mirror of
https://gitcode.com/JianFeeeee/ModelRouter.git
synced 2026-11-07 09:47:23 +00:00
docs: README 拆分为主题文档 + 中性化竞品表述
README 从 642 行降到 131 行，只保留简介、核心能力、最小快速开始、文档索引、 改名说明、目录结构与测试入口。其余内容迁到 docs/，中英文索引一一对应： docs/features.md 实现与完整特性清单（原 README 的"实现/特性"） docs/configuration.md 配置文件、自动探测模型、多密钥与配额 docs/routing.md 模型路由、AUTO 链、冷却与半冷却、源/适配器、源模板 docs/performance.md 实测内存占用与调优、thinking 预算 docs/webui.md WebUI 使用 docs/adapters.md Lua 适配器协议与内置适配器清单 docs/gui.md 桌面端 Headless/Desktop 两种产物与构建 docs/development.md 目录职责、开发协作、测试 docs/deployment.md 部署、发行包、目录权限、llmsproxy 迁移与验证清单 中性化：README / README_EN / GUI 注释里的 `Clash Verge` 类比全部改为 「桌面壳 / desktop shell」，仓库内（源码与文档）已无该表述。
This commit is contained in:
JianFeeeee
674
README.md
674
README.md
@ -4,103 +4,46 @@
|
||||
>
|
||||
> **AI-assisted**: 本项目使用 AI 辅助编程（代码与文档由 AI 协作完成，经人工复核）。
|
||||
|
||||
面向多 llms 订阅者，部署在内网，实现一次配置多个服务共同使用的效果。支持指定模型或 auto 模式，按照配置的优先级选择可用模型提供服务。
|
||||
面向多 LLM 订阅者的内网网关：一次配置，多个服务共用一组上游额度。支持指定模型或
|
||||
`AUTO` 模式，按配置的优先级选择可用模型提供服务。
|
||||
|
||||
> **轻量且增源无需重编译**：网关本体是单 Go 二进制（约 10MB，零运行时依赖）。新增/切换上游
|
||||
> 只需在 `config.yaml`（或 WebUI）加一个 `sources` 条目或挂一个 `.lua` 适配器——**不改 Go、
|
||||
> 不重编译**。源与适配器的改动经 WebUI 提交时即时生效（热更新）；直接编辑
|
||||
> `config.yaml` 或 `adapter_dir` 下的 `.lua` 文件则需要重启进程生效。
|
||||
网关本体是单个 Go 二进制（约 10 MB，零运行时依赖）。新增或切换上游只需在
|
||||
`config.yaml`（或 WebUI）加一个 `sources` 条目、或挂一个 `.lua` 适配器——**不改 Go、
|
||||
不重编译**。源与适配器的改动经 WebUI 提交时即时生效（热更新）；直接编辑
|
||||
`config.yaml` 或 `adapter_dir` 下的 `.lua` 文件需要重启进程生效。
|
||||
|
||||
## 核心优势
|
||||
## 核心能力
|
||||
|
||||
### 极致轻量
|
||||
|
||||
- **单二进制**：编译后 8~12MB（`-s -w` strip 后约 8MB），零运行时依赖（仅依赖系统 libc），部署即用
|
||||
- **低内存占用**：与源数量相关，非与运行时长相关——单源 ~10MB、16 源生产实例 ~32–35MB（[实测分解与调优](#内存占用实测与调优)）
|
||||
- **日志按需加载**：统计数字来自**全量**审计日志（启动时流式扫一遍即释放，29MB/22 万行约 260ms），但原始记录不常驻内存——默认只加载首屏，下滚自动分页，CSV 导出流式写出（O(1) 内存），离页即释放
|
||||
- **零运行时依赖**：纯 Go + LuaJIT 静态链接，无需安装 Python/Node/Java 等运行时
|
||||
- **启动极快**：冷启动 < 200ms，热重载配置 < 10ms
|
||||
|
||||
### 强大的多租户调度能力
|
||||
|
||||
- **多密钥多租户**：支持无限密钥，每个密钥独立角色、模型范围、Token 配额、重置周期
|
||||
- **按模型配额**：每把 key 的每个模型单独配 token + 请求数配额与重置周期（小时/周/月/自定义 N 小时）；一个模型用满只影响该模型，同 key 其它模型照常；耗尽返 429 + `Retry-After` 并点名模型，admin key 永不受限
|
||||
- **AUTO 智能调度**：基于优先级档位的分级调度，同优先级源自动轮询负载均衡，故障自动毫秒级故障转移
|
||||
- **自愈冷却**：冷却上限 5 分钟，过半后放行 1 个探测请求，上游/额度恢复即刻回归轮询，无需等满冷却窗口
|
||||
- **Token 配额管理**：精确到模型级别的 Token 配额控制，支持小时/周/月/自定义小时周期自动重置
|
||||
- **同优先级源负载均衡**：同一优先级档位的多个源，请求自动 Round-Robin 均匀分发，故障毫秒级故障转移
|
||||
- **Source-Model 前缀路由**：支持 `source-model`/`source:model`/`source/model` 精确指定上游源
|
||||
|
||||
### 生产级可靠性
|
||||
|
||||
- **热加载配置**：WebUI 修改源/密钥/适配器/AUTO链即时生效，无需重启
|
||||
- **密钥加密落盘**：AES-256-GCM 加密存储 `api_key`、网关密钥、请求头，master.key 0600 权限保护
|
||||
- **实时源探测**：`GET {base}/models` 定期探测，毫秒级感知上游状态，不污染调度退避状态
|
||||
- **全链路审计**：HTTP 访问日志、登录/配置变更审计、请求记录 CSV 导出、密钥用量统计 CSV 导出
|
||||
|
||||
### 灵活的协议适配
|
||||
|
||||
- **LuaJIT VM**：每个适配器独立 VM + **弹性 worker 池**（按需扩缩容，空闲时自动回收），安全并发，Lua 脚本热加载无需重启
|
||||
- **Lua 适配器协议**：`transform_request` / `transform_response` / `transform_stream_chunk` 双向转换
|
||||
- **动态请求头钩子**：`build_headers(meta)` 支持 HMAC 签名、动态 Header 注入
|
||||
- **多模态透传**：`image_url` 等多模态内容在多源间无损透传，Anthropic/Gemini/Ollama 自动转换
|
||||
|
||||
## 实现
|
||||
|
||||
统一 OpenAI 兼容网关：把多个上游 LLM 源（DeepSeek、Qijiar、OpenAI、Anthropic、Gemini、
|
||||
Groq、Mistral、Ollama、KimiCode…）通过 **Lua 适配器** 做协议转换，对内网暴露
|
||||
一个标准 `OpenAI Chat Completions` 接口（`/v1/chat/completions` + `/v1/models`），
|
||||
支持单次结算、SSE 流式、**AUTO 模型路由**与**多模态**透传。
|
||||
|
||||
从 [HomeAgent](https://gitcode.com/JianFeeeee/HomeAgent) 的多源 LLM 适配层
|
||||
（`internal/agent/api/provider.go` + `internal/lua/adapters/*`）抽离而来并独立演进。
|
||||
|
||||
## 特性
|
||||
|
||||
- **多源**：一个进程内配置任意多个上游源，按请求的 `model` 自动路由。
|
||||
- **AUTO 模式**：`default_model: AUTO` 时按优先级页保存的 AUTO 链档位逐档调度并发请求。
|
||||
- **统一输出**：所有源对外都是 OpenAI 格式（含 `reasoning_content`、`tool_calls`、`usage`）。
|
||||
- **生图**：`POST /v1/images/generations`，`kind: image` 的模型独立路由。
|
||||
- **多模态**：`content` 数组（`image_url` 等）在多源间无损透传，Anthropic/Gemini/Ollama 自动转换。
|
||||
- **LuaJIT VM**：基于 golua 绑定的 LuaJIT，每个适配器独立 VM + worker 池，安全并发。
|
||||
池是 **弹性** 的：`Σ max_concurrent` 只是上限而非预分配，状态按需创建、空闲时自动回收，
|
||||
因此空载网关几乎不持有 Lua 状态（启动 0 个）。
|
||||
- 扩容步长由适配器最大并发决定：`clamp(ceil(max/8), 1, 8)`，且仅在**真实并发争用**
|
||||
（所有现有状态均已占用）时批量预热，顺序流量始终只用 1 个状态。
|
||||
- 缩容步长由当前连接数决定：`clamp(ceil(冗余/(1+占用)), 1, 冗余)`——无连接时一轮收到
|
||||
常驻下限，忙时每轮只释放 1 个，保护热路径。WebUI「适配器」页展示实时池指标。
|
||||
- **thinking 预算**：请求里的推理开关/预算归一为中立的 `thinking:{enabled,budget_tokens}`，
|
||||
适配器映射到各自协议（Anthropic `thinking.budget_tokens`、Gemini `thinkingConfig`、
|
||||
DeepSeek `thinking.type="disabled"`）。
|
||||
- **Lua 适配协议**：每个源挂一个 `.lua` 适配器，完成 `transform_request` /
|
||||
`transform_response` / `transform_stream_chunk` 双向转换，协议差异全在 Lua 层。
|
||||
- **签名 / 请求头钩子**：适配器可定义 `build_headers(meta)`，在 Go 发 HTTP 前
|
||||
动态注入/签名请求头——用于云端 API 校验调用方 app（如 KimiCode 只放行特定
|
||||
agent）。提供 `hmac_sha256_hex` / `sha256_hex` / `base64_encode` 等签名辅助。
|
||||
- **WebUI**：内置管理界面，可在线查看/新增/编辑上游源与模型、配置 AUTO 优先级链、
|
||||
管理密钥模型范围，写入运行时文件持久化。
|
||||
- **密钥加密存储**：运行时文件中的上游 `api_key`、自定义请求头值、网关 key 均以
|
||||
AES-256-GCM 加密落盘（`master.key` 独立 0600，或 `LLMS_PROXY_MASTER_KEY`）。
|
||||
- **实时源探测**：状态页探测各源可达性（`GET {base}/models`，失败回退最小请求），
|
||||
不污染正常调度的退避状态，错误信息在 UI 可悬停查看。
|
||||
- **鉴权**：网关自身用 `gateway_keys` 校验客户端 Bearer key；与上游各自的 key 相互独立。
|
||||
- **流式**：SSE `chat.completion.chunk`，含角色首包与 `[DONE]` 收尾。
|
||||
- **全链路审计**：HTTP 访问日志、登录/配置变更审计、请求记录 CSV 导出、密钥用量统计 CSV 导出
|
||||
- **AES-256-GCM 加密存储**：运行时文件敏感字段加密落盘，master.key 0600 权限，支持环境变量注入主密钥
|
||||
- **统一接口**：对内暴露 OpenAI 兼容的 `/v1/chat/completions`、`/v1/models`、
|
||||
`/v1/images/generations`；所有源对外都是同一套格式（含 `reasoning_content`、
|
||||
`tool_calls`、`usage`）。
|
||||
- **多源 + AUTO 链**：一个进程内任意多个上游源；`default_model: AUTO` 时按优先级页
|
||||
保存的档位逐档调度，源故障自动降级并冷却，恢复后半冷却探测自愈。
|
||||
- **协议适配在 Lua 层**：每个源挂一个 `.lua` 适配器，完成请求/响应/流式分片的双向
|
||||
转换，并可用 `build_headers(meta)` 注入签名头。协议差异不进 Go 内核。
|
||||
- **插件独立**：业务扩展（渠道接入、计费、额度巡检）是外置 Lua 插件，由插件目录加载，
