YG-Soft/X-Financial
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: 重构知识库系统，移除Hermes集成，增强RAG和同步功能

Browse Source 
主要变更:
- 移除Hermes智能体及相关回调服务
- 新增知识库RAG、同步、调度、规范化和索引任务服务
- 重构orchestrator服务，增强运行时聊天功能
- 更新前端聊天、政策制度、设置等页面样式和逻辑
- 更新expense_claims和document_intelligence服务
- 删除llm_wiki相关服务和测试文件
- 更新docker-compose配置和启动脚本
This commit is contained in:
caoxiaozhu
2026-05-17 08:38:41 +00:00
parent 212c935308
commit 68f663f2f4
308 changed files with 83729 additions and 13588 deletions
Download Patch File Download Diff File Expand all files Collapse all files

257
server/tests/test_user_agent_service.py
View File

@@ -83,7 +83,7 @@ def test_user_agent_sanitizes_model_thinking_blocks() -> None:
)
def test_user_agent_rejects_visible_reasoning_drafts() -> None:
def test_user_agent_rejects_visible_reasoning_drafts() -> None:
session_factory = build_session_factory()
with session_factory() as db:
service = UserAgentService(db)
@@ -93,10 +93,43 @@ def test_user_agent_rejects_visible_reasoning_drafts() -> None:
"用户问的是：住宿费怎么算？\n让我分析一下：\n1. 实体识别..."
)
is None
)
def test_user_agent_knowledge_prompt_enforces_knowledge_boundary() -> None:
session_factory = build_session_factory()
with session_factory() as db:
ontology = SemanticOntologyService(db).parse(
OntologyParseRequest(
query="住宿费标准是多少？",
user_id="pytest",
context_json={"session_type": "knowledge"},
)
)
service = UserAgentService(db)
messages = service._build_model_messages(
UserAgentRequest(
run_id=ontology.run_id,
user_id="pytest",
message="住宿费标准是多少？",
ontology=ontology,
tool_payload={"result_type": "knowledge_search", "hits": []},
),
citations=[],
suggested_actions=[],
risk_flags=[],
draft_payload=None,
fallback_answer="",
)
assert "只能依据 facts.tool_payload.hits、facts.knowledge_answer_evidence" in messages[0]["content"]
assert "不能用常识、外部知识或主观推断补齐缺失条件" in messages[0]["content"]
assert "不能只依赖排在最前面的片段" in messages[0]["content"]
assert "不能把第一列的数值直接套给后面的列名" in messages[0]["content"]
assert "knowledge_evidence_blocks" in messages[0]["content"]
assert '"knowledge_answer_evidence": []' in messages[1]["content"]
def test_user_agent_knowledge_prompt_enforces_llm_wiki_boundary() -> None:
def test_user_agent_knowledge_fallback_is_honest_and_personalized() -> None:
session_factory = build_session_factory()
with session_factory() as db:
ontology = SemanticOntologyService(db).parse(
@@ -107,7 +140,48 @@ def test_user_agent_knowledge_prompt_enforces_llm_wiki_boundary() -> None:
)
)
service = UserAgentService(db)
messages = service._build_model_messages(
answer = service._build_knowledge_search_answer(
UserAgentRequest(
run_id=ontology.run_id,
user_id="pytest",
message="住宿费标准是多少？",
ontology=ontology,
context_json={"name": "张三"},
tool_payload={
"result_type": "knowledge_search",
"hits": [{"title": "差旅费制度", "content": "住宿费标准正文"}],
},
),
citations=[],
)
assert answer.startswith("张三，您好。")
assert "答案整理阶段本轮没有及时返回" in answer
assert "先给你当前最直接的依据" in answer
assert "《差旅费制度》" in answer
def test_user_agent_knowledge_answer_generation_uses_fast_timeouts(monkeypatch) -> None:
session_factory = build_session_factory()
with session_factory() as db:
ontology = SemanticOntologyService(db).parse(
OntologyParseRequest(
query="住宿费标准是多少？",
user_id="pytest",
context_json={"session_type": "knowledge"},
)
)
service = UserAgentService(db)
captured: dict[str, object] = {}
def fake_complete(messages, **kwargs):
captured["messages"] = messages
captured.update(kwargs)
return "测试回答"
monkeypatch.setattr(service.runtime_chat_service, "complete", fake_complete)
answer = service._generate_answer_with_model(
UserAgentRequest(
run_id=ontology.run_id,
user_id="pytest",
@@ -122,14 +196,177 @@ def test_user_agent_knowledge_prompt_enforces_llm_wiki_boundary() -> None:
fallback_answer="",
)
assert "只能依据 tool_payload.hits 中的 LLM Wiki 内容作答" in messages[0]["content"]
assert answer == "测试回答"
assert captured["timeout_seconds"] == 5
assert captured["slot_timeouts"] == {"main": 3, "backup": 5}
assert captured["max_attempts"] == 1
def test_user_agent_prefers_structured_knowledge_hit_for_answer_generation() -> None:
selected = UserAgentService._select_knowledge_model_hits(
{
"hits": [
