docs: add deployment guide, production config, and changelog
This commit is contained in:
@@ -72,6 +72,58 @@ python3 -m uvicorn server:app --host 127.0.0.1 --port 9190
|
|||||||
| SIDECAR_COOLDOWN_BASE | 30 | 冷却基础时长 (秒) |
|
| SIDECAR_COOLDOWN_BASE | 30 | 冷却基础时长 (秒) |
|
||||||
| SIDECAR_COOLDOWN_MAX | 600 | 冷却最大时长 (秒) |
|
| SIDECAR_COOLDOWN_MAX | 600 | 冷却最大时长 (秒) |
|
||||||
|
|
||||||
|
## Docker 部署
|
||||||
|
|
||||||
|
```bash
|
||||||
|
# 构建并启动
|
||||||
|
cd /home/vincent/sidecar-v2
|
||||||
|
docker compose up -d --build
|
||||||
|
|
||||||
|
# 查看日志
|
||||||
|
docker logs sidecar-v2 -f
|
||||||
|
|
||||||
|
# 重启
|
||||||
|
docker compose restart
|
||||||
|
```
|
||||||
|
|
||||||
|
### Nginx 反向代理
|
||||||
|
|
||||||
|
```nginx
|
||||||
|
# 参考 deploy/nginx-sidecar-v2.conf
|
||||||
|
# Dashboard: http://192.168.1.99:19190/dashboard
|
||||||
|
# API: http://192.168.1.99:19190/v1/chat/completions
|
||||||
|
# Health: http://192.168.1.99:19190/health
|
||||||
|
```
|
||||||
|
|
||||||
|
## 生产环境 (当前)
|
||||||
|
|
||||||
|
- **部署地址**: 192.168.1.99:9190 (Docker)
|
||||||
|
- **Nginx 代理**: 192.168.1.99:19190
|
||||||
|
- **Dashboard**: http://192.168.1.99:19190/dashboard
|
||||||
|
- **5 个 NVIDIA Primary Backends**
|
||||||
|
- **Git 仓库**: http://192.168.1.99:12299/vincent/sidecar-v2
|
||||||
|
- **BIZ-50 部署任务**: 严维序 (opengineer)
|
||||||
|
- **开发**: 徐聪 (costcodev)
|
||||||
|
|
||||||
|
## 已知问题 & 待办
|
||||||
|
|
||||||
|
- [ ] 配置 Fallback 池(至少 1 个非 NVIDIA provider)
|
||||||
|
- [ ] 清理 openclaw.json 中冗余的 NVIDIA API key
|
||||||
|
- [ ] 排查无效模型 404(nemotron-70b, nemotron-340b)
|
||||||
|
- [ ] 评估 RPM limit 调整降低冷却频率
|
||||||
|
|
||||||
|
## 变更记录
|
||||||
|
|
||||||
|
### v2.0.1 (2026-06-25) — BIZ-50 部署修复
|
||||||
|
- proxy.py: 修复路由路径重复问题 (v1/v1 → v1)
|
||||||
|
- proxy.py: 修复 _emergency_count 全局变量
|
||||||
|
- server.py: 添加 logging.basicConfig 启用 INFO 日志
|
||||||
|
|
||||||
|
### v2.0.0 — 初始版本
|
||||||
|
- 多池路由、429 冷却、RPM 限流
|
||||||
|
- Dashboard + Admin API
|
||||||
|
- API Key 加密存储
|
||||||
|
|
||||||
## 存储
|
## 存储
|
||||||
- SQLite (WAL 模式)
|
- SQLite (WAL 模式)
|
||||||
- 表:backends, backend_usage_logs, cooldown_events, backend_health, system_config, daily_stats
|
- 表:backends, backend_usage_logs, cooldown_events, backend_health, system_config, daily_stats
|
||||||
Reference in New Issue
Block a user