|
||||
不编进二进制、不随发行包分发；插件只通过 `modelrouter` SDK 与内核交互。
|
||||
- **多租户密钥**：网关 key 与上游 key 相互独立，按 key 限定模型范围、AUTO 链与配额。
|
||||
- **加密存储**：上游 key、自定义请求头、网关 key 均以 AES-256-GCM 落盘，
|
||||
`master.key` 单独 0600 权限（或由 `MODELROUTER_MASTER_KEY` 注入）。
|
||||
- **可观测**：状态页实时探测源可达性；HTTP 访问日志、配置变更审计、请求与用量 CSV 导出；
|
||||
写操作不落在请求热路径上。
|
||||
|
||||
## 快速开始
|
||||
|
||||
```bash
|
||||
GOMODCACHE=... GOPROXY=off go build -tags luajit -o llmsproxy ./cmd/llmsproxy
|
||||
./llmsproxy -config config.yaml # 首次运行自动生成默认配置并打印随机 admin key
|
||||
# 需要 LuaJIT 开发库；-tags luajit 是必须的
|
||||
GOMODCACHE=... GOPROXY=off go build -tags luajit -o modelrouter ./cmd/modelrouter
|
||||
|
||||
# 首次运行会生成默认配置，并在日志里打印随机 admin key
|
||||
./modelrouter -config config.yaml
|
||||
```
|
||||
|
||||
> 依赖 [golua](https://github.com/aarzilli/golua)（LuaJIT 绑定）。**必须**带 `-tags luajit`
|
||||
> 构建，否则默认走内置超集 gopher-lua 路径（行为略有差异）。
|
||||
> 依赖 [golua](https://github.com/aarzilli/golua)（LuaJIT 绑定）。不带 `-tags luajit`
|
||||
> 会走 gopher-lua 路径，行为略有差异。
|
||||
|
||||
> **配置不入库**：仓库不携带任何 config 文件（配置文件含密钥）。二进制首次运行时
|
||||
> 会在 `-config` 指定的位置生成默认配置：随机 admin key（打印在启动日志中）、
|
||||
> 仅绑定 `127.0.0.1:8080`。首次登录后请在 WebUI「密钥」页更换管理员密钥。
|
||||
> **配置不入库**：仓库不携带任何 config 文件（其中含密钥）。默认配置只绑定
|
||||
> `127.0.0.1:8080`，首次登录后请在 WebUI「密钥」页更换管理员密钥。
|
||||
|
||||
```bash
|
||||
# 无 key -> 401
|
||||
@ -117,526 +60,73 @@ curl -N -H "Authorization: Bearer $KEY" \
|
||||
http://127.0.0.1:8080/v1/chat/completions
|
||||
```
|
||||
|
||||
任何 OpenAI SDK 把 `base_url` 指到网关地址、`api_key` 用任意一个已创建并授权的网关 key 即可。
|
||||
任何 OpenAI SDK 把 `base_url` 指向网关、`api_key` 换成任一已授权的网关 key 即可。
|
||||
|
||||
## 配置
|
||||
## 文档
|
||||
|
||||
首次运行在 `-config` 路径自动生成默认配置（默认 `./config.yaml`），核心字段：
|
||||
| 主题 | 文档 |
|
||||
|---|---|
|
||||
| 实现与完整特性清单 | [docs/features.md](docs/features.md) |
|
||||
| 配置：`config.yaml`、自动探测模型、多密钥 | [docs/configuration.md](docs/configuration.md) |
|
||||
| 路由：模型路由、AUTO 链、冷却与半冷却、源与适配器、源模板 | [docs/routing.md](docs/routing.md) |
|
||||
| 适配器：Lua 协议、内置适配器清单 | [docs/adapters.md](docs/adapters.md) · [docs/lua-adapters.md](docs/lua-adapters.md)（[EN](docs/lua-adapters-en.md)） |
|
||||
| 插件：SDK 契约、钩子、UI 扩展、HTTP 端点 | [docs/plugins.md](docs/plugins.md) |
|
||||
| 桌面端：Headless / Desktop 两种产物、构建与运行 | [docs/gui.md](docs/gui.md) |
|
||||
| 性能：实测内存占用与调优、thinking 预算 | [docs/performance.md](docs/performance.md) |
|
||||
| WebUI 使用 | [docs/webui.md](docs/webui.md) |
|
||||
| 开发：目录职责、测试、分支与发布纪律 | [docs/development.md](docs/development.md) · [docs/git-workflow.md](docs/git-workflow.md)（[EN](docs/git-workflow-en.md)） |
|
||||
| 部署与迁移 | [docs/deployment.md](docs/deployment.md) |
|
||||
|
||||
```yaml
|
||||
listen: 127.0.0.1:8080 # 网关监听地址（建议绑内网/回环）
|
||||
gateway_keys: [sk-gw-<随机>] # 初始 admin 密钥（每次安装随机生成，仅首启写入运行时存储用）
|
||||
default_model: AUTO # model 无法路由时自动按优先级链选源
|
||||
adapter_dir: adapters # Lua 适配目录；目录不存在时首启创建并 seed 内置，存在则只读
|
||||
runtime_file: runtime.json # WebUI 编辑的源/密钥/AUTO 链持久化到此文件
|
||||
## 从 llmsproxy 改名
|
||||
|
||||
sources:
|
||||
- name: deepseek
|
||||
base_url: https://api.deepseek.com
|
||||
api_key: sk-... # 也可用 api_key_env: SOME_ENV 引用环境变量（不落盘明文）
|
||||
adapter: deepseek
|
||||
max_concurrent: 8
|
||||
models:
|
||||
- id: deepseek-v4-flash
|
||||
priority: 100 # YAML 源首次启动会 seed 进 AUTO 链的初值
|
||||
kind: chat
|
||||
- id: deepseek-v4-pro
|
||||
priority: 60
|
||||
kind: chat
|
||||
# 静态请求头（比适配器默认优先）
|
||||
headers: { X-Tenant: prod }
|
||||
# 传给 Lua build_headers 的透传元数据
|
||||
meta: { app_id: x, app_secret: y }
|
||||
temperature: 0.7
|
||||
max_tokens: 4096
|
||||
timeout: 120s # 请求超时，默认 120s
|
||||
```
|
||||
产品名早已是 ModelRouter，二进制、配置目录、服务名与环境变量现已统一为
|
||||
`modelrouter`：
|
||||
|
||||
| 旧 | 新 |
|
||||
|---|---|
|
||||
| `llmsproxy`（二进制） | `modelrouter` |
|
||||
| `/etc/llmsproxy/` | `/etc/modelrouter/` |
|
||||
| `llmsproxy.service` | `modelrouter.service` |
|
||||
| `LLMSPROXY_*` | `MODELROUTER_*` |
|
||||
|
||||
### 自动探测模型列表
|
||||
|
||||
不想手抄模型 id 时，只填 URL 和 key，让网关去扫：
|
||||
已部署的机器用迁移脚本升级（**默认 dry-run**，复制不移动，可回滚）：
|
||||
|
||||
```bash
|
||||
# 只探测，不写任何东西
|
||||
curl -X POST http://127.0.0.1:8080/api/sources/discover \
|
||||
-H "Authorization: Bearer sk-gw-…" -H "Content-Type: application/json" \
|
||||
-d '{"base_url":"https://api.vendor.com/v1","api_key":"sk-…"}'
|
||||
# → {"ok":true,"base_url":"…","endpoint":"","models":["gpt-x",…],
|
||||
# "image_models":["gpt-image-2"],"models_path":"/models","latency_ms":1387,…}
|
||||
|
||||
# 探测后直接落库
|
||||
curl -X POST http://127.0.0.1:8080/api/sources/discover/apply \
|
||||
-H "Authorization: Bearer sk-gw-…" -H "Content-Type: application/json" \
|
||||
-d '{"base_url":"https://api.vendor.com/v1","api_key":"sk-…","name":"vendor"}'
|
||||
sudo ./packaging/migrate-to-modelrouter.sh # 预演
|
||||
sudo ./packaging/migrate-to-modelrouter.sh --apply # 执行
|
||||
sudo ./packaging/migrate-to-modelrouter.sh --rollback --apply
|
||||
```
|
||||
|
||||
WebUI 源编辑框里也有「自动探测模型」按钮：填好 URL 和 key 点一下，列表自动
|
||||
填好，仍需点保存才生效。
|
||||
|
||||
几点值得知道：
|
||||
|
||||
- **URL 形态不用猜。** 厂商对「`/v1` 放 base 还是放 path」没有共识，探测会
|
||||
依次尝试若干候选并返回答对的那个，同时给出归一化后的 `base_url` +
|
||||
`endpoint`（`endpoint` 平时留空，因为 openai adapter 自带 `/chat/completions`）。
|
||||
- **`image_endpoint` 是另一套规则。** 网关的图像默认路径是**绝对**的
|
||||
`/v1/images/generations`，而 chat 默认是相对的。所以 `base_url` 带 `/v1` 时
|
||||
必须显式给 `image_endpoint: /images/generations`，否则图像请求 404。
|
||||
探测会替你填好。
|
||||
- **图像模型只看上游自己声明的 modality**，绝不从模型名猜。`gpt-image-2`
|
||||
这类 id 也出现在同时提供对话服务的目录上，猜错会让 AUTO 链选中一个只支持
|
||||
图像生成的模型，其后每一层都被它拖垮。
|
||||
- **不会逐模型探活。** 34 个模型的目录跑一轮要花钱且耗时数分钟，`/models`
|
||||
是足够便宜的信号；模型是否真能应答留给正常请求去发现。
|
||||
- **对已存在的源，必须显式 `replace_models: true`** 才会覆盖模型列表。重新
|
||||
探测顺手抹掉为 AUTO 链手工调过的 priority，是最难排查的一类数据丢失。
|
||||
- **`api_key` 留空表示沿用已存值**，不会把凭据清空。
|
||||
|
||||
|
||||
`config.yaml`、`runtime.json`、`master.key`、`adapters/` 均不提交版本库（见
|
||||
`.gitignore`），请随备份带走——`master.key` 丢失后运行时密文无法解密。
|
||||
|
||||
运行文件生成在 `-config` 所在目录：`adapter_dir` / `runtime_file` 自动指向
