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

refactor(server): user_agent/steward/ocr 等服务重构并适配关联任务

Browse Source 
- user_agent 拆分 application/locations/knowledge/response/review 四个子模块，接入申请位置语义与关联草稿分支
- steward planner/runtime/slot/plan_builder 决策链路重构，travel_reimbursement_calculator/orchestrator_expense_query 适配
- ocr/document_preview/document_intelligence/receipt_folder 复用预览与资产缓存，expense_claim_draft_flow/application_handoff 适配
- pyproject.toml 新增依赖，paddleocr bootstrap 脚本与 server_start.sh 调整
- 更新差旅/交通/通信等财务规则表，同步 document_intelligence/ocr/receipt_folder/user_agent 等测试
This commit is contained in:
caoxiaozhu
2026-06-24 10:42:24 +08:00
parent 332f77389d
commit 0264a4b5b4
41 changed files with 1273 additions and 182 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
server/tests/test_agent_asset_service.py
View File

@@ -888,6 +888,34 @@ def test_travel_reimbursement_calculator_rejects_unrecognized_location() -> None
)
def test_travel_reimbursement_calculator_normalizes_location_mixed_with_business_content() -> None:
with build_session() as db:
db.add(
Employee(
employee_no="E9004",
name="混合地点员工",
email="mixed-location@example.com",
position="产品经理",
grade="P4",
)
)
db.commit()
result = TravelReimbursementCalculatorService(db).calculate(
TravelReimbursementCalculatorRequest(days=4, location="上海辅助国网仿生产服务器"),
CurrentUserContext(
username="mixed-location@example.com",
name="混合地点员工",
role_codes=[],
is_admin=False,
),
)
assert result.location == "上海市"
assert result.matched_city == "上海"
assert result.hotel_amount > 0
def test_agent_run_service_lists_seeded_trace_data() -> None:
with build_session() as db:
service = AgentRunService(db)

27
server/tests/test_document_intelligence.py
View File

@@ -84,6 +84,33 @@ def test_document_intelligence_prefers_train_ticket_for_railway_e_ticket_invoice
assert any(field.label == "金额" and field.value == "354元" for field in insight.fields)
def test_document_intelligence_train_ticket_uses_railway_merchant_not_invoice_title() -> None:
insight = build_document_insight(
filename="2月20_武汉-上海.pdf",
summary="电子发票（铁路电子客票）；发票监；统一 制",
text=(
"电子发票（铁路电子客票）\n"
"发票号码:26429165800002785705 湖北\n"
"开票日期:2026年05月18日\n"
"武汉站 G458 上海虹桥站\n"
"Wuhan Shanghaihongqiao\n"
"2026年02月20日 07:55开 06车01B号 二等座\n"
"票价:￥354.00\n"
"4201061987****1615 曹笑竹\n"
"电子客票号:6580061086021391007342026\n"
"购买方名称:曹笑竹 统一社会信用代码:\n"
"买票请到12306 发货请到95306\n"
"中国铁路祝您旅途愉快"
),
)
assert insight.document_type == "train_ticket"
fields = {field.label: field.value for field in insight.fields}
assert fields["商户"] == "中国铁路"
assert fields["金额"] == "354元"
assert fields["列车出发时间"] == "2026-02-20 07:55"
def test_document_intelligence_recovers_train_ticket_from_english_station_ocr_text() -> None:
insight = build_document_insight(
filename="2月20_武汉-上海.pdf",

