feat(server): 工程骨架后端——上传 → 解析 → 清单最小链路

- core:配置(prod 拒绝默认密钥 / dev 短信 / 假模型)、租户事务 + 行级安全策略、
  problem+json 错误码、信封加密与签名下载、对象存储、事务内入队、Valkey stream 事件流、权限点
- identity:手机号 HMAC 存储、注册、登录(挤下线需确认,API3)、服务端会话
- billing:额度流水只追加,预扣 / 结算 / 退回,一笔预扣只有一个结局
- document:分片上传、按文件头识别格式、Gotenberg 转换、文字层与稳定句子编号
- checklist:标注规则识别、模型抽取、三道校验、标注类条款规则兜底、重跑对齐
- ai:提示词注册表(extract.substantive v1 来自 S4 v2)、假模型、OpenAI 兼容供应商、ai_call 流水
- 迁移:data-model 基线、套餐、Procrastinate、会话状态函数
- 测试 30 个:租户隔离、额度、失败退回、对象加密、契约一致性
- Compose(含 migrate)、镜像、CI

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
YANG JIANKUAN
2026-11-03 18:48:10 +08:00
parent 5747a523bb
commit ad7fba6021
106 changed files with 8094 additions and 0 deletions

View File

@@ -0,0 +1,161 @@
"""最小链路:上传 → 解析 → 清单;以及额度的预扣、结算、退回(红线 9)。"""
from __future__ import annotations
import httpx
from tests.conftest import FIXTURES, User, balance, project_with_tender, register, run_parse, upload
async def test_upload_parse_and_checklist(user: User) -> None:
project_id, doc_id = await project_with_tender(user)
assert (await user.get(f"/projects/{project_id}")).json()["next_action"] == "wait_parse"
job = await run_parse(user, doc_id)
assert job["status"] == "succeeded", job
assert job["result"]["items_added"] == 3
items = (await user.get(f"/projects/{project_id}/checklist/items")).json()["items"]
formal = [i for i in items if i["zone"] == "formal"]
assert len(formal) == 3
for it in formal:
assert it["importance"] == "substantive" and "invalidity" in it["categories"]
(c,) = it["citations"]
assert c["page_no"] == 1 and c["quads"] and c["quote"].startswith("★")
assert set(c["sentence_keys"][0].split(":")[:1]) == {doc_id}
assert any("32GB" in i["citations"][0]["quote"] for i in formal)
checklist = (await user.get(f"/projects/{project_id}/checklist")).json()
assert [r["marker"] for r in checklist["marking_rules"]] == ["star"]
project = (await user.get(f"/projects/{project_id}")).json()
assert project["status"] == "drafting" and project["next_action"] == "confirm_checklist"
doc = (await user.get(f"/documents/{doc_id}")).json()
assert doc["status"] == "ready" and doc["page_count"] == 1 and doc["normalized_sha256"]
sentences = (await user.get(f"/documents/{doc_id}/pages/1/sentences")).json()["items"]
assert any(s["text"].startswith("★交换机") for s in sentences)
notes = (await user.get("/notifications")).json()
assert notes["unread_count"] == 1 and notes["items"][0]["kind"] == "parse_succeeded"
async def test_quota_hold_then_settle(user: User) -> None:
_, doc_id = await project_with_tender(user)
assert await balance(user) == 3
await run_parse(user, doc_id)
assert await balance(user) == 2
accounts = (await user.get("/quota")).json()["accounts"]
assert accounts[0]["period_used"] == 1 and accounts[0]["held"] == 0
async def test_failed_parse_refunds_quota(user: User) -> None:
"""没有文字层的 PDF(扫描件):现在报「需要 OCR」,额度退回,净扣点为 0(FR-M6-03)。"""
_, doc_id = await project_with_tender(user, "blank.pdf")
job = await run_parse(user, doc_id)
assert job["status"] == "failed"
assert job["error"] == {"code": "ocr_required", "message": job["error"]["message"], "refunded": True}
assert await balance(user) == 3
from sqlalchemy import text
from bid_doctor.core.db import TenantContext, tenant_tx
ctx = TenantContext(tenant_id=user.me["tenant"]["id"], user_id=user.me["user"]["id"])
async with tenant_tx(ctx) as s:
kinds = (
await s.execute(text("SELECT kind, amount FROM quota_entry WHERE kind <> 'grant' ORDER BY created_at"))
).all()
assert [tuple(k) for k in kinds] == [("hold", -1), ("release", 1)]
async def test_retry_after_failure_holds_again(user: User) -> None:
_, doc_id = await project_with_tender(user, "blank.pdf")
first = await run_parse(user, doc_id)
second = await run_parse(user, doc_id)