|
||||
配置文件旁边的 `adapters/` 与 `runtime.json`，从任意工作目录启动都可用。
|
||||
|
||||
运行时文件（`runtime_file`）中的敏感字段自动加密：
|
||||
|
||||
- 加密算法 AES-256-GCM，格式 `enc:v1:<base64>`。
|
||||
- 主密钥来源：环境变量 `LLMS_PROXY_MASTER_KEY`（64 位 hex）；否则读取
|
||||
`runtime_file` 同目录的 `master.key`；都不存在时首次启动自动生成 `master.key`（0600）。
|
||||
- 升级时旧明文文件自动兼容：首次运行正常读取，任何 UI 保存操作触发全文件加密迁移。
|
||||
- 注意：**master.key 丢失后密文无法解密**，请随配置一起备份；切勿提交到版本库。
|
||||
|
||||
### 网关密钥（多密钥）
|
||||
|
||||
网关鉴权采用「多密钥 + 角色 + 模型范围」架构，密钥持久化在 `runtime_file` 的
|
||||
`keys` 字段（加密存储）：
|
||||
|
||||
- `gateway_keys` 配置只是**初始 admin 密钥种子**：首次启动迁移为运行时 admin
|
||||
key，之后不再参与鉴权管理。
|
||||
- **重要：首次启动后请在 WebUI「密钥」页更换管理员密钥**——初始密钥明文写在
|
||||
`config.yaml` 里（每次安装随机生成），继续使用存在被盗风险；用新密钥登录后删除初始密钥。
|
||||
- WebUI **密钥页**可创建/删除密钥；每个密钥可指定 `admin`（管理全部）或
|
||||
`user`（仅看自己的 key）角色，并配置**模型范围**（模型 + 源 + token 配额 +
|
||||
重置周期）。
|
||||
- 客户端用任意一个已授权的密钥明文作为 Bearer（`Authorization: Bearer <key>`）。
|
||||
- 删除密钥即从运行时存储移除，立即失效。
|
||||
|
||||
#### 密钥用量配额
|
||||
|
||||
**配额按模型单独设置**：每个密钥的 `models[]` 里，每一条模型范围各自带一份
|
||||
token 配额与请求数配额。一个模型用满只影响该模型，同一密钥的其它模型照常工作。
|
||||
|
||||
```yaml
|
||||
keys:
|
||||
- key: sk-gw-<hex>
|
||||
role: user
|
||||
name: agent-alice
|
||||
models:
|
||||
- model: deepseek-v4-flash
|
||||
token_quota: 1000000 # 本周期内该 key 用这个模型的 token 预算
|
||||
req_quota: 20000 # 本周期内的请求次数
|
||||
period: nhour # "" | hour | week | month | nhour
|
||||
hours: 6 # 仅 nhour
|
||||
- model: AUTO
|
||||
token_quota: 5000000 # AUTO 也是一条独立配额
|
||||
period: day # ← 这种写法会被拒绝（词表只有 hour/week/month/nhour）
|
||||
- model: kimi-k3 # 未列配额 = 无限
|
||||
```
|
||||
|
||||
- **没有「整钥总配额」**：这是刻意的设计。整钥总额会让「把 A 模型的额度挪给
|
||||
B 模型」变成一次全局重分配；按模型独立计费则每个模型各自可控，运维可以
|
||||
看出哪个模型吃掉了预算。
|
||||
- `period` 词表：空 = 永不过期（累计总量），`hour` / `week` / `month` = 固定窗口，
|
||||
`nhour` + `hours` = 自定义小时数。**拼错的周期在写入时就被拒**，不会静默变成
|
||||
永不过期。
|
||||
- 配额严格按密钥隔离，且**同一密钥内按模型隔离**：A 密钥用满 `m1` 不会消耗
|
||||
B 密钥的额度，同一密钥的 `m2` 也不受影响。
|
||||
- 配额统计含聊天、流式、生图，跨重启从审计日志回放（保留 40 天，覆盖最长的
|
||||
month 窗口）。
|
||||
- 配额耗尽返回 **429 + `Retry-After`**（`rate_limit_exceeded`），消息里点名是哪个
|
||||
模型用满了，客户端可等窗口重置后自动恢复；模型越权才是 403。
|
||||
**admin 密钥永不受配额限制**（其 scope 上的配额也不强制），避免把管理员
|
||||
锁在门外。
|
||||
- 窗口用量按整点小时分桶统计，实际释放比配置窗口最多晚 1 小时（配额宁可晚释放
|
||||
也不超发）。
|
||||
- `PUT /api/keys/{key}` 提交 `models` 即同时提交它们的配额（配额就是 scope 的一部分，
|
||||
不存在会与模型列表脱节的第二份预算）。显式 `0` = 解除该模型的限制。
|
||||
|
||||
##### 配额拒绝 vs 容量拒绝：两种「拒绝」含义不同
|
||||
|
||||
配额在**入口**检查，不占用任何上游槽位，因此是廉价拒绝；容量不足则要走
|
||||
`busyWait` 有界等待后才返回 503。实测（容量 8、0.6s/请求、100 并发）：
|
||||
|
||||
| 情况 | 状态码 | 延迟 | 是否打到上游 |
|
||||
|---|---|---|---|
|
||||
| 配额用尽（`req_quota` 50 → 100 并发） | `429 rate_limit_exceeded` + `Retry-After` | **19–21ms** | 否 |
|
||||
| 容量打满（同一拓扑，未限配额） | `503 upstream_error` | ~2.6s（`busyWait`） | 超出部分否 |
|
||||
|
||||
客户端因此可以区分「等窗口重置」与「等上游腾出容量」：前者只需等待，后者
|
||||
通常该降并发或换源。
|
||||
|
||||
##### 配额的开销
|
||||
|
||||
配额桶按 (密钥, 模型, 整点小时) 分桶保留 40 天，实测（AMD 7840HS）：
|
||||
|
||||
- 窗口查询按窗口长度而非保留总量扫描：1h 窗口 49ns、24h 窗口 55ns、
|
||||
30d 窗口 3.9μs（此前全扫保留总量，960 桶时 5.9μs）。
|
||||
- 记录一条请求：283ns、3 分配（与引入配额前相同，分配来自 ring buffer）。
|
||||
- 内存：生产形态（7 密钥 × 8 模型 × 2 源 × 满 40 天 retention）约 **3.7MB**。
|
||||
按源 pin 的 `source::model` 桶**惰性创建**——只有当某条配额真的 pin 了
|
||||
某个源时才维护，否则每条记录多写一份桶，在 20 密钥 × 8 模型 × 3 源下会
|
||||
多占 18MB。
|
||||
代价：配置 pinned 配额之前的历史用量无法事后按源拆分，所以 **pinned 配额的
|
||||
首个窗口可能少算**。
|
||||
|
||||
```bash
|
||||
# 配额耗尽时客户端看到（点名了具体模型）
|
||||
HTTP/1.1 429 Too Many Requests
|
||||
Retry-After: 2100
|
||||
{"error":{"type":"rate_limit_exceeded","message":"token quota exceeded for \"deepseek-v4-flash\" (5000000/5000000)"}}
|
||||
```
|
||||
|
||||
### 模型路由
|
||||
|
||||
`/v1/chat/completions` 的 `model` 解析顺序：
|
||||
|
||||
1. `source/model` 或 `source:model` 前缀 → 指定源；
|
||||
2. 精确匹配某个源的 `model`；
|
||||
3. 配置 `default_model: AUTO` 时 → 走优先级页保存的 AUTO 链（见下）；
|
||||
4. 否则回落（返回错误）。
|
||||
|
||||
任何 OpenAI 客户端，只要 `model` 设为某个源的 `name/任意名`，即可锁定走该源；
|
||||
设为 `AUTO`（或网关配了 `default_model: AUTO`）即自动按优先级链选源。
|
||||
|
||||
### AUTO 链（优先级页）
|
||||
|
||||
AUTO 调度**只**由优先级页保存的规则决定，源配置里的 `models[].priority` 数字不再
|
||||
参与调度、也不再显示。**聊天与生图是两套独立的链**，在优先级页面通过「聊天 / 生图」
|
||||
按钮切换编辑：
|
||||
|
||||
- **聊天链**（`runtime_file` 的 `auto` 字段）：服务 `POST /v1/chat/completions`
|
||||
的 `model: AUTO` 请求；只接受 `kind: chat` 模型，配置在链里的生图槽位会被忽略。
|
||||
- **生图链**（`runtime_file` 的 `auto_image` 字段）：服务 `POST /v1/images/generations`
|
||||
的 `model: AUTO` 请求；只接受 `kind: image` 模型。未配置生图链时，AUTO 生图回落
|
||||
到「所有含生图模型的源、按注册顺序」的旧行为，兼容旧配置。
|
||||
|
||||
每条规则 = 一个「槽位」：`{ model, source, tier, token_quota, period, hours }`。
|
||||
|
||||
- `tier` 表示优先级档位：同一档的模型并排、共享该优先级；档位从上到下递减。
|
||||
- 同一模型可配置多个槽位（如 A 源低配 → B 源低配 → A 源高配 → B 源高配），按档位顺延。
|
||||
- `token_quota` > 0 时该槽位在重置周期内用满即顺延到下一槽位；`period` 支持
|
||||
`hour` / `week` / `month` / `nhour`（配合 `hours`），空 = 不限。
|
||||
- 同 tier 内按「偏好分」排序，失败槽位冷却后自动跳过；冷却 / 配额耗尽 / hard
|
||||
错误会顺延到下一 tier，全部失败时返回 503 并附上每档失败摘要。
|
||||
|
||||
### 冷却与半冷却探测（自愈调度）
|
||||
|
||||
槽位失败后进入指数退避冷却（5s → 10s → 20s …，上限 **5 分钟**）。冷却不是全时段
|
||||
硬闸：
|
||||
|
||||
- **前半段完全静默**：刚失败的槽位不接任何流量，避免对故障上游持续施压。
|
||||
- **后半段放行 1 个探测**：窗口过半后，同一 `(源, 模型)` 最多允许 **一个** 请求作为
|
||||
探测通过。探测槽数 = `clamp(max_concurrent / 10, 1, 2)`——`max_concurrent: 10`
|
||||
的源恰好放行 1 个探测请求。
|
||||
- **探测排在最后**：同 tier 内探测候选永远排在健康候选之后，只有没有正常槽可用时
|
||||
才承接真实流量，因此探测不会抢走可用容量。
|
||||
- **探测即真实请求**：成功则立即清零失败计数与冷却，槽位当场回到正常轮询——
|
||||
上游恢复/额度恢复后不必等满整个冷却窗口。失败则重开一个新窗口，下次探测顺延到
|
||||
新窗口的中点，不会连续重试。
|
||||
|
||||
分类冷却策略：
|
||||
|
||||
| 情况 | 冷却 | 说明 |
|
||||
|---|---|---|
|
||||
| 传输错误 / 5xx | 5s→10s→20s…，上限 5min | 指数退避 |
|
||||
| 401 / 403 凭据错误 | 10min | 换 key 后由下一次探测自动接回 |
|
||||
| 429 限流 | 30s 固定 | 「太快」不等于「坏了」，不进指数阶梯 |
|
||||
| 配额耗尽（`insufficient_quota` / `余额不足` 等） | 对齐配额窗口，上限 30min | **不进指数阶梯、不涨失败计数**，额度恢复即可用 |
|
||||
|
||||
WebUI「优先级」页的槽位健康标签会区分 `冷却` / `待探测` / `探测中`，悬停可看窗口
|
||||
起点、探测放行时刻与完全恢复时刻。
|
||||
|
||||
### 源（Source）与适配器（Adapter）的关系
|
||||
|
||||
- **源** 是到某个上游的连接描述：`name`、`base_url`、`api_key`、模型列表与优先级。
|
||||