160
server/tests/test_expense_claim_service.py
View File

@@ -28,6 +28,7 @@ from app.schemas.reimbursement import (
)
from app.services.agent_conversations import AgentConversationService
from app.services.budget import BudgetService
from app.services.document_preview import DocumentPreviewAssets
from app.services.expense_claim_attachment_storage import ExpenseClaimAttachmentStorage
from app.services.expense_claim_budget_flow import ExpenseClaimBudgetFlowMixin
from app.services.expense_claim_workflow_constants import (
@@ -3314,6 +3315,68 @@ def test_attachment_preview_resolves_legacy_filename_in_claim_item_directory(mon
assert filename == "legacy-ticket.pdf"
def test_attachment_pdf_preview_falls_back_to_source_when_render_fonts_missing(monkeypatch, tmp_path) -> None:
current_user = CurrentUserContext(
username="emp-1",
name="张三",
role_codes=[],
is_admin=False,
)
monkeypatch.setattr(ExpenseClaimAttachmentStorage, "root", lambda self: tmp_path)
with build_session() as db:
claim = build_claim(expense_type="train", location="上海")
db.add(claim)
db.commit()
attachment_dir = tmp_path / claim.id / claim.items[0].id
attachment_dir.mkdir(parents=True)
file_path = attachment_dir / "2月20_武汉-上海.pdf"
preview_path = attachment_dir / "2月20_武汉-上海.preview.png"
file_path.write_bytes(b"%PDF-1.7 fake")
preview_path.write_bytes(b"broken-preview")
claim.items[0].invoice_id = f"{claim.id}/{claim.items[0].id}/{file_path.name}"
db.commit()
storage = ExpenseClaimAttachmentStorage()
storage.write_meta(
file_path,
{
"file_name": file_path.name,
"storage_key": storage.to_storage_key(file_path),
"media_type": "application/pdf",
"previewable": True,
"preview_kind": "image",
"preview_storage_key": storage.to_storage_key(preview_path),
"preview_media_type": "image/png",
"preview_file_name": preview_path.name,
"preview_rendered_with": "pdftoppm-png-r160-poppler-data",
},
)
def fake_render_pdf_first_page(*, pdf_path, preview_path, timeout_seconds):
raise RuntimeError("Missing language pack for 'Adobe-GB1' mapping")
monkeypatch.setattr(DocumentPreviewAssets, "render_pdf_first_page", fake_render_pdf_first_page)
resolved_path, media_type, filename = ExpenseClaimService(db).get_claim_item_attachment_preview_content(
claim_id=claim.id,
item_id=claim.items[0].id,
current_user=current_user,
)
assert resolved_path == file_path
assert media_type == "application/pdf"
assert filename == file_path.name
refreshed_meta = storage.read_meta(file_path)
assert refreshed_meta["preview_kind"] == "pdf"
assert refreshed_meta["preview_storage_key"] == storage.to_storage_key(file_path)
assert refreshed_meta["preview_media_type"] == "application/pdf"
assert refreshed_meta["preview_file_name"] == file_path.name
assert refreshed_meta["preview_rendered_with"] == ""
def test_submit_claim_runs_ai_review_and_routes_to_direct_manager() -> None:
current_user = CurrentUserContext(
username="emp-submit@example.com",
@@ -5199,6 +5262,103 @@ def test_admin_delete_claim_unlinks_receipt_folder_items(monkeypatch, tmp_path)
get_settings.cache_clear()
def test_admin_delete_linked_reimbursement_resets_application_link_status() -> None:
