Merge branch 'main' into fix-bugfix0731

## 变更说明

- below_minimum 归入成功，保留原始业务结局
- store_closed/store_not_found/items_not_found/no_delivery/unsupported 归入失败
- running 保持进行中生命周期状态
- 历史细分状态迁移为三态终态，迁移可逆
- 后台成功/失败筛选及汇总兼容迁移前历史值

## 验证

- 相关回归：48 passed
- Ruff：通过
- Alembic：单一 head

---------

Co-authored-by: guke <guke@wonderable.ai>
Co-authored-by: linkeyu <798648091@qq.com>
Co-authored-by: unknown <798648091@qq.com>
Reviewed-on: #209
Co-authored-by: linkeyu <linkeyu@wonderable.ai>
Co-committed-by: linkeyu <linkeyu@wonderable.ai>

alembic/versions/deepseek_v4_flash_price.py

@@ -1,211 +0,0 @@
-"""correct DeepSeek V4 Flash token price and frozen historical costs
-
-Revision ID: deepseek_v4_flash_price
-Revises: comparison_below_min_success
-"""
-from __future__ import annotations
-
-import json
-from collections.abc import Sequence
-from typing import Any
-
-import sqlalchemy as sa
-from sqlalchemy.dialects import postgresql
-
-from alembic import op
-
-revision: str = "deepseek_v4_flash_price"
-down_revision: str | Sequence[str] | None = "comparison_below_min_success"
-branch_labels: str | Sequence[str] | None = None
-depends_on: str | Sequence[str] | None = None
-
-_JSON = sa.JSON().with_variant(postgresql.JSONB(), "postgresql")
-_CONFIG_KEY = "llm_token_price"
-_MODEL = "deepseek-v4-flash"
-_OLD_INPUT_PRICE = 3.0
-_OLD_OUTPUT_PRICE = 15.0
-_NEW_INPUT_PRICE = 1.0
-_NEW_OUTPUT_PRICE = 2.0
-_CORRECTION_MARKER = "deepseek_v4_flash_price"
-_CONFIG_MARKER_KEY = "migration_deepseek_v4_flash_price"
-
-
-def _decode_object(value: Any) -> dict[str, Any] | None:
-    if isinstance(value, dict):
-        return value
-    if isinstance(value, str):
-        try:
-            parsed = json.loads(value)
-        except (TypeError, ValueError):
-            return None
-        return parsed if isinstance(parsed, dict) else None
-    return None
-
-
-def _model_tokens(calls: Any) -> tuple[int, int]:
-    if not isinstance(calls, list):
-        return 0, 0
-    input_tokens = 0
-    output_tokens = 0
-    for call in calls:
-        if not isinstance(call, dict) or call.get("error") or call.get("model") != _MODEL:
-            continue
-        usage = _decode_object(call.get("usage"))
-        if usage is None:
-            continue
-        input_tokens += int(usage.get("prompt_tokens") or 0)
-        output_tokens += int(usage.get("completion_tokens") or 0)
-    return input_tokens, output_tokens
-
-
-def _app_config_table() -> sa.TableClause:
-    return sa.table(
-        "app_config",
-        sa.column("key", sa.String(64)),
-        sa.column("value", _JSON),
-        sa.column("updated_at", sa.DateTime(timezone=True)),
-    )
-
-
-def _comparison_table() -> sa.TableClause:
-    return sa.table(
-        "comparison_record",
-        sa.column("id", sa.Integer),
-        sa.column("llm_calls", _JSON),
-        sa.column("llm_cost_yuan", sa.Float),
-        sa.column("llm_price_snapshot", _JSON),
-    )
-
-
-def _update_config(conn, *, upgrade: bool) -> None:
-    table = _app_config_table()
-    row = conn.execute(
-        sa.select(table.c.value).where(table.c.key == _CONFIG_KEY)
-    ).mappings().first()
-    if row is None or not isinstance(row["value"], dict):
-        return
-
-    config = dict(row["value"])
-    per_model = dict(config.get("per_model") or {})
-    current = per_model.get(_MODEL)
-    if upgrade:
-        # Preserve an operator's explicit model price. The production defect is specifically
-        # the missing key falling through to the generic 3/15 price.
-        if current is not None:
-            return
-        per_model[_MODEL] = {
-            "input_per_1m": _NEW_INPUT_PRICE,
-            "output_per_1m": _NEW_OUTPUT_PRICE,
-        }
-        conn.execute(table.insert().values(key=_CONFIG_MARKER_KEY, value=True))
-    else:
-        marker_exists = conn.execute(