- **适配器** 是协议转换逻辑（Lua 脚本）：把统一 OpenAI 格式请求转成上游原生格式，
|
||||
再把上游响应/流式分块转回统一格式。
|
||||
- 一个适配器可被多个源复用（如 `openai.lua` 同时服务多个 OpenAI 兼容站点）；
|
||||
同一个源也可以切换不同适配器（改 `adapter` 字段即可）。
|
||||
- 源决定“连谁、暴露哪些模型”，适配器决定“怎么对话”——二者在 `sources[]` 条目中
|
||||
通过 `adapter` 字段关联。
|
||||
|
||||
加载流程（`internal/core` 的 `Core` 负责装配）：
|
||||
|
||||
1. 启动时 `lua.NewVM(adapter_dir)` 加载全部适配器：内置适配器（编译期 embed）+
|
||||
`adapter_dir` 下的同名覆盖文件；内置文件写在代码内，覆盖文件要求更高优先级。
|
||||
2. `config.Load` 读取 `config.yaml`，`config.NewStore(runtime_file)` 读 WebUI 改动的
|
||||
运行时源，二者按名称合并成完整源列表。
|
||||
3. `rebuildRegistry` 为每个源创建 `provider.Provider`（持有目标适配器），并按源的
|
||||
并发上限配置适配器 worker 池**上限**（不预分配，见下）。
|
||||
4. 请求进来时 `Registry` 按 `model` 路由到 Provider，Provider 调适配器
|
||||
`transform_request` → HTTP 发送 → `transform_response` / `transform_stream_chunk`。
|
||||
|
||||
WebUI 上的"新增/编辑源"、"上传 Lua 适配器"、"改 AUTO 优先级链"与"密钥模型范围"
|
||||
都即时生效（写入运行时文件或 `adapter_dir` 后重新装配，无需重启）；直接编辑
|
||||
`config.yaml` / `adapter_dir` 下的文件则需要重启进程才会重新加载。
|
||||
|
||||
在 WebUI 删除源/适配器时：运行时（WebUI 创建）的源与 `.lua` 文件会真正移除；
|
||||
`config.yaml` 中定义的基源无法改写配置文件，采用删除标记隐藏（重启后仍隐藏），
|
||||
在 UI 里重新添加同名源即可恢复。
|
||||
|
||||
### 源模板（多 key 负载均衡）
|
||||
|
||||
多个上游 key 共享同一套配置（URL / 适配器 / 模型列表 / 并发 / rpm）时，逐一复制
|
||||
整段 source 太冗余。**源模板**保存除 `name` 和 `api_key` 以外的全部源字段，
|
||||
从模板一键创建多个仅 key 不同的源，天然支持多 key 负载均衡（每个展开源独立调度、
|
||||
独立冷却、独立健康状态）。
|
||||
|
||||
- WebUI「源」页右上角「**模板管理**」按钮：列出全部模板，可编辑 / 删除 / 新建。
|
||||
- 添加源弹窗右上角两个按钮：
|
||||
- **从模板创建** → 选模板 → 表单自动填充（name + key 仍需你填）；
|
||||
- **存为模板** → 把当前表单非 key/name 字段存为新模板。
|
||||
- 模板持久化在 `runtime_file` 的 `source_templates` 字段，与运行时源一样热管理，
|
||||
改完即时生效。模板本身不含密钥，可以安全共享 / 纳管。
|
||||
|
||||
模板不会自己变成源——它只是「配方」。真正承担流量的是从模板创建出来的、带 key
|
||||
的具体源。
|
||||
|
||||
### 内存占用（实测与调优）
|
||||
|
||||
内存不是一个定数：它随**配置的源数量**增长（每个源一个 `http.Transport` 与连接池、
|
||||
一组健康状态），与运行时长无关。本机实测（Linux x86_64，12 核）：
|
||||
|
||||
| 部署形态 | 启动 RSS | 稳态 RSS |
|
||||
|---|---|---|
|
||||
| 1 源 / 1 适配器（最小配置） | ~4 MB | ~10 MB |
|
||||
| 1 源 + 29 MB 历史审计日志 | ~19 MB | ~20 MB |
|
||||
| **16 源 / 13 适配器 / 59 模型（本机生产）** | ~28 MB | **~32–35 MB** |
|
||||
|
||||
> 历史参考：本项优化前同一生产配置为 **~105 MB**。降幅来自三处：审计日志的**原始记录**
|
||||
> 不再常驻内存（约 25 MB；统计聚合仍扫全量，但扫完即释放）、Lua 状态池不再单调增长、
|
||||
> 以及下面两个运行时开关。
|
||||
|
||||
内存构成（生产实例分段测量，`/proc/<pid>/smaps`）：
|
||||
|
||||
| 区域 | RSS | 说明 |
|
||||
|---|---|---|
|
||||
| Go 堆 | ~14 MB | provider/registry/scheduler 结构 + 连接池缓冲 |
|
||||
| 其他匹名（线程栈 / LuaJIT chunk / runtime） | ~12 MB | 与线程数、已加载适配器数相关 |
|
||||
| 二进制 text+rodata | ~8 MB | 映射的可执行文件页（只读、可被内核回收） |
|
||||
| Go runtime 预留 | ~4 MB | VSZ 上看到的 2 GB+ 是地址空间预留，不占物理内存 |
|
||||
| 共享库 | ~3 MB | libc / libluajit / libm |
|
||||
|
||||
**两个推荐的部署开关**（只适用于环境变量，无需改代码）：
|
||||
|
||||
```ini
|
||||
# /etc/systemd/system/llmsproxy.service
|
||||
Environment=GOGC=50
|
||||
Environment=MALLOC_ARENA_MAX=2
|
||||
```
|
||||
|
||||
- `MALLOC_ARENA_MAX=2`：LuaJIT 的分配走 cgo → glibc malloc，glibc 默认允许 `8×nproc`
|
||||
个 per-thread arena，每个碰到 malloc 的 OS 线程会占用一个（各约 1 MB，且**不归还给系统**）。
|
||||
实测从 8–15 个 arena（约 7–12 MB）降到 **0**。
|
||||
- `GOGC=50`：把 Go 堆增长目标减半。**单独使用无效**（省下的堆会立即被更多 glibc
|
||||
arena 吃掉，实测 20.3 → 21.5 MB 反而变大），必须与 `MALLOC_ARENA_MAX` 配合，
|
||||
两者同时开启才降 **~19%**。网关是 I/O 密集型（本机 9 小时仅消耗 1min10s CPU），
|
||||
多出的 GC 周期与它的空闲 CPU 相比可忽略。
|
||||
|
||||
> 桌面版（Electron）已在 `cmd/gui/main.js` 里默认为内嵌核心注入这两个开关；手动 export 同名
|
||||
> 环境变量可覆盖。如果你用自己的 systemd unit / 容器，建议照上面加上。
|
||||
|
||||
可选：`GOMEMLIMIT=48MiB` 作软上限，实测再省 ~1 MB；代价是逐近上限时 GC 转激进，
|
||||
温和场景不必开。
|
||||
|
||||
### thinking 预算
|
||||
|
||||
请求体带 `"thinking": {"enabled": false}` 时关闭推理；`"thinking": {"enabled": true,
|
||||
"budget_tokens": 4096}` 指定预算。内核只定义这个**中立**语义字段，映射到具体协议由
|
||||
适配器负责（Anthropic `thinking.budget_tokens`、Gemini `thinkingConfig.thinkingBudget`、
|
||||
DeepSeek `thinking.type = "disabled"`）。内核不再向适配器泄漏 `disable_thinking` /
|
||||
`extra_body` 这类私有字段。
|
||||
|
||||
### WebUI
|
||||
|
||||
内置管理界面（`GET /`），登录后可在浏览器完成：
|
||||
|
||||
- **状态页**：源在线状态（实时探测 + 悬停看错误）、模型/源/key 用量统计、请求记录，
|
||||
支持按时间范围导出 CSV；点击模型可生成 pin 到该模型的连接配置。
|
||||
- **对话页**：流式/非流式调试。
|
||||
- **密钥页**：创建/编辑网关 key，为每个 key 配模型范围（模型 + 源 + token 配额 + 周期），
|
||||
管理员管理全部 key，用户只看到自己的 key。每个模型砖块显示自己的配额徽标
|
||||
（如 `1.0K 77×·1h`，未设配额显示 `∞`），点开可编辑该模型的 token 配额、
|
||||
请求数配额与重置周期 —— 配额按模型独立生效，一个用满不影响同一 key 的其它模型。
|
||||
「我的密钥」页对用户展示本 key 的模型范围与各自配额。
|
||||
- **优先级页**：拖拽积木配置 AUTO 链档位。
|
||||
- **源页**：在线增删改上游源（API key 等敏感字段加密落盘）。
|
||||
- **适配器页**：上传 / 删除 Lua 适配器脚本。
|
||||
|
||||
改动写入 `runtime_file`（重启仍生效）。
|
||||
|
||||
## Lua 适配器协议
|
||||
|
||||
完整 API 见 **[Lua 适配器 API 文档](docs/lua-adapters.md)**（[English](docs/lua-adapters-en.md)）。
|
||||
|
||||
每个适配器是一个返回 table 的 Lua 脚本（`internal/lua/adapters/<name>.lua`），
|
||||
加载时可被 `adapter_dir` 下的同名脚本覆盖——**改 Lua 脚本无需重编译**，
|
||||
重启即生效；通过 WebUI 上传的适配器与在线编辑的源配置则即时生效。
|
||||
|
||||
```lua
|
||||
return {
|
||||
name = "mysrc",
|
||||
version = "1.0.0",
|
||||
endpoint = "/chat/completions", -- 上游路径（可被 source.endpoint 覆盖）
|
||||
headers = { ["X-Static"] = "v" }, -- 静态默认请求头（build_headers 缺省时使用）
|
||||
|
||||
-- 请求转换：把 OpenAI 格式 req 转成上游原生格式，返回字符串
|
||||
transform_request = function(raw_json) ... end,
|
||||
|
||||
-- 响应转换：把上游原生响应转成统一格式字符串
|
||||
-- { content, reasoning_content, finish_reason, token_usage{...}, tool_calls[{...}] }
|
||||
transform_response = function(raw_json) ... end,
|
||||
|
||||
-- 流式分块转换：把上游 SSE data 转成 { content, done, ... }，返回 "" 则跳过
|
||||
transform_stream_chunk = function(raw_chunk) ... end,
|
||||
|
||||
-- [可选] 动态请求头/签名钩子
|
||||
-- meta = { url, method, body, api_key, timestamp, source={ name, meta={...} } }
|
||||
build_headers = function(meta) return { ["X-App-Sign"] = sign } end,
|
||||
|
||||
-- [可选] 出向序列化（契约 v2）：上游不是 OpenAI 协议时接管写给客户端的字节。
|
||||
-- 不声明则沿用内核的 OpenAI 兜底；声明即完全取代，内核不再混合字段。
|
||||
serialize_response = function(result) return body_str end,
|
||||
serialize_open_preamble = function(result) return bytes end,
|
||||
serialize_chunk = function(delta) return bytes end,
|
||||