admin_user = CurrentUserContext(
username="superadmin",
name="系统管理员",
role_codes=["admin"],
is_admin=True,
)
with build_session() as db:
application_claim = ExpenseClaim(
id="application-delete-linked-reimbursement",
claim_no="APP-DEL-LINKED-APPLICATION",
employee_name="张三",
department_name="交付部",
project_code="PRJ-A",
expense_type="travel_application",
reason="支撑国网仿生产环境部署",
location="上海",
amount=Decimal("3000.00"),
currency="CNY",
invoice_count=0,
occurred_at=datetime(2026, 6, 21, 22, 30, tzinfo=UTC),
submitted_at=datetime(2026, 6, 21, 22, 35, tzinfo=UTC),
status="approved",
approval_stage=APPLICATION_LINK_STATUS_STAGE,
risk_flags_json=[
{
"source": "manual_approval",
"event_type": "expense_application_approval",
"operator": "向万红",
"previous_approval_stage": DIRECT_MANAGER_APPROVAL_STAGE,
"next_status": "approved",
"next_approval_stage": APPLICATION_LINK_STATUS_STAGE,
"generated_draft_claim_id": "reimbursement-delete-linked-application",
"generated_draft_claim_no": "RDELETE01",
"created_at": "2026-06-21T22:45:00+00:00",
}
],
)
reimbursement_claim = ExpenseClaim(
id="reimbursement-delete-linked-application",
claim_no="RDELETE01",
employee_name="张三",
department_name="交付部",
project_code="PRJ-A",
expense_type="travel",
reason="支撑国网仿生产环境部署报销",
location="上海",
amount=Decimal("3000.00"),
currency="CNY",
invoice_count=1,
occurred_at=datetime(2026, 6, 21, 22, 30, tzinfo=UTC),
submitted_at=None,
status="draft",
approval_stage="待提交",
risk_flags_json=[
{
"source": "application_handoff",
"event_type": "expense_application_to_reimbursement_draft",
"application_claim_id": application_claim.id,
"application_claim_no": application_claim.claim_no,
}
],
)
db.add_all([application_claim, reimbursement_claim])
db.commit()
deleted = ExpenseClaimService(db).delete_claim(reimbursement_claim.id, admin_user)
assert deleted is not None
assert deleted.claim_no == "RDELETE01"
assert db.get(ExpenseClaim, reimbursement_claim.id) is None
db.refresh(application_claim)
assert application_claim.status == "approved"
assert application_claim.approval_stage == APPLICATION_LINK_STATUS_STAGE
approval_flag = next(
flag
for flag in application_claim.risk_flags_json
if isinstance(flag, dict) and flag.get("event_type") == "expense_application_approval"
)
assert "generated_draft_claim_id" not in approval_flag
assert "generated_draft_claim_no" not in approval_flag
sync_flag = next(
flag
for flag in application_claim.risk_flags_json
if isinstance(flag, dict) and flag.get("event_type") == "expense_application_reimbursement_deleted"
)
assert sync_flag["source"] == "application_link_sync"
assert sync_flag["severity"] == "info"
assert sync_flag["actionability"] == "system_trace"
assert sync_flag["deleted_reimbursement_claim_id"] == "reimbursement-delete-linked-application"
assert sync_flag["deleted_reimbursement_claim_no"] == "RDELETE01"
assert sync_flag["next_approval_stage"] == APPLICATION_LINK_STATUS_STAGE
def test_direct_manager_can_return_subordinate_claim_to_pending_submission() -> None:
current_user = CurrentUserContext(
username="manager-return@example.com",

