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).
303 lines
9.0 KiB
Python
303 lines
9.0 KiB
Python
"""LangGraph JRXML 生成代理的状态图定义。"""
|
|
|
|
import functools
|
|
import os
|
|
from typing import Literal
|
|
|
|
from dotenv import load_dotenv
|
|
from langgraph.graph import StateGraph, END
|
|
|
|
from agent.state import AgentState
|
|
from agent.nodes import (
|
|
load_session_node,
|
|
process_input,
|
|
manage_context,
|
|
save_state_snapshot,
|
|
classify_intent,
|
|
retrieve,
|
|
generate,
|
|
generate_skeleton,
|
|
refine_layout,
|
|
map_fields,
|
|
modify_jrxml,
|
|
handle_consult,
|
|
handle_undo,
|
|
handle_reset,
|
|
save_session_node,
|
|
validate,
|
|
explain_error,
|
|
correct_jrxml,
|
|
finalize,
|
|
)
|
|
from backend.logger import get_logger
|
|
|
|
load_dotenv()
|
|
MAX_RETRY = int(os.getenv("MAX_RETRY", "3"))
|
|
|
|
_graph_log = get_logger("agent")
|
|
|
|
|
|
def _log_route(route_name: str):
|
|
"""装饰器：自动记录路由决策。"""
|
|
def decorator(func):
|
|
@functools.wraps(func)
|
|
def wrapper(state: AgentState, *args, **kwargs):
|
|
target = func(state, *args, **kwargs)
|
|
_graph_log.info(
|
|
f"[路由] {route_name} → {target}",
|
|
extra={
|
|
"route": route_name,
|
|
"target": target,
|
|
"session_id": state.get("session_id", ""),
|
|
"intent": state.get("intent", ""),
|
|
"status": state.get("status", ""),
|
|
"has_jrxml": bool(state.get("current_jrxml", "").strip()),
|
|
"retry_count": state.get("retry_count", 0),
|
|
},
|
|
)
|
|
return target
|
|
return wrapper
|
|
return decorator
|
|
|
|
# ============================================================
|
|
# 路由函数
|
|
# ============================================================
|
|
|
|
@_log_route("route_by_intent")
|
|
def route_by_intent(state: AgentState) -> Literal[
|
|
"retrieve", "modify_jrxml", "save_session",
|
|
"handle_consult", "handle_undo", "handle_reset"
|
|
]:
|
|
"""根据 classify_intent 的结果路由到对应的处理节点。"""
|
|
intent = state.get("intent", "initial_generation")
|
|
|
|
if intent == "initial_generation":
|
|
return "retrieve"
|
|
elif intent == "modify_report":
|
|
return "modify_jrxml"
|
|
elif intent in ("preview_report", "export_pdf", "export_jrxml"):
|
|
return "save_session"
|
|
elif intent == "consult_question":
|
|
return "handle_consult"
|
|
elif intent == "undo_modification":
|
|
return "handle_undo"
|
|
elif intent == "reset_session":
|
|
return "handle_reset"
|
|
else:
|
|
# 兜底：根据是否有报表判断
|
|
if state.get("current_jrxml"):
|
|
return "modify_jrxml"
|
|
return "retrieve"
|
|
|
|
|
|
@_log_route("route_after_retrieve")
|
|
def route_after_retrieve(state: AgentState) -> Literal["generate", "generate_skeleton"]:
|
|
"""当 layout_schema 存在时走三层精确生成，否则走原有 1-shot。"""
|
|
layout_schema = state.get("layout_schema")
|
|
if layout_schema and isinstance(layout_schema, dict) and layout_schema.get("total_rows", 0) > 0:
|
|
return "generate_skeleton"
|
|
return "generate"
|
|
|
|
|
|
@_log_route("route_after_generate")
|
|
def route_after_generate(state: AgentState) -> Literal["save_session"]:
|
|
return "save_session"
|
|
|
|
|
|
@_log_route("route_after_modify")
|
|
def route_after_modify(state: AgentState) -> Literal["save_session"]:
|
|
return "save_session"
|
|
|
|
|
|
@_log_route("route_after_undo")
|
|
def route_after_undo(state: AgentState) -> Literal["save_session"]:
|
|
return "save_session"
|
|
|
|
|
|
@_log_route("route_after_save")
|
|
def route_after_save(state: AgentState) -> Literal["validate", "finalize"]:
|
|
# 预览/导出意图跳过验证，直接完成
|
|
intent = state.get("intent", "")
|
|
if intent in ("preview_report", "export_pdf", "export_jrxml"):
|
|
return "finalize"
|
|
return "validate"
|
|
|
|
|
|
@_log_route("route_after_validate")
|
|
def route_after_validate(state: AgentState) -> Literal["finalize", "explain_error"]:
|
|
if state.get("status") == "pass":
|
|
return "finalize"
|
|
return "explain_error"
|
|
|
|
|
|
@_log_route("route_after_explain")
|
|
def route_after_explain(state: AgentState) -> Literal["correct_jrxml"]:
|
|
return "correct_jrxml"
|
|
|
|
|
|
@_log_route("route_after_correct")
|
|
def route_after_correct(state: AgentState) -> Literal["validate", "finalize"]:
|
|
retry = state.get("retry_count", 0)
|
|
if retry >= MAX_RETRY:
|
|
return "finalize"
|
|
return "validate"
|
|
|
|
|
|
# ============================================================
|
|
# 图构建
|
|