serialize_close = function(result) return bytes end,
|
||||
}
|
||||
```
|
||||
|
||||
内置信号辅助：`hmac_sha256_hex(key, data)`、`sha256_hex(data)`、`base64_encode(s)`、
|
||||
`tohex(s)`、`json.encode/decode`、`log(level, msg)`。
|
||||
|
||||
### 内置适配器
|
||||
|
||||
`openai` `deepseek` `anthropic` `gemini` `github` `groq` `mistral` `ollama` `kimicode`
|
||||
`qwen38` `sensenova` `trae`。
|
||||
|
||||
`anthropic`/`gemini`/`ollama` 适配器内置多模态转换（`image_url` → 各自上游格式）；
|
||||
推理开关由中立的 `thinking` 字段表达，`deepseek` 适配器把它映射为
|
||||
`thinking.type = "disabled"`。
|
||||
|
||||
**kimicode** 是展示 `build_headers` 的样例：云端校验调用方 app，需要按
|
||||
`meta.app_secret` 对时间戳+URL+请求体哈希做 HMAC 签名并附 `X-App-Sign` 等头。
|
||||
配好 `sources[].meta.{app_id, app_secret, app_agent}` 即可。
|
||||
|
||||
## 桌面 GUI（Electron）
|
||||
|
||||
可选独立产物，面向内网桌面端用户：**内嵌 ModelRouter 核心**（Clash Verge 形态），
|
||||
一键启停、系统托盘常驻、开机自启、静默启动，内容区一比一内嵌完整 WebUI
|
||||
（状态/对话/密钥/优先级/源/适配器 六个页面，免登录）。服务器用户继续使用纯 Go 二进制。
|
||||
|
||||
### 两种安装包：Headless 与 Desktop
|
||||
|
||||
发布到 Release 的安装包分为两类，面向不同用户：
|
||||
|
||||
- **Headless（服务器版）**：纯核心二进制（无 Electron），配置驱动，适合服务器、
|
||||
容器、systemd 等无人值守场景。发布包含 `deb` + `rpm`（含 systemd unit、内存调优
|
||||
与示例配置）与 `tar.gz`（二进制 + `config.example.yaml` + `adapters/` + README），
|
||||
Windows 为 `.zip`。
|
||||
- **Desktop（桌面版）**：Electron GUI 安装包，内嵌核心，适合个人桌面日常使用。
|
||||
|
||||
### 选择 GUI 还是纯后端
|
||||
|
||||
| 场景 | 推荐 | 理由 |
|
||||
| ------ | ------ | ------ |
|
||||
| 服务器 / 内网网关 / 无人值守常驻 | **Headless**（单二进制） | 极轻量（二进制 ~8MB；RSS 随源数量，单源 ~10MB、16 源 ~32–35MB，见[内存占用](#内存占用实测与调优)），零依赖单进程，直接跑在 systemd/任意容器里，远程管理 |
|
||||
| 个人桌面日常使用 / 多设备内网共享 | **Desktop**（GUI） | 免登录内嵌 WebUI、系统托盘一键启停、开机自启、静默后台，适合不懂命令行的使用者 |
|
||||
| Windows 桌面 | **Desktop** | 纯后端在 Windows 上需自行注册服务，GUI 提供原生托盘/自启体验 |
|
||||
| CI 一键出三平台安装包 | **Desktop 打包脚本** | `make gui-*` 系列出 deb/AppImage/NSIS，可进发行流水线 |
|
||||
|
||||
两者完全同源：GUI 内嵌的就是纯后端的同一份 `llmsproxy` 核心（LuaJIT 版），
|
||||
配置/适配器格式完全一致，可随时互换。
|
||||
|
||||
### 构建与运行
|
||||
|
||||
```bash
|
||||
cd cmd/gui
|
||||
npm install
|
||||
make gui # 或 npm run dev —— 开发运行（需先 make build 生成 bin/llmsproxy）
|
||||
make gui-deb # 本机构建 deb（分享给其它 Linux 用户）
|
||||
make gui-dist # 构建 deb + rpm + AppImage（linux，rpm 需系统 rpmbuild）
|
||||
make core-dist # 构建核心 deb + rpm + tar.gz（服务器版，需 nfpm）
|
||||
make gui-win # 构建 Windows nsis 安装包（本机 mingw + wine）
|
||||
make gui-win-docker # 构建 Windows nsis 安装包（全 docker 自足，含 wine）
|
||||
make gui-dist-dir # 仅解包产物（调试用，不产安装包）
|
||||
```
|
||||
|
||||
产物在 `cmd/build/gui-dist/`：`ModelRouter-<ver>.AppImage`、
|
||||
`modelrouter-gui_<ver>_amd64.deb`、`ModelRouter Setup <ver>.exe`（win）；
|
||||
核心包在 `cmd/build/dist/`：`llmsproxy_<ver>_amd64.deb`、`llmsproxy-<ver>.x86_64.rpm`。
|
||||
rpm 需系统 `rpmbuild`。每次打包后都会跑 `packaging/verify-dist.sh` 产物体检
|
||||
（大小下限门禁——1.3.0 曾因宿主 wine 损坏产出 264KB 的残缺 exe 却没有被拦住）。
|
||||
发布时会把产物重命名为 `ModelRouter-Desktop-*`（GUI）与 `ModelRouter-Headless-*`（纯后端）
|
||||
两类上传到 GitCode Release。
|
||||
|
||||
#### Windows 打包（docker 固化方案，推荐）
|
||||
|
||||
`make gui-win-docker` 一步完成：交叉编译核心 + NSIS 打包**全在 docker 内**，宿主机
|
||||
**无需 mingw、无需 wine、无需 node**。镜像内置 wine（含 i386 运行时），修复了
|
||||
此前宿主 wine 损坏（缺 syswow64 → `c0000135`）导致产出 264KB 残缺 exe 的问题。
|
||||
|
||||
- 构建镜像：`cmd/gui/docker/win-builder/Dockerfile`（`golang:1.25` + mingw-w64 +
|
||||
node + wine32/wine64 + 预编译 LuaJIT windows 版 `lua51.dll` 与 import lib）。
|
||||
LuaJIT 源码取自 gitcode 镜像（国内可达），github 兜底；Go 依赖走 `https://goproxy.cn`。
|
||||
- 一键发布：`cmd/gui/scripts/dist-win-docker.sh` —— 核心编译 + electron-builder
|
||||
NSIS 全部在容器内执行（`USE_SYSTEM_WINE=true` 走镜像内的 wine）；electron-builder
|
||||
的 ~400MB 工具链缓存挂到 docker volume，二次构建增量复用。
|
||||
- 首次运行自动构建 `modelrouter/win-builder` 镜像（约 5–10 分钟）；已存在则直接复用。
|
||||
- 打包完成后自动跑 `packaging/verify-dist.sh` 门禁：exe 低于 5MB / deb·rpm·7z 低于
|
||||
10MB 即失败——残缺安装包不可能再溜进 Release。
|
||||
|
||||
```bash
|
||||
# 全 Windows 发布（核心重编 + NSIS）：
|
||||
make gui-win-docker
|
||||
```
|
||||
|
||||
### 特性
|
||||
|
||||
- **内嵌核心**：自动拉起捆绑的 `llmsproxy`（LuaJIT 版），配置/密钥/adapter 存于
|
||||
用户数据目录 `<userData>/profile/`，首次运行自动生成随机 admin key 写入 `keys`
|
||||
（非 seed）并自注入，WebUI 免登录、不会提示"更换初始密钥"。端口可在设置中更改（默认 8787）。
|
||||
- **系统托盘**：内核状态点、开机自启/静默启动开关、启停/重启内核、退出。
|
||||
- **静默启动**：`--silent` 参数或设置项，启动仅驻托盘不弹窗；首次运行始终显示窗口。
|
||||
- **开机自启**：Windows/macOS 用 `setLoginItemSettings`；Linux 写
|
||||
`~/.config/autostart/modelrouter-gui.desktop`，静默开启时自动附加 `--silent`。
|
||||
|
||||
## 目录
|
||||
|
||||
```
|
||||
cmd/llmsproxy # 入口（纯 Go 服务端）
|
||||
cmd/gui # Electron 桌面 GUI（内嵌核心 + 托盘 + 打包）
|
||||
internal/config # YAML 配置加载/校验
|
||||
internal/lua # LuaJIT VM + worker 池 + AdapterCache + 内置适配器 (embed)
|
||||
旧路径与旧环境变量在读取侧仍被接受，因此升级不会因为顺序问题立刻失效。
|
||||
`deploy.sh` 在检测到旧 `/etc/llmsproxy` 而新配置不存在时会**中止**，避免部署出
|
||||
一个「能启动但没有 key / 没有源」的空实例。
|
||||
|
||||
插件 SDK 同名调整：唯一公开入口是 `modelrouter`（`llmsproxy` 仅作兼容别名），
|
||||
详见 [docs/plugins.md](docs/plugins.md)。
|
||||
|
||||
## 目录结构
|
||||
|
||||
```text
|
||||
cmd/modelrouter # 入口（纯 Go 服务端）
|
||||
cmd/gui # Electron 桌面壳（内嵌同一份核心）
|
||||
internal/config # 配置解析、加密存储、运行时文件
|
||||
internal/core # 调度、内核动作、状态与审计
|
||||
internal/lua # LuaJIT VM + 弹性 worker 池 + 内置适配器 (embed)
|
||||
internal/lua/plugins # 可选插件示例（不随二进制/发行包分发）
|
||||
internal/provider # Provider(HTTP) + Registry(路由)
|
||||
internal/gateway # OpenAI 兼容 HTTP 服务 + 鉴权 + SDK/流式 + WebUI
|
||||
internal/gateway # OpenAI 兼容 HTTP 服务 + 鉴权 + 流式 + WebUI
|
||||
internal/types # 统一格式 & OpenAI wire 类型
|
||||
packaging/ # deb/rpm/tar 打包、systemd unit、迁移脚本
|
||||
docs/ # 主题文档
|
||||
```
|
||||
|
||||
## 开发协作
|
||||
|
||||
- **Git 分支工作流**：主分支 / 特性分支 / 发布分支的规范见
|
||||
**[docs/git-workflow.md](docs/git-workflow.md)**（[English](docs/git-workflow-en.md)）。
|
||||
主线：`main` 永可部署 → 特性合入 `main` → 切 `release/vX.Y.Z` 打 tag 发布 →
|
||||
hotfix 提交发布分支并 cherry-pick 回 `main`。
|
||||
- **Lua 适配器协议**：见 **[docs/lua-adapters.md](docs/lua-adapters.md)**
|
||||
（[English](docs/lua-adapters-en.md)）。
|
||||
|
||||
## 测试
|
||||
## 开发与测试
|
||||
|
||||
```bash
|
||||
go test -tags luajit ./...