25
server/tests/test_notification_states.py
View File

@@ -85,6 +85,31 @@ def test_notification_state_service_persists_user_scoped_read_and_hidden_state()
assert other_saved.states[0].hidden_at is None
def test_notification_state_storage_ready_runs_once_per_database_bind(monkeypatch) -> None:
with build_session() as db:
service = NotificationStateService(db)
user = CurrentUserContext(username="alice", name="Alice", role_codes=[], is_admin=False)
calls: list[object] = []
original_create_all = Base.metadata.create_all
def track_create_all(*args, **kwargs):
calls.append(kwargs.get("bind"))
return original_create_all(*args, **kwargs)
monkeypatch.setattr(Base.metadata, "create_all", track_create_all)
service.list_states(user)
service.list_states(user)
service.patch_states(
NotificationStateBatchPatch(
states=[NotificationStatePatch(notification_id="workbench:todo:EXP-002", read=True)]
),
user,
)
assert len(calls) == 1
def test_notification_state_endpoint_reads_and_updates_current_user_state() -> None:
client = build_client()
headers = {"x-auth-username": "alice", "x-auth-name": "Alice"}

168
server/tests/test_ocr_service.py
View File

@@ -5,19 +5,23 @@ import subprocess
from pathlib import Path
from app.core.config import get_settings
from app.services import document_preview
from app.services.ocr import OcrService
def test_ocr_runtime_installers_include_poppler_cjk_data() -> None:
def test_ocr_runtime_installers_include_cjk_safe_pdf_rendering_tools() -> None:
repo_root = Path(__file__).resolve().parents[2]
dependency_sources = [
repo_root / "docker-compose.yml",
repo_root / "docker-compose.full.yml",
repo_root / "server" / "scripts" / "bootstrap_paddleocr_mobile.sh",
repo_root / "server" / "scripts" / "bootstrap_paddleocr_gpu.sh",
]
for path in dependency_sources:
assert "poppler-data" in path.read_text(encoding="utf-8")
content = path.read_text(encoding="utf-8")
assert "poppler-data" in content
assert "mupdf-tools" in content
def test_ocr_service_uses_worker_runtime_and_keeps_unsupported_files_as_warnings(
@@ -163,6 +167,7 @@ def test_ocr_service_passes_configured_device_to_worker(
text: bool,
timeout: int,
check: bool,
env: dict[str, str] | None = None,
) -> subprocess.CompletedProcess[str]:
captured_commands.append(command)
return subprocess.CompletedProcess(
@@ -194,12 +199,12 @@ def test_ocr_service_converts_pdf_to_images_and_returns_image_preview(
monkeypatch,
tmp_path: Path,
) -> None:
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> list[Path]:
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> tuple[list[Path], bool]:
first = output_dir / "page-1.png"
second = output_dir / "page-2.png"
first.write_bytes(b"fake-page-1")
second.write_bytes(b"fake-page-2")
return [first, second]
return [first, second], True
def fake_invoke_worker(
self,
@@ -281,26 +286,143 @@ def test_ocr_service_converts_pdf_to_images_and_returns_image_preview(
assert recognized.lines[1].page_index == 1
def test_ocr_service_uses_pdf_text_layer_without_worker_runtime(
def test_ocr_service_rejects_pdf_ocr_when_rendered_image_fonts_are_broken(
monkeypatch,
tmp_path: Path,
) -> None:
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> list[Path]:
page = output_dir / "page-1.png"
page.write_bytes(b"fake-rendered-page")
return [page]
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> tuple[list[Path], bool]:
raise RuntimeError("PDF 转图片失败：检测到中文字体映射缺失，未生成可 OCR 的图片。")
def fail_resolve_python(self) -> str:
raise AssertionError("PDF 文本层可用时不应强制解析 OCR worker。")
def fail_invoke_worker(self, **kwargs) -> dict:
raise AssertionError("PDF 文本层可用时不应调用 OCR worker。")
def fake_invoke_worker(
self,
*,
python_bin: str,
worker_path: str,
input_paths: list[Path],
) -> dict:
raise AssertionError("PDF 转图片已确认丢中文时，不应继续调用 OCR worker。")
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
monkeypatch.setattr(OcrService, "_resolve_python_bin", fail_resolve_python)
monkeypatch.setattr(OcrService, "_resolve_python_bin", lambda self: "python")
monkeypatch.setattr(OcrService, "_resolve_worker_path", lambda self: "worker.py")
monkeypatch.setattr(OcrService, "_convert_pdf_to_images", fake_convert_pdf_to_images)
monkeypatch.setattr(OcrService, "_invoke_worker", fail_invoke_worker)
monkeypatch.setattr(OcrService, "_invoke_worker", fake_invoke_worker)