{"content": "raw hit 1"},
{"content": "raw hit 2"},
{"content": "# 问答线索补充\n\n- 第二章 报销时限：费用发生后 30 日内提交申请。"},
{"content": "# 结构化表格补充\n\n| 项目 | 金额 |"},
]
}
)
assert selected[0]["content"].startswith("# 问答线索补充")
assert selected[1]["content"].startswith("# 结构化表格补充")
assert selected[2]["content"] == "raw hit 1"
def test_user_agent_uses_fast_knowledge_answer_without_model(monkeypatch) -> None:
session_factory = build_session_factory()
with session_factory() as db:
ontology = SemanticOntologyService(db).parse(
OntologyParseRequest(
query="报销时限是多少？",
user_id="pytest",
context_json={"session_type": "knowledge"},
)
)
service = UserAgentService(db)
monkeypatch.setattr(
service,
"_generate_answer_with_model",
lambda *args, **kwargs: (_ for _ in ()).throw(AssertionError("model should not be called")),
)
response = service.respond(
UserAgentRequest(
run_id=ontology.run_id,
user_id="pytest",
message="报销时限是多少？",
ontology=ontology,
context_json={
"name": "张三",
"session_type": "knowledge",
"user_input_text": "报销时限是多少？",
},
tool_payload={
"result_type": "knowledge_search",
"hits": [
{
"title": "费用报销制度",
"content": (
"# 问答线索补充\n\n"
"- 第二章 报销时限：员工应在费用发生后 30 日内提交报销申请。\n"
"- 第二章 报销时限：超过 30 日需补充审批说明。"
),
},
],
},
)
)
assert response.answer.startswith("张三，您好。")
assert "当前能直接确认的是" in response.answer
assert "30 日内提交报销申请" in response.answer
assert "答案整理阶段本轮没有及时返回" not in response.answer
def test_user_agent_fast_knowledge_answer_renders_relevant_table_preview() -> None:
session_factory = build_session_factory()
with session_factory() as db:
ontology = SemanticOntologyService(db).parse(
OntologyParseRequest(
query="餐补标准是多少？",
user_id="pytest",
context_json={"session_type": "knowledge"},
)
)
service = UserAgentService(db)
answer = service._build_fast_knowledge_answer(
UserAgentRequest(
run_id=ontology.run_id,
user_id="pytest",
message="餐补标准是多少？",
ontology=ontology,
context_json={
"session_type": "knowledge",
"user_input_text": "餐补标准是多少？",
},
tool_payload={
"result_type": "knowledge_search",
"hits": [
{
"title": "费用报销制度",
"content": (
"# 结构化表格补充\n\n"
"## 表3 出差补贴标准\n\n"
"| 项目 | 港澳台 | 其他地区 | 国外 |\n"
"| --- | --- | --- | --- |\n"
"| 餐补 | 75 | 55 | 140 |\n"
"| 住宿补贴 | 35 | 35 | 35 |\n"
),
}
],
},
),
citations=[],
)
assert answer is not None
assert "| 项目 | 港澳台 | 其他地区 | 国外 |" in answer
assert "| 餐补 | 75 | 55 | 140 |" in answer
def test_user_agent_fast_knowledge_answer_notes_missing_location_grounding() -> None:
session_factory = build_session_factory()
with session_factory() as db:
ontology = SemanticOntologyService(db).parse(
OntologyParseRequest(
query="前往北京出差的报销标准是什么？",
user_id="pytest",
context_json={"session_type": "knowledge"},
)
)
service = UserAgentService(db)
answer = service._build_fast_knowledge_answer(
UserAgentRequest(
run_id=ontology.run_id,
user_id="pytest",
message="前往北京出差的报销标准是什么？",
ontology=ontology,
context_json={
"session_type": "knowledge",
"user_input_text": "前往北京出差的报销标准是什么？",
},
tool_payload={
"result_type": "knowledge_search",
"hits": [
{
"title": "费用报销制度",
"content": (
"# 结构化表格补充\n\n"
"## 表3 出差补贴标准\n\n"
"| 项目 | 港澳台 | 直辖市/特区/西藏 | 其他地区 |\n"
"| --- | --- | --- | --- |\n"
"| 餐补 | 75 | 65 | 55 |\n"
"| 基本补贴 | 35 | 35 | 35 |\n"
),
}
],
},
),
citations=[],
)
assert answer is not None
assert "没有直接写出“北京”对应的地区档位或映射关系" in answer
def test_user_agent_model_prompt_supports_contextual_personalization() -> None:
session_factory = build_session_factory()
with session_factory() as db:
ontology = SemanticOntologyService(db).parse(
OntologyParseRequest(
session_factory = build_session_factory()
with session_factory() as db:
ontology = SemanticOntologyService(db).parse(
OntologyParseRequest(
query="我能坐什么舱位？",
user_id="pytest",
)
@@ -159,8 +396,8 @@ def test_user_agent_model_prompt_supports_contextual_personalization() -> None:
system_prompt = messages[0]["content"]
user_prompt = messages[1]["content"]
assert "user_grade" in system_prompt
assert "conversation_history" in system_prompt
assert "context.user_grade" in system_prompt
assert "conversation_history" in user_prompt
assert '"user_name": "张三"' in user_prompt
assert '"user_position": "财务分析师"' in user_prompt
assert '"user_grade": "P5"' in user_prompt