|
||||
go test -tags luajit ./... # 全量
|
||||
go test -tags luajit ./internal/lua # 单包（LuaJIT 相关改动必跑）
|
||||
```
|
||||
|
||||
覆盖：配置校验、适配器加载/变换、签名钩子、Gateway 鉴权、SDK 结算、SSE 流式、
|
||||
模型路由、多模态透传与 thinking 预算。
|
||||
- 分支与发布纪律见 [docs/git-workflow.md](docs/git-workflow.md)。
|
||||
- 插件放置在 `plugin_dir`，网关启动时**不会**创建、覆盖或删除任何插件文件；
|
||||
升级桌面端也不会动它们。
|
||||
|
||||
689
README_EN.md
689
README_EN.md
@ -2,625 +2,148 @@
|
||||
|
||||
> **中文**: [README.md](./README.md)
|
||||
>
|
||||
> **AI-assisted**: This project is developed with AI-assisted coding (code and docs are produced in collaboration with AI, reviewed by humans).
|
||||
> **AI-assisted**: code and documentation were produced with AI assistance and
|
||||
> reviewed by a human.
|
||||
|
||||
A lightweight, unified OpenAI-compatible LLM gateway for internal networks —
|
||||
configure once, and let multiple services (agents, SDKs, bots) share many
|
||||
upstreams (DeepSeek, Qijiar, OpenAI, Anthropic, Gemini, Groq, Mistral, Ollama,
|
||||
KimiCode…) behind a single endpoint. Pick a specific model or use **AUTO** mode,
|
||||
which routes to the best healthy upstream by configured priority.
|
||||
An intranet gateway for people with several LLM subscriptions: configure once and
|
||||
let multiple services share the upstream capacity. Pick a model explicitly, or let
|
||||
`AUTO` choose by priority among the sources that are currently healthy.
|
||||
|
||||
> **Lightweight, no recompile to add sources**: the gateway is a single Go
|
||||
> binary (8-12 MB, ~8 MB stripped; zero runtime dependencies). Adding or switching an upstream is
|
||||
> just a `sources` entry in `config.yaml` (or via the WebUI) or a `.lua`
|
||||
> adapter — **no Go changes, no recompile**. Changes submitted through the
|
||||
> WebUI take effect immediately (hot reload); editing `config.yaml` or a `.lua`
|
||||
> file under `adapter_dir` by hand requires a process restart.
|
||||
The gateway itself is a single Go binary (~10 MB, no runtime dependencies). Adding
|
||||
or switching an upstream is a `sources` entry in `config.yaml` (or the WebUI), or a
|
||||
dropped-in `.lua` adapter — **no Go changes, no rebuild**. Changes submitted through
|
||||
the WebUI take effect immediately; editing `config.yaml` or a `.lua` adapter file
|
||||
directly requires a restart.
|
||||
|
||||
## Overview
|
||||
## What it does
|
||||
|
||||
Exposes a standard `OpenAI Chat Completions` API (`/v1/chat/completions` +
|
||||
`/v1/models`) to your internal network, translating between multiple upstream
|
||||
protocols via **Lua adapters**. Supports one-shot calls, SSE streaming,
|
||||
**AUTO model routing**, multimodal passthrough, and tool calls.
|
||||
|
||||
Extracted and independently evolved from the multi-source LLM adapter layer of
|
||||
[HomeAgent](https://gitcode.com/JianFeeeee/HomeAgent)
|
||||
(`internal/agent/api/provider.go` + `internal/lua/adapters/*`).
|
||||
|
||||
## Features
|
||||
|
||||
- **Multi-source**: any number of upstream sources in one process, routed by the
|
||||
request's `model`.
|
||||
- **AUTO mode**: with `default_model: AUTO`, scheduling follows the tiered AUTO
|
||||
chain saved on the Priority page (see below).
|
||||
- **Unified output**: every source speaks OpenAI format (incl.
|
||||
`reasoning_content`, `tool_calls`, `usage`).
|
||||
- **Image generation**: `POST /v1/images/generations`, routed to models with
|
||||
`kind: image`.
|
||||
- **Per-model quota**: each key gives every model its own token and request caps
|
||||
plus a reset period (hour/week/month/custom N hours). One model running out
|
||||
affects only that model — the key's other models keep working. Exhaustion
|
||||
answers 429 + `Retry-After` naming the model, so a client resumes when the
|
||||
window rolls over; admin keys are never capped.
|
||||
- **Multimodal**: `content` arrays (`image_url` etc.) pass through losslessly;
|
||||
Anthropic/Gemini/Ollama are translated automatically.
|
||||
- **LuaJIT VM**: golua-binding LuaJIT; each adapter has its own VM + worker
|
||||
pool for safe concurrency. Pools are **elastic**: the sum of `max_concurrent`
|
||||
over an adapter's sources is a ceiling, not a preallocation — states are booted
|
||||
on demand and reclaimed when demand drops, so an idle gateway holds close to
|
||||
zero Lua states. The grow step follows the adapter's max concurrency
|
||||
(`clamp(ceil(max/8), 1, 8)`, and never exceeds the number of queued callers);
|
||||
the shrink step follows the live connection count
|
||||
(`clamp(ceil(slack/(1+in_use)), 1, slack)`), so an idle pool collapses in one
|
||||
round while a busy one gives up a single state at a time.
|
||||
- **On-demand request logs**: dashboard totals are computed from the **full**
|
||||
audit history (streamed once at startup and released — 29 MB / 221k lines in
|
||||
~260 ms), while the raw records are never held in memory: the dashboard loads
|
||||
one screen, scrolling pages the rest straight off disk, CSV export streams in
|
||||
O(1) memory, and leaving the page releases everything.
|
||||
- **Self-healing cooldown**: cooldown is capped at 5 minutes and, past the
|
||||
window's midpoint, exactly one probe request is allowed through; a recovered
|
||||
upstream (or a reset quota) returns to full rotation on that probe instead of
|
||||
waiting out the window.
|
||||
- **disable_thinking**: `disable_thinking: true` toggles reasoning per-request.
|
||||
- **Lua adapter protocol**: each source mounts a `.lua` adapter with
|
||||
`transform_request` / `transform_response` / `transform_stream_chunk` — all
|
||||
protocol differences live in Lua, Go only schedules and proxies.
|
||||
- **Signature / header hooks**: adapters may define `build_headers(meta)` to
|
||||
inject/sign request headers before the HTTP call (e.g. KimiCode style
|
||||
app-validation), with helpers like `hmac_sha256_hex`, `sha256_hex`,
|
||||
`base64_encode`.
|
||||
- **WebUI**: built-in management page to view/add/edit sources & models, edit
|
||||
the AUTO chain, and manage per-key model scopes, persisted to a runtime file.
|
||||
- **Encrypted secrets at rest**: upstream `api_key`, custom header values and
|
||||
gateway keys are stored AES-256-GCM encrypted (`master.key` 0600 next to the
|
||||
runtime file, or `LLMS_PROXY_MASTER_KEY`).
|
||||
- **Live source probing**: the status page probes each source
|
||||
(`GET {base}/models`, falling back to a minimal chat call) without touching
|
||||
the scheduler's backoff state; errors are shown on hover.
|
||||
- **Auth**: the gateway validates client keys via `gateway_keys`; independent
|
||||
from each upstream's own key.
|
||||
- **Streaming**: SSE `chat.completion.chunk` with a role-first chunk and
|
||||
`[DONE]` terminator.
|
||||
- **One OpenAI-compatible API**: `/v1/chat/completions`, `/v1/models`,
|
||||
`/v1/images/generations`. Every source speaks the same shape outwards, including
|
||||
`reasoning_content`, `tool_calls` and `usage`.
|
||||
- **Many sources, one AUTO chain**: configure any number of upstreams; `AUTO`
|
||||
follows the priority tiers saved in the WebUI, fails over between them, cools
|
||||
down broken sources and probes them back to health at half-cool.
|
||||
- **Protocol differences live in Lua**: each source binds a `.lua` adapter that
|
||||
transforms requests, responses and stream chunks both ways, and can sign headers
|
||||
via `build_headers(meta)`. Vendor knowledge stays out of the Go kernel.
|
||||
- **Plugins are independent**: business extensions (channel login, billing, quota
|
||||
probing) are external Lua plugins loaded from `plugin_dir`. They are not compiled
|
||||
into the binary and not shipped in release archives, and they talk to the kernel
|
||||
only through the `modelrouter` SDK.
|
||||
- **Multi-tenant keys**: gateway keys are separate from upstream keys, and can be
|
||||
scoped to a model set, an AUTO chain and a quota.
|
||||
- **Encrypted at rest**: upstream keys, custom header values and gateway keys are
|
||||
sealed with AES-256-GCM; `master.key` is 0600, or injected via
|
||||
`MODELROUTER_MASTER_KEY`.
|
||||
- **Observability**: live source probing, HTTP access log, config-change audit,
|
||||
request and usage CSV export. Writes are kept off the request hot path.
|
||||
|
||||
## Quick start
|
||||
|
||||
```bash
|
||||
GOMODCACHE=... GOPROXY=off go build -tags luajit -o llmsproxy ./cmd/llmsproxy
|
||||
./llmsproxy -config config.yaml # first run generates a default config and prints a random admin key
|
||||
# LuaJIT development libraries are required; -tags luajit is mandatory
|
||||
GOMODCACHE=... GOPROXY=off go build -tags luajit -o modelrouter ./cmd/modelrouter
|
||||
|
||||
# First run generates a default config and prints a random admin key
|
||||
./modelrouter -config config.yaml
|
||||
```
|
||||
|
||||
> Depends on [golua](https://github.com/aarzilli/golua) (LuaJIT bindings).
|
||||
> You **must** build with `-tags luajit`; otherwise the built-in superset
|
||||
> gopher-lua path is used (behavior differs slightly).
|
||||
> Depends on [golua](https://github.com/aarzilli/golua) (LuaJIT bindings). Without
|
||||
> `-tags luajit` the build falls back to gopher-lua, which behaves slightly
|
||||
> differently.
|
||||
|
||||
> **No config is shipped in the repo** (config files carry real keys). On
|
||||
> first run the binary generates a default config at the `-config` path: a
|
||||
> random admin key (printed to the startup log), loopback-only `127.0.0.1:8080`,
|
||||
> and a keyless source ready to chat. Rotate the admin key in the
|
||||
> WebUI after first login.