get_settings.cache_clear()
try:
result = OcrService().recognize_files(
[
("2月20_武汉-上海.pdf", b"%PDF-1.7 fake", "application/pdf"),
]
)
finally:
get_settings.cache_clear()
failed = result.documents[0]
assert failed.line_count == 0
assert failed.preview_kind == ""
assert failed.preview_data_url == ""
assert failed.warnings == ["PDF 转图片失败：检测到中文字体映射缺失，未生成可 OCR 的图片。"]
def test_ocr_pdf_conversion_tries_next_renderer_when_poppler_font_mapping_fails(
monkeypatch,
tmp_path: Path,
) -> None:
output_dir = tmp_path / "pages"
output_dir.mkdir()
calls: list[str] = []
def fake_run(
command: list[str],
*,
capture_output: bool,
text: bool,
timeout: int,
check: bool,
) -> subprocess.CompletedProcess[str]:
calls.append(Path(command[0]).name)
if Path(command[0]).name == "pdftoppm":
(output_dir / "page-1.png").write_bytes(b"broken-preview")
return subprocess.CompletedProcess(
args=command,
returncode=0,
stdout="",
stderr="Syntax Error: Missing language pack for 'Adobe-GB1' mapping",
)
(output_dir / "page-1.png").write_bytes(b"rendered-with-chinese")
return subprocess.CompletedProcess(
args=command,
returncode=0,
stdout="",
stderr="",
)
monkeypatch.setattr(
document_preview.shutil,
"which",
lambda name: f"/usr/bin/{name}" if name in {"pdftoppm", "mutool"} else None,
)
monkeypatch.setattr(subprocess, "run", fake_run)
pages, preview_usable = OcrService()._convert_pdf_to_images(
pdf_path=tmp_path / "ticket.pdf",
output_dir=output_dir,
)
assert pages == [output_dir / "page-1.png"]
assert preview_usable is True
assert calls == ["pdftoppm", "mutool"]
def test_ocr_service_invokes_worker_even_when_pdf_text_layer_is_usable(
monkeypatch,
tmp_path: Path,
) -> None:
calls = {"worker": 0}
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> tuple[list[Path], bool]:
page = output_dir / "page-1.png"
page.write_bytes(b"fake-rendered-page")
return [page], True
def fake_invoke_worker(
self,
*,
python_bin: str,
worker_path: str,
input_paths: list[Path],
) -> dict:
calls["worker"] += 1
return {
"engine": "paddleocr_mobile",
"model": "PP-OCRv5_mobile",
"documents": [
{
"input_path": str(input_paths[0]),
"engine": "paddleocr_mobile",
"model": "PP-OCRv5_mobile",
"text": "电子发票（铁路电子客票） 武汉站 上海虹桥站 G458 票价 ￥354.00",
"summary": "铁路电子客票",
"avg_score": 0.95,
"line_count": 1,
"page_count": 1,
"warnings": [],
"lines": [
{
"text": "电子发票（铁路电子客票） 武汉站 上海虹桥站 G458 票价 ￥354.00",
"score": 0.95,
"box": [[1, 2], [10, 2], [10, 8], [1, 8]],
}
],
}
],
}
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
monkeypatch.setattr(OcrService, "_resolve_python_bin", lambda self: "python")
monkeypatch.setattr(OcrService, "_resolve_worker_path", lambda self: "worker.py")
monkeypatch.setattr(OcrService, "_convert_pdf_to_images", fake_convert_pdf_to_images)
monkeypatch.setattr(OcrService, "_invoke_worker", fake_invoke_worker)
monkeypatch.setattr(
OcrService,
"_extract_pdf_text_layer",
@@ -326,9 +448,9 @@ def test_ocr_service_uses_pdf_text_layer_without_worker_runtime(
recognized = result.documents[0]
assert result.success_count == 1
assert calls["worker"] == 1
assert recognized.document_type == "train_ticket"
assert "电子发票（铁路电子客票）" in recognized.text
assert "电子客票号:6580061086021391007342026" in recognized.text
assert any(field.label == "金额" and field.value == "354元" for field in recognized.document_fields)
assert recognized.preview_kind == "image"
assert recognized.preview_data_url.startswith("data:image/png;base64,")
@@ -392,14 +514,22 @@ def test_ocr_service_reuses_cached_document_for_same_content(
assert second.documents[0].summary == first.documents[0].summary
def test_ocr_cache_key_includes_pdf_render_pipeline_version() -> None:
cache_key = OcrService()._build_cache_key(b"same-pdf-content")
assert "pdf-image-ocr:" in cache_key
assert document_preview.DocumentPreviewAssets.PDF_RENDERER_ID in cache_key
assert "no-pdf-direct" in cache_key
def test_ocr_service_prefers_pdf_text_layer_when_rendered_ocr_is_placeholder_heavy(
monkeypatch,
tmp_path: Path,
) -> None:
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> list[Path]:
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> tuple[list[Path], bool]:
page = output_dir / "page-1.png"
page.write_bytes(b"fake-page")
return [page]
return [page], True
def fake_invoke_worker(
self,