# ============================================================
|
|
|
|
def build_graph() -> StateGraph:
|
|
workflow = StateGraph(AgentState)
|
|
|
|
# 现有节点
|
|
workflow.add_node("load_session", load_session_node)
|
|
workflow.add_node("process_input", process_input)
|
|
workflow.add_node("manage_context", manage_context)
|
|
workflow.add_node("save_session", save_session_node)
|
|
workflow.add_node("retrieve", retrieve)
|
|
workflow.add_node("generate", generate)
|
|
workflow.add_node("modify_jrxml", modify_jrxml)
|
|
workflow.add_node("validate", validate)
|
|
workflow.add_node("explain_error", explain_error)
|
|
workflow.add_node("correct_jrxml", correct_jrxml)
|
|
workflow.add_node("finalize", finalize)
|
|
|
|
# 新增节点：意图识别
|
|
workflow.add_node("save_state_snapshot", save_state_snapshot)
|
|
workflow.add_node("classify_intent", classify_intent)
|
|
workflow.add_node("handle_consult", handle_consult)
|
|
workflow.add_node("handle_undo", handle_undo)
|
|
workflow.add_node("handle_reset", handle_reset)
|
|
|
|
# 新增节点：分层精确生成（阶段一~三）
|
|
workflow.add_node("generate_skeleton", generate_skeleton)
|
|
workflow.add_node("refine_layout", refine_layout)
|
|
workflow.add_node("map_fields", map_fields)
|
|
|
|
# ---- 入口和前置流程 ----
|
|
workflow.set_entry_point("load_session")
|
|
workflow.add_edge("load_session", "process_input")
|
|
workflow.add_edge("process_input", "manage_context")
|
|
workflow.add_edge("manage_context", "save_state_snapshot")
|
|
workflow.add_edge("save_state_snapshot", "classify_intent")
|
|
|
|
# ---- 意图路由 ----
|
|
workflow.add_conditional_edges(
|
|
"classify_intent",
|
|
route_by_intent,
|
|
{
|
|
"retrieve": "retrieve",
|
|
"modify_jrxml": "modify_jrxml",
|
|
"save_session": "save_session",
|
|
"handle_consult": "handle_consult",
|
|
"handle_undo": "handle_undo",
|
|
"handle_reset": "handle_reset",
|
|
},
|
|
)
|
|
|
|
# ---- 初始生成分支 ----
|
|
workflow.add_conditional_edges(
|
|
"retrieve",
|
|
route_after_retrieve,
|
|
{
|
|
"generate": "generate",
|
|
"generate_skeleton": "generate_skeleton",
|
|
},
|
|
)
|
|
# 原有 1-shot 路径
|
|
workflow.add_conditional_edges(
|
|
"generate",
|
|
route_after_generate,
|
|
{"save_session": "save_session"},
|
|
)
|
|
# 分层精确生成 3 阶段路径
|
|
workflow.add_edge("generate_skeleton", "refine_layout")
|
|
workflow.add_edge("refine_layout", "map_fields")
|
|
workflow.add_conditional_edges(
|
|
"map_fields",
|
|
route_after_generate,
|
|
{"save_session": "save_session"},
|
|
)
|
|
|
|
# ---- 修改分支 ----
|
|
workflow.add_conditional_edges(
|
|
"modify_jrxml",
|
|
route_after_modify,
|
|
{"save_session": "save_session"},
|
|
)
|
|
|
|
# ---- 撤销分支 ----
|
|
workflow.add_conditional_edges(
|
|
"handle_undo",
|
|
route_after_undo,
|
|
{"save_session": "save_session"},
|
|
)
|
|
|
|
# ---- 保存后进入验证 ----
|
|
workflow.add_conditional_edges(
|
|
"save_session",
|
|
route_after_save,
|
|
{"validate": "validate"},
|
|
)
|
|
|
|
# ---- 验证 → 修正循环 ----
|
|
workflow.add_conditional_edges(
|
|
"validate",
|
|
route_after_validate,
|
|
{"finalize": "finalize", "explain_error": "explain_error"},
|
|
)
|
|
workflow.add_conditional_edges(
|
|
"explain_error",
|
|
route_after_explain,
|
|
{"correct_jrxml": "correct_jrxml"},
|
|
)
|
|
workflow.add_conditional_edges(
|
|
"correct_jrxml",
|
|
route_after_correct,
|
|
{"validate": "validate", "finalize": "finalize"},
|
|
)
|
|
|
|
# ---- 咨询 / 重置 → 直接结束 ----
|
|
workflow.add_edge("handle_consult", "finalize")
|
|
workflow.add_edge("handle_reset", "finalize")
|
|
|
|
# ---- 结束 ----
|
|
workflow.add_edge("finalize", END)
|
|
|
|
return workflow.compile()
|
|
|
|
|
|
# ============================================================
|
|
# 初始状态
|
|
# ============================================================
|
|
|
|
def create_initial_state() -> AgentState:
|
|
return AgentState(
|
|
conversation_history=[],
|
|
current_jrxml="",
|
|
user_input="",
|
|
status="",
|
|
error_msg="",
|
|
natural_explanation="",
|
|
retry_count=0,
|
|
user_modification_request="",
|
|
final_jrxml="",
|
|
stage="initial_generation",
|
|
retrieved_context="",
|
|
full_conversation_history=[],
|
|
compressed_history="",
|
|
current_token_count=0,
|
|
session_id="",
|
|
session_name="",
|
|
created_at="",
|
|
updated_at="",
|
|
intent="",
|
|
history_states=[],
|
|
jrxml_versions=[],
|
|
last_error_case={},
|
|
pending_failure_context={},
|
|
layout_schema={},
|
|
ocr_elements=[],
|
|
)
|