|
||||
> **No config in the repo**: the repository ships no config file (they contain
|
||||
> keys). The generated default binds `127.0.0.1:8080` only; rotate the admin key in
|
||||
> the WebUI after the first login.
|
||||
|
||||
```bash
|
||||
# no key -> 401
|
||||
# No key -> 401
|
||||
curl http://127.0.0.1:8080/v1/models
|
||||
|
||||
# one-shot ($KEY = the admin key printed at first startup)
|
||||
# Non-streaming ($KEY = the admin key printed at first start)
|
||||
curl -H "Authorization: Bearer $KEY" \
|
||||
-d '{"model":"AUTO","messages":[{"role":"user","content":"hi"}]}' \
|
||||
http://127.0.0.1:8080/v1/chat/completions
|
||||
|
||||
# streaming
|
||||
# Streaming
|
||||
curl -N -H "Authorization: Bearer $KEY" \
|
||||
-d '{"model":"AUTO","stream":true,"messages":[{"role":"user","content":"hi"}]}' \
|
||||
http://127.0.0.1:8080/v1/chat/completions
|
||||
```
|
||||
|
||||
Any OpenAI SDK works: point `base_url` at the gateway, use any authorized
|
||||
gateway key as the API key.
|
||||
|
||||
## Configuration
|
||||
|
||||
Generated on first run at the `-config` path (default `./config.yaml`). Core fields:
|
||||
|
||||
```yaml
|
||||
listen: 127.0.0.1:8080 # bind address (keep internal/loopback)
|
||||
gateway_keys: [sk-gw-<random>] # initial admin key seed (random per install), migrated to the store on first start
|
||||
default_model: AUTO # when model is unroutable, follow the AUTO chain
|
||||
adapter_dir: adapters # Lua adapter dir; created+seeded if missing, read-only otherwise
|
||||
runtime_file: runtime.json # WebUI-edited sources/keys/AUTO chain persist here
|
||||
|
||||
sources:
|
||||
- name: deepseek
|
||||
base_url: https://api.deepseek.com
|
||||
api_key: sk-... # or api_key_env: SOME_ENV to read from an env var
|
||||
adapter: deepseek
|
||||
max_concurrent: 8
|
||||
models:
|
||||
- id: deepseek-v4-flash
|
||||
priority: 100 # YAML sources seed the AUTO chain on first start
|
||||
kind: chat
|
||||
- id: deepseek-v4-pro
|
||||
priority: 60
|
||||
kind: chat
|
||||
# static headers (take precedence over adapter defaults)
|
||||
headers: { X-Tenant: prod }
|
||||
# passthrough metadata for the Lua build_headers hook
|
||||
meta: { app_id: x, app_secret: y }
|
||||
temperature: 0.7
|
||||
max_tokens: 4096
|
||||
timeout: 120s # request timeout, default 120s
|
||||
```
|
||||
|
||||
Sensitive fields in the runtime file (`runtime_file`) are encrypted at rest:
|
||||
|
||||
- AES-256-GCM, stored as `enc:v1:<base64>`.
|
||||
- Master key source: env var `LLMS_PROXY_MASTER_KEY` (64 hex chars), else a
|
||||
`master.key` file next to the runtime file; if neither exists it is generated
|
||||
on first start (mode 0600).
|
||||
- Old plaintext files still load; the first UI save migrates the whole file to
|
||||
ciphertext.
|
||||
- Back up `master.key` with your config — losing it makes the secrets
|
||||
undecryptable. Do not commit it.
|
||||
|
||||
### Gateway keys (multi-key)
|
||||
|
||||
Gateway auth uses a "multi-key + role + model scope" model. Keys are persisted
|
||||
under the `keys` field of the runtime file (encrypted at rest):
|
||||
|
||||
- The `gateway_keys` config is only an **initial admin key seed** — it is
|
||||
migrated into the runtime store on first start and no longer drives auth.
|
||||
- **Important: after first start, replace the admin key via the WebUI Keys
|
||||
page.** The seed key is written in plaintext in `config.yaml`, so keeping it
|
||||
active is a security risk; create a new admin key, log in with it, then
|
||||
delete the seed key.
|
||||
- The WebUI **Keys page** creates/deletes keys. Each key has a role (`admin`
|
||||
manages everything, `user` sees only its own key) and an optional **model
|
||||
scope** (model + source + token quota + reset period). Each model brick shows
|
||||
its own budget badge (e.g. `1.0K 77×·1h`, `∞` when uncapped); clicking it
|
||||
edits that model's token quota, request quota and reset period. Quotas apply
|
||||
per model, so one model running out never blocks the key's others. The
|
||||
"My key" view shows a user its model scopes and their budgets.
|
||||
- Clients authenticate with any authorized key's plaintext as
|
||||
`Authorization: Bearer <key>`.
|
||||
- Deleting a key removes it from the store immediately.
|
||||
|
||||
#### Per-key usage quota
|
||||
|
||||
**Quotas are per model.** Each key's `models[]` list gives every model its own
|
||||
token budget and request budget. One model running out affects only that model
|
||||
— the key's other models keep working.
|
||||
|
||||
```yaml
|
||||
keys:
|
||||
- key: sk-gw-<hex>
|
||||
role: user
|
||||
name: agent-alice
|
||||
models:
|
||||
- model: deepseek-v4-flash
|
||||
token_quota: 1000000 # this key's token budget for this model
|
||||
req_quota: 20000 # requests within the window
|
||||
period: nhour # "" | hour | week | month | nhour
|
||||
hours: 6 # n-hour only
|
||||
- model: AUTO
|
||||
token_quota: 5000000 # AUTO is a quota entry like any other
|
||||
period: hour
|
||||
- model: kimi-k3 # no quota listed = unlimited
|
||||
```
|
||||
|
||||
- **There is deliberately no key-wide total.** A key-wide cap would make
|
||||
"move A's budget to B" a global reallocation; per-model budgets keep each
|
||||
model independently controllable, so it stays visible which model is
|
||||
actually consuming the spend.
|
||||
- `period`: empty = never resets (lifetime total); `hour` / `week` / `month` =
|
||||
fixed windows; `nhour` + `hours` = a custom hour count. **A misspelled
|
||||
period is rejected at write time** rather than silently becoming a
|
||||
never-resetting quota.
|
||||
- Quotas are isolated per key *and* per model within a key: one key exhausting
|
||||
`m1` never draws on another key's budget, and never blocks the same key's
|
||||
`m2`.
|
||||
- Usage counts chat, streaming and image requests, and survives a restart by
|
||||
replaying the audit log (40 days retained, covering the longest `month`
|
||||
window).
|
||||
- An exhausted quota returns **429 + `Retry-After`**
|
||||
(`rate_limit_exceeded`) and the message names the model that ran out, so a
|
||||
client resumes when the window rolls over; a model the key may not use stays
|
||||
403. **Admin keys are never capped** (quotas on their scopes are not
|
||||
enforced either), so a cap can never lock the operator out.
|
||||
- Buckets are whole unix hours, so a window frees up at most an hour late
|
||||
(deliberately freeing late rather than overspending).
|
||||
- `PUT /api/keys/{key}` submits quotas by submitting `models` — the caps are
|
||||
part of the scope, so there is no second budget that can drift out of sync
|
||||
with the model list. An explicit `0` lifts that model's cap.
|
||||
|
||||
##### Quota rejection vs capacity rejection
|
||||
|
||||
A quota is checked at the entry point and occupies no upstream slot, so it is a
|
||||
cheap rejection. Running out of capacity instead waits out the bounded
|
||||
`busyWait` before answering 503. Measured (capacity 8, 0.6 s/request, 100
|
||||
concurrent clients):
|
||||
|
||||
| Case | Status | Latency | Reaches upstream |
|
||||
|---|---|---|---|
|
||||
| Quota spent (`req_quota` 50, 100 concurrent) | `429 rate_limit_exceeded` + `Retry-After` | **19–21 ms** | no |
|
||||
| Capacity exhausted (same topology, no quota) | `503 upstream_error` | ~2.6 s (`busyWait`) | overflow: no |
|
||||
|
||||
So a client can tell "wait for the window to roll over" apart from "wait for
|
||||
upstream capacity" — the first only needs patience, the second usually means
|
||||
back off concurrency or switch sources.
|
||||
|
||||
##### What the quota costs
|
||||
|
||||
Buckets are kept per (key, model, whole unix hour) for 40 days. Measured on an
|
||||
AMD 7840HS:
|
||||
|
||||
- A window query scans the window, not the whole retention: 49 ns for 1 h,
|
||||
55 ns for 24 h, 3.9 µs for 30 d.
|
||||
- Recording one request: 283 ns, 3 allocations (unchanged from before this
|
||||
feature; the allocations come from the record ring buffer).
|
||||
- Memory: the production shape (7 keys x 8 models x 2 sources at full 40-day
|
||||
retention) costs about **3.7 MB**. The `source::model` bucket used by a
|
||||
source-pinned quota is created **lazily** — it is maintained only once some
|
||||
quota actually pins that source, because writing it on every record would
|
||||
cost an extra 18 MB at 20 keys x 8 models x 3 sources.
|
||||
The trade-off: usage recorded before a pinned quota was configured cannot be
|
||||
split by source afterwards, so **a pinned quota may under-count its first
|
||||
window**.
|
||||
|
||||
```
|
||||
HTTP/1.1 429 Too Many Requests
|
||||
Retry-After: 2100
|
||||
{"error":{"type":"rate_limit_exceeded","message":"token quota exceeded for \"deepseek-v4-flash\" (5000000/5000000)"}}
|
||||
```
|
||||
|
||||
### Model routing
|
||||
|
||||
`/v1/chat/completions` `model` resolution order:
|
||||
|
||||
1. `source/model` or `source:model` prefix → pinned source;
|
||||
2. exact match of a source's `model`;
|
||||
3. with `default_model: AUTO` → follow the AUTO chain from the Priority page
|
||||
(see below);
|
||||
4. otherwise the request errors.
|
||||
|
||||
Any OpenAI client can pin to a source by setting `model` to its
|
||||
`name/anything`; `AUTO` (or the gateway's `default_model: AUTO`) follows the
|
||||
tiered AUTO chain.
|
||||
|
||||
### AUTO chain (Priority page)
|
||||
|
||||
AUTO scheduling is driven **only** by the rules saved on the Priority page. The
|
||||
numeric `models[].priority` in source configs no longer participates in
|
||||
scheduling and is no longer shown. **Chat and image generation are two
|
||||
independent chains**, switched in the Priority page via the `Chat / Image` toggle:
|
||||
|
||||
- **Chat chain** (`auto` field): serves `POST /v1/chat/completions` with
|
||||
`model: AUTO`; only `kind: chat` models are honored (image slots are skipped).
|
||||
- **Image chain** (`auto_image` field): serves `POST /v1/images/generations` with
|
||||
`model: AUTO`; only `kind: image` models are honored. With no image chain
|
||||
configured, AUTO image generation falls back to legacy discovery (all sources
|
||||
exposing an image model, in registry order) for backward compatibility.