assert (first["attempt_no"], second["attempt_no"]) == (1, 2)
assert await balance(user) == 3
async def test_duplicate_parse_is_rejected(user: User) -> None:
_, doc_id = await project_with_tender(user)
assert (await user.post(f"/documents/{doc_id}/parse-jobs", json={"expected_charge": 1})).status_code == 201
r = await user.post(f"/documents/{doc_id}/parse-jobs", json={"expected_charge": 1})
assert r.status_code == 409 and r.json()["code"] == "job_already_running" and r.json()["job_id"]
assert await balance(user) == 2 # 只预扣了一次
async def test_billing_mismatch_is_rejected_without_charge(user: User) -> None:
_, doc_id = await project_with_tender(user)
r = await user.post(f"/documents/{doc_id}/parse-jobs", json={"expected_charge": 0})
assert r.status_code == 409 and r.json()["code"] == "billing_changed"
assert await balance(user) == 3
async def test_quota_exhausted(app_client: httpx.AsyncClient) -> None:
u = await register(app_client)
_, doc_id = await project_with_tender(u)
for _ in range(3):
await run_parse(u, doc_id)
r = await u.post(f"/documents/{doc_id}/parse-jobs", json={"expected_charge": 1})
assert r.status_code == 409 and r.json()["code"] == "quota_exhausted"
assert await balance(u) == 0
async def test_reparse_keeps_item_ids(user: User) -> None:
"""重新解析时清单项编号不变(架构 §3.8)。"""
project_id, doc_id = await project_with_tender(user)
await run_parse(user, doc_id)
before = {i["id"] for i in (await user.get(f"/projects/{project_id}/checklist/items")).json()["items"]}
job = await run_parse(user, doc_id)
assert job["result"]["items_updated"] == 3 and job["result"]["items_added"] == 0
after = {i["id"] for i in (await user.get(f"/projects/{project_id}/checklist/items")).json()["items"]}
assert before == after
async def test_free_plan_allows_one_project(user: User) -> None:
assert (await user.post("/projects", json={})).status_code == 201
r = await user.post("/projects", json={})
assert r.status_code == 409 and r.json()["code"] == "plan_limit_exceeded"
async def test_upload_rejects_unsupported_and_corrupt_files(user: User) -> None:
r = await upload(user, "a.txt", "这不是招标文件".encode())
assert r.status_code == 422 and r.json()["code"] == "file_format_unsupported"
r = await upload(user, "a.pdf", b"%PDF-1.7\nnot really a pdf")
assert r.status_code == 422 and r.json()["code"] == "file_corrupted"
r = await upload(user, "看起来是.doc", (FIXTURES / "tender-mini.pdf").read_bytes())
assert r.status_code == 200
assert r.json()["detected_format"] == "pdf" and r.json()["extension_mismatch"] is True
async def test_upload_size_limit(user: User) -> None:
r = await user.post(
"/uploads", json={"filename": "big.pdf", "size_bytes": 51 * 1024 * 1024, "purpose": "project_document"}
)
assert r.status_code == 409 and r.json()["code"] == "plan_limit_exceeded"
async def test_signed_download(user: User, app_client: httpx.AsyncClient) -> None:
_, doc_id = await project_with_tender(user)
await run_parse(user, doc_id)
url = (await user.get(f"/documents/{doc_id}/file", params={"variant": "normalized"})).json()["url"]
path = url.split("/api/v1", 1)[1]
r = await app_client.get(path)
assert r.status_code == 200 and r.content.startswith(b"%PDF")
tampered = path.replace("sig=", "sig=x")
assert (await app_client.get(tampered)).status_code == 404
async def test_objects_are_encrypted_at_rest(user: User) -> None:
import os
from pathlib import Path
_, doc_id = await project_with_tender(user)
await run_parse(user, doc_id)
root = Path(os.environ["BD_STORAGE_DIR"]) / "t" / user.tenant_id
files = [p for p in root.rglob("*") if p.is_file()]
assert files and not any(p.read_bytes().startswith(b"%PDF") for p in files)
plain = (FIXTURES / "tender-mini.pdf").read_bytes()[:64]
assert not any(plain in p.read_bytes() for p in files)