选字段，返回带最终状态的候选。
+
+    三态语义（计划作业 4-5）：
+    - 语义明确且校验通过 → CONFIRMED（可直接写入正式字段并跳过已知项）；
+    - 语义模糊/格式不合规/未来时间 → 保持 CANDIDATE（待核实节点确认）；
+    - 空值 → UNKNOWN。
+    未知 key 直接拒绝（ValueError），不得写入正式字段。
+    """
+    if candidate.key not in KNOWN_FIELD_KEYS:
+        raise ValueError(f"未知字段 key: {candidate.key!r}")
+    if candidate.value in (None, ""):
+        return candidate.model_copy(update={"status": FieldStatus.UNKNOWN})
+
+    if candidate.key == "jdcsl":
+        count = _parse_count(candidate.value)
+        if count is not None:
+            return candidate.model_copy(
+                update={"value": count, "status": FieldStatus.CONFIRMED}
+            )
+        return candidate  # CANDIDATE：数量未知，不得推断
+
+    if candidate.key in TIME_FIELDS:
+        text = str(candidate.value)
+        checked = validate_accident_time(text, (clock or SystemClock()).now())
+        status = FieldStatus.CONFIRMED if checked else FieldStatus.CANDIDATE
+        return candidate.model_copy(update={"value": text, "status": status})
+
+    if candidate.key in _BOOL_FIELD_KEYS:
+        parsed = _parse_bool(candidate.value)
+        if parsed is not None:
+            return candidate.model_copy(
+                update={"value": parsed, "status": FieldStatus.CONFIRMED}
+            )
+        return candidate  # CANDIDATE：模糊/ASR 破碎，不猜
+
+    return candidate.model_copy(update={"status": FieldStatus.CONFIRMED})
+
+
+__all__ = [
+    "ACCIDENT_FIELD_KEYS",
+    "PERSON1_FIELD_KEYS",
+    "PERSON2_FIELD_KEYS",
+    "KNOWN_FIELD_KEYS",
+    "FieldCandidate",
+    "FieldStatus",
+    "validate_candidate",
+    "Clock",
+    "SystemClock",
+    "validate_accident_time",
+]