From 220901ff6198304b9b189ecb537e3067c9490a8d Mon Sep 17 00:00:00 2001
From: yumoqing <yumoqing@gmail.com>
Date: Fri, 11 Sep 2026 15:30:24 +0800
Subject: [PATCH] =?UTF-8?q?feat(migration):=20m0021=E5=88=A0llm=5Fusage=20?=
 =?UTF-8?q?token=E5=86=97=E4=BD=99=E5=88=97+m0025=E5=8A=A0project=5Fid?=
 =?UTF-8?q?=E5=88=97=E7=B4=A2=E5=BC=95(2026-09-10=E7=94=A8=E6=88=B7?=
 =?UTF-8?q?=E5=AE=9A=E5=A4=BA,=E9=85=8D=E5=AF=B9pipeline-llm@427cd7c)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- m0021: usages JSON全量不截断为计费唯一事实源,req_tokens/resp_tokens同源冗余删除,
  统计查询走JSON_EXTRACT+JSON_VALID守卫;backup_tables含llm_usage,dmig幂等(列不存在跳过)
- m0025: 按项目统计费用,project_id NOT NULL DEFAULT '0'(哨兵=非项目调用,对齐org_id平台约定),
  idx_llm_usage_project(project_id,created_at);token绑定project_id全链透传落库
- 两迁移id字段与文件名一致性已校验;llm_call_trace新表走create_tables.py动态扫描models建表,无需迁移
---
 .../m0021_llm_usage_drop_token_cols.json      | 13 +++++++++
 .../m0025_llm_usage_project_id.json           | 27 +++++++++++++++++++
 2 files changed, 40 insertions(+)
 create mode 100644 deploy/migrations/m0021_llm_usage_drop_token_cols.json
 create mode 100644 deploy/migrations/m0025_llm_usage_project_id.json

diff --git a/deploy/migrations/m0021_llm_usage_drop_token_cols.json b/deploy/migrations/m0021_llm_usage_drop_token_cols.json
new file mode 100644
index 0000000..ec5db37
--- /dev/null
+++ b/deploy/migrations/m0021_llm_usage_drop_token_cols.json
@@ -0,0 +1,13 @@
+{
+  "id": "m0021",
+  "title": "llm_usage 删除 req_tokens/resp_tokens 冗余列（2026-09-10 用户定夺：usages JSON 是计费唯一事实源，token 计数在其中；两列与 usages 同源冗余且曾被截断病根牵连，统计查询改 JSON_EXTRACT）。dmig 幂等守卫：列已不存在自动跳过",
+  "backup_tables": ["llm_usage"],
+  "up": [
+    {"op": "sql", "sql": "ALTER TABLE llm_usage DROP COLUMN req_tokens"},
+    {"op": "sql", "sql": "ALTER TABLE llm_usage DROP COLUMN resp_tokens"}
+  ],
+  "down": [
+    {"op": "sql", "sql": "ALTER TABLE llm_usage ADD COLUMN req_tokens INT NOT NULL DEFAULT 0 COMMENT '请求token(历史冗余列,计费看usages)'"},
+    {"op": "sql", "sql": "ALTER TABLE llm_usage ADD COLUMN resp_tokens INT NOT NULL DEFAULT 0 COMMENT '响应token(历史冗余列,计费看usages)'"}
+  ]
+}
diff --git a/deploy/migrations/m0025_llm_usage_project_id.json b/deploy/migrations/m0025_llm_usage_project_id.json
new file mode 100644
index 0000000..7034d2c
--- /dev/null
+++ b/deploy/migrations/m0025_llm_usage_project_id.json
@@ -0,0 +1,27 @@
+{
+  "id": "m0025",
+  "title": "llm_usage 加 project_id 列+索引（2026-09-10 用户定夺：按项目统计费用；非项目调用哨兵值'0'（NOT NULL DEFAULT，对齐org_id '0'平台约定）；查询不传project_id=不区分项目。token绑定的project_id沿llm_bridge→token→推理端点→govern_settle全链透传落库）。dmig 幂等守卫：列/索引已存在自动跳过（create_tables.py 也会按 models 自动补列）",
+  "backup_tables": [
+    "llm_usage"
+  ],
+  "up": [
+    {
+      "op": "sql",
+      "sql": "ALTER TABLE llm_usage ADD COLUMN project_id VARCHAR(32) NOT NULL DEFAULT '0' COMMENT '项目ID(sd_projects.id;''0''=非项目调用,查询不传=不区分项目)' AFTER ppid"
+    },
+    {
+      "op": "sql",
+      "sql": "CREATE INDEX idx_llm_usage_project ON llm_usage (project_id, created_at)"
+    }
+  ],
+  "down": [
+    {
+      "op": "sql",
+      "sql": "DROP INDEX idx_llm_usage_project ON llm_usage"
+    },
+    {
+      "op": "sql",
+      "sql": "ALTER TABLE llm_usage DROP COLUMN project_id"
+    }
+  ]
+}
\ No newline at end of file