-            sa.select(table.c.key).where(table.c.key == _CONFIG_MARKER_KEY)
-        ).scalar_one_or_none()
-        if marker_exists is None:
-            return
-        if current != {
-            "input_per_1m": _NEW_INPUT_PRICE,
-            "output_per_1m": _NEW_OUTPUT_PRICE,
-        }:
-            conn.execute(table.delete().where(table.c.key == _CONFIG_MARKER_KEY))
-            return
-        per_model.pop(_MODEL, None)
-        conn.execute(table.delete().where(table.c.key == _CONFIG_MARKER_KEY))
-    config["per_model"] = per_model
-    conn.execute(
-        table.update()
-        .where(table.c.key == _CONFIG_KEY)
-        # 这是对历史误配置的追溯修正，不是从部署时刻开始的新价格。保留原 updated_at，
-        # 否则缺失成本回填会把部署前的记录全部排除。
-        .values(value=config)
-    )
-
-
-def _correct_frozen_costs(conn, *, upgrade: bool) -> None:
-    table = _comparison_table()
-    rows = conn.execute(
-        sa.select(
-            table.c.id,
-            table.c.llm_calls,
-            table.c.llm_cost_yuan,
-            table.c.llm_price_snapshot,
-        ).where(
-            table.c.llm_cost_yuan.is_not(None),
-            table.c.llm_price_snapshot.is_not(None),
-        )
-    ).mappings()
-
-    for row in rows:
-        snapshot = row["llm_price_snapshot"]
-        if not isinstance(snapshot, dict):
-            continue
-        prices = snapshot.get("prices")
-        if not isinstance(prices, dict):
-            continue
-        model_price = prices.get(_MODEL)
-        if not isinstance(model_price, dict):
-            continue
-
-        if upgrade:
-            if not (
-                model_price.get("_source") == "default"
-                and model_price.get("input_per_1m") == _OLD_INPUT_PRICE
-                and model_price.get("output_per_1m") == _OLD_OUTPUT_PRICE
-            ):
-                continue
-        elif snapshot.get("pricing_correction") != _CORRECTION_MARKER:
-            continue
-
-        input_tokens, output_tokens = _model_tokens(row["llm_calls"])
-        if input_tokens == 0 and output_tokens == 0:
-            continue
-        if upgrade:
-            delta = (
-                input_tokens / 1_000_000 * (_OLD_INPUT_PRICE - _NEW_INPUT_PRICE)
-                + output_tokens / 1_000_000 * (_OLD_OUTPUT_PRICE - _NEW_OUTPUT_PRICE)
-            )
-            corrected_price = {
-                "input_per_1m": _NEW_INPUT_PRICE,
-                "output_per_1m": _NEW_OUTPUT_PRICE,
-                "_source": "per_model",
-            }
-            snapshot["pricing_correction"] = _CORRECTION_MARKER
-            new_cost = max(0.0, float(row["llm_cost_yuan"]) - delta)
-        else:
-            delta = (
-                input_tokens / 1_000_000 * (_OLD_INPUT_PRICE - _NEW_INPUT_PRICE)
-                + output_tokens / 1_000_000 * (_OLD_OUTPUT_PRICE - _NEW_OUTPUT_PRICE)
-            )
-            corrected_price = {
-                "input_per_1m": _OLD_INPUT_PRICE,
-                "output_per_1m": _OLD_OUTPUT_PRICE,
-                "_source": "default",
-            }
-            snapshot.pop("pricing_correction", None)
-            new_cost = float(row["llm_cost_yuan"]) + delta
-
-        updated_prices = dict(prices)
-        updated_prices[_MODEL] = corrected_price
-        updated_snapshot = dict(snapshot)
-        updated_snapshot["prices"] = updated_prices
-        conn.execute(
-            table.update()
-            .where(table.c.id == row["id"])
-            .values(
-                llm_cost_yuan=round(new_cost, 6),
-                llm_price_snapshot=updated_snapshot,
-            )
-        )
-
-
-def upgrade() -> None:
-    conn = op.get_bind()
-    _update_config(conn, upgrade=True)
-    _correct_frozen_costs(conn, upgrade=True)
-
-
-def downgrade() -> None:
-    conn = op.get_bind()
-    _correct_frozen_costs(conn, upgrade=False)
-    _update_config(conn, upgrade=False)