258
server/tests/test_receipt_folder_service.py
View File

@@ -4,7 +4,7 @@ import base64
from app.api.deps import CurrentUserContext
from app.core.config import get_settings
from app.schemas.ocr import OcrRecognizeDocumentRead
from app.schemas.ocr import OcrRecognizeDocumentRead, OcrRecognizeFieldRead
from app.services.document_preview import DocumentPreviewAssets
from app.services.receipt_folder import ReceiptFolderService
@@ -72,6 +72,55 @@ def test_receipt_folder_train_ticket_uses_invoice_date_and_enriches_fields(monke
get_settings.cache_clear()
def test_receipt_folder_pdf_save_eagerly_renders_image_preview(monkeypatch, tmp_path) -> None:
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
get_settings.cache_clear()
try:
current_user = CurrentUserContext(
username="pytest",
name="Py Test",
role_codes=[],
is_admin=False,
)
def fake_render_pdf_first_page(*, pdf_path, preview_path, timeout_seconds):
preview_path.write_bytes(b"rendered-preview")
return preview_path
monkeypatch.setattr(DocumentPreviewAssets, "render_pdf_first_page", fake_render_pdf_first_page)
service = ReceiptFolderService()
receipt = service.save_receipt(
filename="2月20_武汉-上海.pdf",
content=b"%PDF-1.4 fake",
media_type="application/pdf",
current_user=current_user,
document=OcrRecognizeDocumentRead(
filename="2月20_武汉-上海.pdf",
media_type="application/pdf",
text="铁路电子客票 武汉 上海虹桥 354.00",
summary="铁路电子客票，武汉至上海虹桥。",
),
)
receipt_dir = next(service.root.glob("pytest/*"))
preview_path = receipt_dir / "preview.png"
meta = service._read_meta(receipt_dir)
assert receipt.preview_kind == "image"
assert preview_path.read_bytes() == b"rendered-preview"
assert meta["preview_file_name"] == "preview.png"
assert meta["preview_media_type"] == "image/png"
assert meta["preview_rendered_with"] == DocumentPreviewAssets.PDF_RENDERER_ID
resolved_path, media_type, file_name = service.resolve_preview(receipt.id, current_user)
assert resolved_path == preview_path
assert media_type == "image/png"
assert file_name == "preview.png"
finally:
get_settings.cache_clear()
def test_receipt_folder_pdf_preview_regenerates_stale_cached_image(monkeypatch, tmp_path) -> None:
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
get_settings.cache_clear()
@@ -123,6 +172,213 @@ def test_receipt_folder_pdf_preview_regenerates_stale_cached_image(monkeypatch,
get_settings.cache_clear()
def test_receipt_folder_pdf_preview_falls_back_to_source_when_render_fonts_missing(
monkeypatch,
tmp_path,
) -> None:
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
get_settings.cache_clear()
try:
current_user = CurrentUserContext(
username="pytest",
name="Py Test",
role_codes=[],
is_admin=False,
)
stale_preview = b"broken-preview"
preview_data_url = f"data:image/png;base64,{base64.b64encode(stale_preview).decode('ascii')}"
service = ReceiptFolderService()
receipt = service.save_receipt(
filename="2月20_武汉-上海.pdf",
content=b"%PDF-1.7 fake",
media_type="application/pdf",
current_user=current_user,
document=OcrRecognizeDocumentRead(
filename="2月20_武汉-上海.pdf",
media_type="application/pdf",
preview_kind="image",
preview_data_url=preview_data_url,
),
)
receipt_dir = next(service.root.glob("pytest/*"))
meta = service._read_meta(receipt_dir)
meta["preview_rendered_with"] = "pdftoppm-png-r160-poppler-data"
service._write_meta(receipt_dir, meta)
def fake_render_pdf_first_page(*, pdf_path, preview_path, timeout_seconds):
raise RuntimeError("Missing language pack for 'Adobe-GB1' mapping")
monkeypatch.setattr(DocumentPreviewAssets, "render_pdf_first_page", fake_render_pdf_first_page)
resolved_path, media_type, file_name = service.resolve_preview(receipt.id, current_user)
assert resolved_path == receipt_dir / "2月20_武汉-上海.pdf"
assert media_type == "application/pdf"
assert file_name == "2月20_武汉-上海.pdf"
refreshed_meta = service._read_meta(receipt_dir)
assert refreshed_meta["preview_kind"] == "pdf"
assert refreshed_meta["preview_file_name"] == "2月20_武汉-上海.pdf"
assert refreshed_meta["preview_media_type"] == "application/pdf"
assert refreshed_meta["preview_rendered_with"] == ""
finally:
get_settings.cache_clear()
def test_receipt_folder_train_ticket_extracts_passenger_from_id_line_and_purchase_name(
