panda
43a0542a11
3-phase pipeline to solve LLM prompt overflow from too many OCR elements:
Phase 1 (generate_skeleton): compressed layout schema → skeleton JRXML
Phase 2 (refine_layout): sampled coordinates → pixel-level position tuning
Phase 3 (map_fields): OCR field names → replace $F{field_N} placeholders
Only triggered when layout_schema.total_rows > 0 on initial_generation intent.
Text requests and all other intents are unaffected (zero behavior change).
54 lines
1.5 KiB
Python
54 lines
1.5 KiB
Python
"""LangGraph JRXML 生成代理工作流的状态定义。"""
|
|
|
|
from typing import TypedDict, List
|
|
|
|
|
|
class AgentState(TypedDict, total=False):
|
|
# 核心工作流字段
|
|
conversation_history: List[dict]
|
|
current_jrxml: str
|
|
user_input: str
|
|
status: str
|
|
error_msg: str
|
|
natural_explanation: str
|
|
retry_count: int
|
|
user_modification_request: str
|
|
final_jrxml: str
|
|
stage: str
|
|
retrieved_context: str
|
|
|
|
# 需求1：智能上下文压缩
|
|
full_conversation_history: List[dict]
|
|
compressed_history: str
|
|
current_token_count: int
|
|
|
|
# 需求2：多会话持久化
|
|
session_id: str
|
|
session_name: str
|
|
created_at: str
|
|
updated_at: str
|
|
|
|
# 需求3：意图识别
|
|
intent: str
|
|
history_states: List[dict]
|
|
|
|
# 需求4：JRXML 版本历史（用于下载历史版本）
|
|
jrxml_versions: List[dict]
|
|
|
|
# 需求5：错误自增长（记录修正前的状态，供 validate 节点判断是否入知识库）
|
|
last_error_case: dict
|
|
|
|
# 需求6：失败上下文传递 — 重试耗尽后暂存失败信息，下次用户输入时自动注入
|
|
pending_failure_context: dict
|
|
|
|
# 需求7：OCR 单据字段精确提取结果
|
|
ocr_extraction_result: dict
|
|
uploaded_file_path: str
|
|
|
|
# 需求8：图片批注检测（圈选/箭头标记）
|
|
annotation_result: dict
|
|
|
|
# 需求9：分层精确生成
|
|
layout_schema: dict # extract_layout_schema() 输出，列+区域结构
|
|
ocr_elements: list # OCR 原始行数据（用于阶段二坐标采样）
|