app/api/v1/feedback.py

@@ -54,11 +54,13 @@ def _app_status(db_status: str) -> str:
 
 
 def _record_out(fb) -> FeedbackRecordOut:
+    images = fb.images or []
     return FeedbackRecordOut(
         id=fb.id,
         content=fb.content,
         scene=getattr(fb, "scene", None),
-        images=fb.images or [],
+        images=images,
+        image_thumbnails=[media.feedback_thumbnail_url(url) for url in images],
         status=_app_status(fb.status),
         reject_reason=getattr(fb, "reject_reason", None),
         reward_coins=getattr(fb, "reward_coins", None),
@@ -84,7 +86,7 @@ async def submit_feedback(
     device_model: str = Form(default=""),
     rom_name: str = Form(default=""),
     android_version: str = Form(default=""),
-    images: list[UploadFile] = File(default=[]),
+    images: list[UploadFile] = File(default=[]),  # noqa: B008 - FastAPI dependency declaration
 ) -> FeedbackOut:
     content = content.strip()
     contact = contact.strip()

app/core/config_schema.py

@@ -227,11 +227,7 @@ CONFIG_DEFS: dict[str, dict[str, Any]] = {
     # 编辑框;set_value 不校验类型,嵌套 JSON 照存。
     "llm_token_price": {
         "default": {
-            "per_model": {
-                "qwen3.5-flash": {"input_per_1m": 0.8, "output_per_1m": 2.0},
-                # DashScope 华北 2 公网调用原价；必须显式配置，不能落到 3/15 的未知模型兜底价。
-                "deepseek-v4-flash": {"input_per_1m": 1.0, "output_per_1m": 2.0},
-            },
+            "per_model": {"qwen3.5-flash": {"input_per_1m": 0.8, "output_per_1m": 2.0}},
             "default": {"input_per_1m": 3.0, "output_per_1m": 15.0},
             "currency": "CNY", "unit": "per_1m_tokens",
         },

app/core/media.py

@@ -11,6 +11,7 @@
 from __future__ import annotations
 
 import json
+import logging
 import os
 import secrets
 import subprocess
@@ -18,8 +19,17 @@ from decimal import ROUND_HALF_UP, Decimal, InvalidOperation
 from pathlib import Path
 from typing import Any
 
+from PIL import Image, ImageOps
+
 from app.core.config import settings
 
+logger = logging.getLogger("shagua.media")
+
+_FEEDBACK_DIR = "feedback"
+_FEEDBACK_THUMB_DIR = "feedback_thumbs"
+_FEEDBACK_THUMB_MAX_PX = 256
+_FEEDBACK_THUMB_QUALITY = 78
+
 
 class MediaError(Exception):
     """上传文件不合法(类型/大小)。调用方转 400。"""
@@ -68,8 +78,86 @@ def save_avatar(user_id: int, data: bytes) -> str:
 