monkeypatch,
tmp_path,
) -> None:
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
get_settings.cache_clear()
try:
current_user = CurrentUserContext(
username="pytest",
name="Py Test",
role_codes=[],
is_admin=False,
)
service = ReceiptFolderService()
receipt = service.save_receipt(
filename="2月20_武汉-上海.pdf",
content=b"%PDF-1.4 fake",
media_type="application/pdf",
current_user=current_user,
document=OcrRecognizeDocumentRead(
filename="2月20_武汉-上海.pdf",
media_type="application/pdf",
text=(
"电子发票（铁路电子客票）\n"
"发票号码:26429165800002785705 湖北\n"
"开票日期:2026年05月18日\n"
"武汉站 G458 上海虹桥站\n"
"Wuhan Shanghaihongqiao\n"
"2026年02月20日 07:55开 06车01B号 二等座\n"
"票价:￥354.00\n"
"4201061987****1615 曹笑竹\n"
"电子客票号:6580061086021391007342026\n"
"购买方名称:曹笑竹 统一社会信用代码:\n"
"买票请到12306 发货请到95306\n"
"中国铁路祝您旅途愉快"
),
summary="电子发票（铁路电子客票）；发票监；统一 制",
document_type="train_ticket",
document_type_label="火车/高铁票",
scene_code="travel",
scene_label="差旅票据",
document_fields=[
OcrRecognizeFieldRead(key="merchant_name", label="商户", value="电子发票（铁路"),
OcrRecognizeFieldRead(key="amount", label="金额", value="354元"),
OcrRecognizeFieldRead(key="date", label="列车出发时间", value="2026-02-20 07:55"),
OcrRecognizeFieldRead(key="trip_no", label="车次", value="G458"),
OcrRecognizeFieldRead(key="route", label="行程", value="武汉-上海"),
],
),
)
assert receipt.merchant_name == "中国铁路"
detail = service.get_receipt(receipt.id, current_user)
fields = {field.label: field.value for field in detail.fields}
assert fields["商户"] == "中国铁路"
assert fields["乘车人"] == "曹笑竹"
assert fields["出发地点"] == "武汉"
assert fields["到达地点"] == "上海虹桥"
assert fields["身份证号"] == "4201061987****1615"
assert fields["电子客票号"] == "6580061086021391007342026"
assert fields["开票日期"] == "2026-05-18"
assert fields["列车出发时间"] == "2026-02-20 07:55"
assert fields["车厢"] == "06车"
assert fields["座位号"] == "01B"
finally:
get_settings.cache_clear()
def test_receipt_folder_train_ticket_repairs_invalid_generated_fields_from_ocr_text(
monkeypatch,
tmp_path,
) -> None:
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
get_settings.cache_clear()
try:
current_user = CurrentUserContext(
username="pytest",
name="Py Test",
role_codes=[],
is_admin=False,
)
service = ReceiptFolderService()
receipt = service.save_receipt(
filename="2月21日_上海-深圳.png",
content=b"fake image",
media_type="image/png",
current_user=current_user,
document=OcrRecognizeDocumentRead(
filename="2月21日_上海-深圳.png",
media_type="image/png",
text=(
"行程单示意\n"
"出票渠道：示例平台\n"
"非官方车票\n"
"不可报销\n"
"仅供演示\n"
"创建日期：2026年02月15日\n"
"订单号：DEMO202602210001\n"
"单据编号：DEMO-IT-000001\n"
"上海虹桥\n"
"G999\n"
"深圳北\n"
"站\n"
"站\n"
"Shanghaihongqiao\n"
"Shenzhenbei\n"
"2026年02月21日\n"
"08:30出发\n"
"全程约7小时30分\n"
"15:00到达\n"
"DEMO\n"
"乘客：示例旅客\n"
"车厢：05车\n"
"席别：二等座\n"
"-\n"
"扫码无效\n"
"证件号：310101199001010000\n"
"座位：08A\n"
"票价：￥438.00\n"
"仅为演示"
),
summary="行程单示意；出票渠道：示例平台；非官方车票",
document_type="train_ticket",
document_type_label="火车/高铁票",
scene_code="travel",
scene_label="差旅票据",
document_fields=[
OcrRecognizeFieldRead(key="amount", label="金额", value="438元"),
OcrRecognizeFieldRead(key="date", label="列车出发时间", value="2026-02-21 08:30"),
OcrRecognizeFieldRead(key="invoice_number", label="票据号码", value="DEMO202602210001"),
OcrRecognizeFieldRead(key="trip_no", label="车次", value="G999"),
OcrRecognizeFieldRead(key="route", label="行程", value="上海-深圳"),
OcrRecognizeFieldRead(key="departure_station", label="出发地点", value="二等座"),
OcrRecognizeFieldRead(key="arrival_station", label="到达地点", value="扫码无效"),
OcrRecognizeFieldRead(key="passenger_name", label="乘车人", value="席别二等座"),
],
),
)
detail = service.get_receipt(receipt.id, current_user)
fields = {field.label: field.value for field in detail.fields}
assert fields["出发地点"] == "上海虹桥"
assert fields["到达地点"] == "深圳北"
assert fields["乘车人"] == "示例旅客"
assert fields["身份证号"] == "310101199001010000"
assert fields["席别"] == "二等座"
assert fields["车厢"] == "05车"
assert fields["座位号"] == "08A"
assert fields["票价"] == "438.00元"
finally:
get_settings.cache_clear()
def test_receipt_folder_delete_removes_duplicate_marker(monkeypatch, tmp_path) -> None:
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
get_settings.cache_clear()