 
 def save_feedback_image(user_id: int, data: bytes) -> str:
-    """保存反馈截图,返回相对 URL(`/media/feedback/<file>`)。"""
-    return _save_image("feedback", user_id, data)
+    """保存反馈截图并预生成历史页缩略图，返回原图相对 URL。"""
+    url = _save_image(_FEEDBACK_DIR, user_id, data)
+    # 缩略图失败不影响反馈受理；读取缩略图 URL 时会按需重试并回退原图。
+    ensure_feedback_thumbnail(url)
+    return url
+
+
+def feedback_thumbnail_url(image_url: str) -> str:
+    """把反馈原图 URL 映射成确定的缩略图 URL，不在 records 接口内做图片解码。
+
+    上传文件名由服务端生成且不会覆盖；旧数据在客户端真正请求可见图片时按需补图。
+    """
+    paths = _feedback_thumbnail_paths(image_url)
+    return paths[2] if paths is not None else image_url
+
+
+def _feedback_thumbnail_paths(image_url: str) -> tuple[Path, Path, str] | None:
+    prefix = f"{settings.MEDIA_URL_PREFIX}/{_FEEDBACK_DIR}/"
+    if not image_url.startswith(prefix):
+        return None
+
+    filename = image_url.removeprefix(prefix)
+    # 只接受当前目录下的单个文件名，避免数据库脏数据造成路径穿越。
+    if not filename or Path(filename).name != filename:
+        return None
+
+    source = _media_dir(_FEEDBACK_DIR) / filename
+    thumb_name = f"{Path(filename).stem}.jpg"
+    thumb = _media_dir(_FEEDBACK_THUMB_DIR) / thumb_name
+    thumb_url = f"{settings.MEDIA_URL_PREFIX}/{_FEEDBACK_THUMB_DIR}/{thumb_name}"
+    return source, thumb, thumb_url
+
+
+def ensure_feedback_thumbnail(image_url: str) -> Path | None:
+    """确保缩略图存在并返回文件；生成失败时回退原图，供动态缩略图路由使用。"""
+    paths = _feedback_thumbnail_paths(image_url)
+    if paths is None:
+        return None
+    source, thumb, _ = paths
+    if thumb.is_file():
+        return thumb
+    if not source.is_file():
+        return None
+
+    temp = thumb.with_name(f".{thumb.name}.{secrets.token_hex(4)}.tmp")
+    try:
+        with Image.open(source) as opened:
+            image = ImageOps.exif_transpose(opened)
+            image.thumbnail(
+                (_FEEDBACK_THUMB_MAX_PX, _FEEDBACK_THUMB_MAX_PX),
+                Image.Resampling.LANCZOS,
+            )
+            if image.mode != "RGB":
+                image = image.convert("RGB")
+            image.save(
+                temp,
+                format="JPEG",
+                quality=_FEEDBACK_THUMB_QUALITY,
+                optimize=True,
+            )
+        os.replace(temp, thumb)
+        return thumb
+    except (Image.DecompressionBombError, OSError, ValueError):
+        logger.warning("生成反馈缩略图失败: %s", source, exc_info=True)
+        return source
+    finally:
+        temp.unlink(missing_ok=True)
+
+
+def feedback_thumbnail_file(filename: str) -> Path | None:
+    """由缩略图文件名找到原反馈图并按需生成，非法/不存在返回 None。"""
+    if not filename or Path(filename).name != filename or Path(filename).suffix.lower() != ".jpg":
+        return None
+    stem = Path(filename).stem
+    for ext in (".jpg", ".png", ".webp"):
+        original = _media_dir(_FEEDBACK_DIR) / f"{stem}{ext}"
+        if original.is_file():
+            original_url = f"{settings.MEDIA_URL_PREFIX}/{_FEEDBACK_DIR}/{original.name}"
+            return ensure_feedback_thumbnail(original_url)
+    return None
 
 
 def save_report_image(user_id: int, data: bytes) -> str:

app/main.py

@@ -10,7 +10,7 @@ from collections.abc import AsyncIterator
 from contextlib import asynccontextmanager
 from pathlib import Path
 
-from fastapi import FastAPI
+from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import FileResponse
 from fastapi.staticfiles import StaticFiles
@@ -44,6 +44,7 @@ from app.api.v1.tasks import router as tasks_router
 from app.api.v1.user import router as user_router
 from app.api.v1.wallet import router as wallet_router
 from app.api.v1.wxpay import router as wxpay_router
+from app.core import media
 from app.core.config import settings
 from app.core.cps_reconcile_worker import (
     start_cps_reconcile_worker,
@@ -82,6 +83,19 @@ setup_logging(debug=settings.APP_DEBUG)
 logger = logging.getLogger("shagua.main")
 
 
+class FeedbackMediaStaticFiles(StaticFiles):
+    """反馈原图/缩略图文件名不可变，可长期缓存，避免列表反复回源。"""
+
+    async def get_response(self, path: str, scope):
+        response = await super().get_response(path, scope)
+        media_path = path.replace("\\", "/").lstrip("/")
+        if response.status_code == 200 and media_path.startswith(
+            ("feedback/", "feedback_thumbs/")
+        ):
+            response.headers["Cache-Control"] = "public, max-age=31536000, immutable"
+        return response
+
+
 @asynccontextmanager
 async def lifespan(_: FastAPI) -> AsyncIterator[None]:
     # 提示而非强制建表:生产用 alembic upgrade head,本地 dev 也建议先跑一次 migration。
@@ -212,8 +226,24 @@ def download_apk() -> FileResponse:
     )
 
 
+@app.get(
+    f"{settings.MEDIA_URL_PREFIX}/feedback_thumbs/{{filename}}",
+    tags=["feedback"],
+    include_in_schema=False,
+)
+def feedback_thumbnail(filename: str) -> FileResponse:
+    """旧反馈图按首次可见请求补缩略图；新图上传时已预生成。"""
+    path = media.feedback_thumbnail_file(filename)
+    if path is None:
+        raise HTTPException(status_code=404, detail="图片不存在")
+    return FileResponse(
+        path,
+        headers={"Cache-Control": "public, max-age=31536000, immutable"},
+    )
+
+
 app.mount(
     settings.MEDIA_URL_PREFIX,
-    StaticFiles(directory=str(_media_root)),
+    FeedbackMediaStaticFiles(directory=str(_media_root)),
     name="media",
 )

app/schemas/feedback.py

@@ -33,6 +33,8 @@ class FeedbackRecordOut(BaseModel):
     # 比价反馈的问题场景(找错商品/优惠不对…);普通反馈为 None
     scene: str | None = None
     images: list[str] = Field(default_factory=list)
+    # 与 images 下标一一对应；生成失败时该项回退原图 URL，兼容历史数据。
+    image_thumbnails: list[str] = Field(default_factory=list)
     status: str
     reject_reason: str | None = None
     reward_coins: int | None = None

pyproject.toml

@@ -35,6 +35,9 @@ dependencies = [
     # multipart form (FastAPI 表单上传依赖)
     "python-multipart>=0.0.9",
 
+    # 用户反馈截图缩略图，避免 App 历史页为 48dp 小图下载数 MB 原图
+    "pillow>=11.0.0",
+
     # admin 后台账号密码 hash(用户侧是手机号+验证码登录,不需要密码;admin 才用)
     "bcrypt>=4.0.0",
 