66
server/tests/test_user_agent_service.py
View File

@@ -15,6 +15,7 @@ from app.models.financial_record import ExpenseClaim
from app.schemas.ontology import OntologyParseRequest
from app.schemas.user_agent import UserAgentCitation, UserAgentRequest, UserAgentReviewRiskBrief
from app.services.agent_assets import AgentAssetService
from app.services.application_location_semantics import resolve_jieba_tokens
from app.services.ontology import SemanticOntologyService
from app.services.user_agent import UserAgentService
from app.services.user_agent_documents import UserAgentDocumentService
@@ -763,6 +764,67 @@ def test_user_agent_application_submit_blocks_overlapping_travel_dates() -> None
assert response.draft_payload is None
def test_user_agent_application_submit_normalizes_location_mixed_with_business_content() -> None:
session_factory = build_session_factory()
with session_factory() as db:
response = build_application_user_agent_response(
db,
"确认提交",
context_overrides={
"manager_name": "向万红",
"application_preview": {
"fields": {
"applicationType": "差旅费用申请",
"time": "2026-02-20 至 2026-02-23",
"location": "上海辅助国网仿生产服务器",
"reason": "辅助国网仿生产服务器部署",
"days": "4天",
"transportMode": "火车",
"amount": "2120元",
}
},
},
)
claim = application_claim_query(db).one()
assert claim.location == "上海市"
assert claim.reason == "辅助国网仿生产服务器部署"
assert "申请单据已生成" in response.answer
assert response.draft_payload is not None
def test_user_agent_application_submit_splits_location_and_reason_from_raw_sentence() -> None:
session_factory = build_session_factory()
with session_factory() as db:
response = build_application_user_agent_response(
db,
"确认提交",
history=[
{
"role": "user",
"content": "2026-02-20 至 2026-02-23，去上海辅助国网仿生产服务器部署，火车",
}
],
context_overrides={
"manager_name": "向万红",
"grade": "P5",
"department_name": "技术部",
},
)
claim = application_claim_query(db).one()
assert claim.location == "上海市"
assert claim.reason == "辅助国网仿生产服务器部署"
assert "申请单据已生成" in response.answer
def test_application_sentence_jieba_tokenizer_recognizes_location_boundary() -> None:
tokens = resolve_jieba_tokens("上海辅助国网仿生产服务器部署")
assert ("上海", "ns") in tokens
assert [word for word, _ in tokens] == ["上海", "辅助", "国网", "仿生产", "服务器", "部署"]
def test_user_agent_application_maps_preview_travel_type_label() -> None:
session_factory = build_session_factory()
with session_factory() as db:
@@ -2155,7 +2217,7 @@ def test_user_agent_returns_draft_limit_message_when_save_is_blocked() -> None:
context_json={"review_action": "save_draft"},
tool_payload={
"draft_limit_reached": True,
"message": "你当前已保存 3 个草稿，请先完成已保存的草稿，才能再次新建草稿。",
"message": "您当前已保存 3 个草稿，请先完成已保存的草稿，才能再次新建草稿。",
"status": "blocked",
},
)
@@ -2163,7 +2225,7 @@ def test_user_agent_returns_draft_limit_message_when_save_is_blocked() -> None:
assert (
response.answer
== "你当前已保存 3 个草稿，请先完成已保存的草稿，才能再次新建草稿。"
== "您当前已保存 3 个草稿，请先完成已保存的草稿，才能再次新建草稿。"
)