tests/test_deepseek_price_migration.py

@@ -1,184 +0,0 @@
-from __future__ import annotations
-
-import importlib.util
-from datetime import datetime
-from pathlib import Path
-
-import sqlalchemy as sa
-
-
-def _load_migration():
-    path = (
-        Path(__file__).parents[1]
-        / "alembic"
-        / "versions"
-        / "deepseek_v4_flash_price.py"
-    )
-    spec = importlib.util.spec_from_file_location("deepseek_v4_flash_price", path)
-    assert spec is not None and spec.loader is not None
-    module = importlib.util.module_from_spec(spec)
-    spec.loader.exec_module(module)
-    return module
-
-
-def test_migration_follows_comparison_status_normalization():
-    migration = _load_migration()
-
-    assert migration.down_revision == "comparison_below_min_success"
-
-
-def test_migration_adds_price_and_corrects_only_mispriced_snapshot(monkeypatch):
-    migration = _load_migration()
-    engine = sa.create_engine("sqlite://")
-    metadata = sa.MetaData()
-    app_config = sa.Table(
-        "app_config",
-        metadata,
-        sa.Column("key", sa.String(64), primary_key=True),
-        sa.Column("value", sa.JSON, nullable=False),
-        sa.Column("updated_at", sa.DateTime),
-    )
-    comparison = sa.Table(
-        "comparison_record",
-        metadata,
-        sa.Column("id", sa.Integer, primary_key=True),
-        sa.Column("llm_calls", sa.JSON),
-        sa.Column("llm_cost_yuan", sa.Float),
-        sa.Column("llm_price_snapshot", sa.JSON),
-    )
-    metadata.create_all(engine)
-
-    calls = [
-        {
-            "model": "deepseek-v4-flash",
-            "error": None,
-            "usage": {"prompt_tokens": 12031, "completion_tokens": 125},
-        },
-        {
-            "model": "qwen3.5-flash",
-            "error": None,
-            "usage": {"prompt_tokens": 2416, "completion_tokens": 119},
-        },
-    ]
-    snapshot = {
-        "mode": "per_model",
-        "prices": {
-            "deepseek-v4-flash": {
-                "input_per_1m": 3.0,
-                "output_per_1m": 15.0,
-                "_source": "default",
-            },
-            "qwen3.5-flash": {
-                "input_per_1m": 0.2,
-                "output_per_1m": 2.0,
-                "_source": "per_model",
-            },
-        },
-    }
-
-    with engine.begin() as conn:
-        original_updated_at = datetime(2026, 7, 13, 18, 20, 10)
-        conn.execute(
-            app_config.insert().values(
-                key="llm_token_price",
-                value={
-                    "per_model": {
-                        "qwen3.5-flash": {
-                            "input_per_1m": 0.2,
-                            "output_per_1m": 2.0,
-                        }
-                    },
-                    "default": {"input_per_1m": 3.0, "output_per_1m": 15.0},
-                },
-                updated_at=original_updated_at,
-            )
-        )
-        conn.execute(
-            comparison.insert().values(
-                id=1,
-                llm_calls=calls,
-                llm_cost_yuan=0.038689,
-                llm_price_snapshot=snapshot,
-            )
-        )
-        monkeypatch.setattr(migration.op, "get_bind", lambda: conn)
-
-        migration.upgrade()
-
-        config = conn.execute(
-            sa.select(app_config.c.value).where(
-                app_config.c.key == "llm_token_price"
-            )
-        ).scalar_one()
-        assert config["per_model"]["deepseek-v4-flash"] == {
-            "input_per_1m": 1.0,
-            "output_per_1m": 2.0,
-        }
-        assert conn.execute(
-            sa.select(app_config.c.updated_at).where(
-                app_config.c.key == "llm_token_price"
-            )
-        ).scalar_one() == original_updated_at
-        corrected = conn.execute(sa.select(comparison)).mappings().one()
-        assert corrected["llm_cost_yuan"] == 0.013002
-        assert corrected["llm_price_snapshot"]["prices"]["deepseek-v4-flash"] == {
-            "input_per_1m": 1.0,
-            "output_per_1m": 2.0,
-            "_source": "per_model",
-        }
-        assert corrected["llm_price_snapshot"]["pricing_correction"] == (
-            "deepseek_v4_flash_price"
-        )
-
-        migration.downgrade()
-
-        reverted_config = conn.execute(
-            sa.select(app_config.c.value).where(
-                app_config.c.key == "llm_token_price"
-            )
-        ).scalar_one()
-        assert "deepseek-v4-flash" not in reverted_config["per_model"]
-        reverted = conn.execute(sa.select(comparison)).mappings().one()
-        assert reverted["llm_cost_yuan"] == 0.038689
-        assert "pricing_correction" not in reverted["llm_price_snapshot"]
-
-
-def test_downgrade_preserves_price_that_existed_before_upgrade(monkeypatch):
-    migration = _load_migration()
-    engine = sa.create_engine("sqlite://")
-    metadata = sa.MetaData()
-    app_config = sa.Table(
-        "app_config",
-        metadata,
-        sa.Column("key", sa.String(64), primary_key=True),
-        sa.Column("value", sa.JSON, nullable=False),
-        sa.Column("updated_at", sa.DateTime),
-    )
-    sa.Table(
-        "comparison_record",
-        metadata,
-        sa.Column("id", sa.Integer, primary_key=True),
-        sa.Column("llm_calls", sa.JSON),
-        sa.Column("llm_cost_yuan", sa.Float),
-        sa.Column("llm_price_snapshot", sa.JSON),
-    )
-    metadata.create_all(engine)
-
-    explicit_price = {"input_per_1m": 1.0, "output_per_1m": 2.0}
-    with engine.begin() as conn:
-        conn.execute(
-            app_config.insert().values(
-                key="llm_token_price",
-                value={
-                    "per_model": {"deepseek-v4-flash": explicit_price},
-                    "default": {"input_per_1m": 3.0, "output_per_1m": 15.0},
-                },
-            )
-        )
-        monkeypatch.setattr(migration.op, "get_bind", lambda: conn)
-
-        migration.upgrade()
-        migration.downgrade()
-
-        config = conn.execute(sa.select(app_config.c.value)).scalar_one()
-        assert config["per_model"]["deepseek-v4-flash"] == explicit_price

tests/test_llm_cost.py

@@ -108,10 +108,6 @@ def test_get_llm_prices_falls_back_to_default_then_uses_override():
         # 无 override → CONFIG_DEFS 默认(含 per_model / default)
         prices = get_llm_prices(db)
         assert "per_model" in prices and "default" in prices
-        assert prices["per_model"]["deepseek-v4-flash"] == {
-            "input_per_1m": 1.0,
-            "output_per_1m": 2.0,
-        }
         # 有 override → 用 DB 值
         app_config.set_value(
             db, "llm_token_price",