|
||||
|
||||
Each rule is one "slot": `{ model, source, tier, token_quota, period, hours }`.
|
||||
|
||||
- `tier` is the priority tier: models in the same tier sit side by side and
|
||||
share it; tiers run high → low.
|
||||
- The same model may appear in several slots (e.g. A low → B low → A high →
|
||||
B high) and is tried in tier order.
|
||||
- `token_quota` > 0 means the slot is skipped once its tokens within the reset
|
||||
window are exhausted; `period` supports `hour` / `week` / `month` / `nhour`
|
||||
(with `hours`); empty = unlimited.
|
||||
- Within a tier, slots run by preference score; cooling / quota-exhausted /
|
||||
hard-failed slots fall through to the next tier, and total failure returns
|
||||
503 with a per-tier summary.
|
||||
|
||||
### Sources vs. adapters
|
||||
|
||||
- A **source** describes a connection to an upstream: `name`, `base_url`,
|
||||
`api_key`, model list and priorities.
|
||||
- An **adapter** is the protocol translation logic (Lua script): converts the
|
||||
unified OpenAI request to the upstream's native format and back.
|
||||
- One adapter serves many sources (e.g. `openai.lua` for any OpenAI-compatible
|
||||
site); one source can switch adapters via its `adapter` field.
|
||||
- Sources decide *whom to talk to and which models to expose*; adapters decide
|
||||
*how to talk*. They are linked by the `adapter` field inside each `sources[]`
|
||||
entry.
|
||||
|
||||
Loading flow (assembled by `Core` in `internal/core`):
|
||||
|
||||
1. On startup `lua.NewVM(adapter_dir)` loads all adapters: built-in ones
|
||||
(embedded at compile time) + same-name override files under `adapter_dir`.
|
||||
2. `config.Load` reads `config.yaml`; `config.NewStore(runtime_file)` reads
|
||||
WebUI-edited runtime sources; both are merged by name.
|
||||
3. `rebuildRegistry` creates a `provider.Provider` per source (holding its
|
||||
adapter), sizing the adapter's worker pool from the source's concurrency
|
||||
limit.
|
||||
4. Incoming requests are routed by `Registry` to a provider, which calls the
|
||||
adapter's `transform_request` → HTTP call → `transform_response` /
|
||||
`transform_stream_chunk`.
|
||||
|
||||
"Add/Edit source", "Upload adapter", "AUTO chain edits" and "per-key model
|
||||
scopes" in the WebUI take effect immediately (written to the runtime file /
|
||||
`adapter_dir`, then reassembled — no restart). Directly editing `config.yaml`
|
||||
or files under `adapter_dir` requires a process restart.
|
||||
|
||||
Deleting a source/adapter in the WebUI truly removes runtime (UI-created)
|
||||
sources and `.lua` files; base sources defined in `config.yaml` cannot rewrite
|
||||
that file, so they are hidden via a deletion tombstone (still hidden after
|
||||
restart) and can be restored by re-adding the same name in the UI.
|
||||
|
||||
### Source templates (multi-key balancing)
|
||||
|
||||
When several upstream keys share one config (URL / adapter / model list /
|
||||
concurrency / rpm), copying the whole source block N times is wasteful. A
|
||||
**source template** stores every source field except `name` and `api_key`, so you
|
||||
create N key-only-different sources from one shared skeleton — natural multi-key
|
||||
load balancing (each expanded source schedules, cools down, and reports health
|
||||
independently).
|
||||
|
||||
- WebUI Sources page, top-right **Templates** button: list all templates with
|
||||
edit / delete / create.
|
||||
- Add-source dialog has two header buttons:
|
||||
- **From template** → pick a template → form auto-fills (name + key still
|
||||
yours to enter);
|
||||
- **As template** → save the current form's non-key/name fields as a template.
|
||||
- Templates persist in the runtime file's `source_templates` field alongside
|
||||
runtime sources (hot-managed, apply immediately). Templates hold no secret
|
||||
keys, so they're safe to share / version-control.
|
||||
|
||||
A template is just a recipe — it doesn't become a source by itself. Only the
|
||||
key-bearing sources created from it carry real traffic.
|
||||
|
||||
### Memory footprint (measured and tuned)
|
||||
|
||||
Memory is not a constant: it scales with the **number of configured sources**
|
||||
(each source owns an `http.Transport` with its connection pool plus per-model
|
||||
health state), not with uptime. Measured on this machine (Linux x86_64, 12 cores):
|
||||
|
||||
| Deployment shape | Startup RSS | Settled RSS |
|
||||
|---|---|---|
|
||||
| 1 source / 1 adapter (minimal) | ~4 MB | ~10 MB |
|
||||
| 1 source + a 29 MB audit history | ~19 MB | ~20 MB |
|
||||
| **16 sources / 13 adapters / 59 models (this host)** | ~28 MB | **~32-35 MB** |
|
||||
|
||||
> For reference, the same production config used **~105 MB** before this round of
|
||||
> work. The reduction comes from three places: raw audit records are no longer kept
|
||||
> resident (~25 MB — the aggregates still scan the full history, but the scan
|
||||
> releases as it goes), Lua state pools no longer grow monotonically, and the two
|
||||
> runtime knobs below.
|
||||
|
||||
Breakdown of the production instance (per-region, from `/proc/<pid>/smaps`):
|
||||
|
||||
| Region | RSS | Notes |
|
||||
|---|---|---|
|
||||
| Go heap | ~14 MB | provider/registry/scheduler state + connection-pool buffers |
|
||||
| Other anonymous (thread stacks / LuaJIT chunks / runtime) | ~12 MB | scales with thread count and loaded adapters |
|
||||
| Binary text+rodata | ~8 MB | mapped executable pages (read-only, reclaimable by the kernel) |
|
||||
| Go runtime reservations | ~4 MB | the 2 GB+ you see in VSZ is address space, not physical memory |
|
||||
| Shared libraries | ~3 MB | libc / libluajit / libm |
|
||||
|
||||
**Two recommended deployment knobs** (environment only, no code change):
|
||||
|
||||
```ini
|
||||
# /etc/systemd/system/llmsproxy.service
|
||||
Environment=GOGC=50
|
||||
Environment=MALLOC_ARENA_MAX=2
|
||||
```
|
||||
|
||||
- `MALLOC_ARENA_MAX=2`: LuaJIT allocates through cgo into glibc malloc, and glibc
|
||||
allows up to `8 x nproc` per-thread arenas. Every OS thread that touches malloc
|
||||
claims one (~1 MB each) and **never returns it to the OS**. Measured: 8-15
|
||||
arenas (7-12 MB) down to **0**.
|
||||
- `GOGC=50`: halves the Go heap growth target. **It does nothing on its own** —
|
||||
the heap it saves is immediately consumed by additional glibc arenas (measured
|
||||
20.3 -> 21.5 MB, i.e. slightly worse) — so it must be paired with
|
||||
`MALLOC_ARENA_MAX`. Together they cut settled RSS by **~19%**. The gateway is
|
||||
I/O bound (1min10s of CPU per 9 hours here), so the extra GC cycles are free.
|
||||
|
||||
> The desktop build (Electron) already injects both when spawning the embedded
|
||||
> core (`cmd/gui/main.js`); exporting either variable yourself overrides it. If you
|
||||
> write your own systemd unit or container spec, add them there.
|
||||
|
||||
Optional: `GOMEMLIMIT=48MiB` as a soft ceiling saves roughly 1 MB more, at the
|
||||
cost of GC turning aggressive as the limit approaches. Not worth it for a relaxed
|
||||
deployment.
|
||||
|
||||
### disable_thinking
|
||||
|
||||
With `"disable_thinking": true` in the request body, the gateway passes it to
|
||||
each adapter; the DeepSeek adapter maps it to `extra_body.thinking.type =
|
||||
"disabled"`, other sources follow their own protocol.
|
||||
|
||||
### WebUI
|
||||
|
||||
Built-in admin page at `GET /`; after login you can:
|
||||
|
||||
- **Status**: source online state (live probe, error on hover), per
|
||||
model/source/key usage stats, request records filtered by key, and CSV export
|
||||
over a time range; click a model to pin the connection snippet to it.
|
||||
- **Chat**: streaming / non-streaming debug.
|
||||
- **Keys**: create/edit gateway keys, set per-key model scopes (model + source +
|
||||
token quota + reset period); admins manage all keys, users see only their own.
|
||||
- **Priority**: drag blocks to build the tiered AUTO chain.
|
||||
- **Sources**: add/edit/delete upstream sources online (secrets stored
|
||||
encrypted).
|
||||
- **Adapters**: upload / delete Lua adapter scripts.
|
||||
|
||||
Changes persist to `runtime_file` (survive restarts).
|
||||
|
||||
## Lua adapter protocol
|
||||
|
||||
Full API: **[Lua adapter docs](docs/lua-adapters.md)**
|
||||
([中文](docs/lua-adapters-en.md)).
|
||||
|
||||
Each adapter is a Lua script returning a table
|
||||
(`internal/lua/adapters/<name>.lua`), overridable by a same-name file under
|
||||
`adapter_dir` — **no recompile needed**, restart to take effect; uploads via
|
||||
the WebUI take effect immediately.
|
||||
|
||||
```lua
|
||||
return {
|
||||
name = "mysrc",
|
||||
version = "1.0.0",
|
||||
endpoint = "/chat/completions", -- upstream path (source.endpoint overrides)
|
||||
headers = { ["X-Static"] = "v" }, -- static default headers (fallback)
|
||||
|
||||
-- request: convert unified OpenAI req -> upstream native format, return string
|
||||
transform_request = function(raw_json) ... end,
|
||||
|
||||
-- response: convert upstream response to unified format string
|
||||
-- { content, reasoning_content, finish_reason, token_usage{...}, tool_calls[{...}] }
|
||||
transform_response = function(raw_json) ... end,
|
||||
|
||||
-- streaming chunk: convert upstream SSE data to { content, done, ... }; "" skips
|
||||
transform_stream_chunk = function(raw_chunk) ... end,
|
||||
|
||||
-- [optional] dynamic headers / signature hook
|
||||
-- meta = { url, method, body, api_key, timestamp, source={ name, meta={...} } }
|
||||
build_headers = function(meta) return { ["X-App-Sign"] = sign } end,
|
||||
}
|
||||
```
|
||||
|
||||
Shared helpers: `hmac_sha256_hex(key, data)`, `sha256_hex(data)`,
|
||||
`base64_encode(s)`, `tohex(s)`, `json.encode/decode`, `log(level, msg)`.
|
||||
|
||||
### Built-in adapters
|
||||
|
||||
`openai` `deepseek` `anthropic` `gemini` `github` `groq` `mistral` `ollama`
|
||||
`kimicode` `qwen38` `sensenova` `trae`.
|
||||
|
||||
`anthropic`/`gemini`/`ollama` include multimodal conversion
|
||||
(`image_url` → their native format); with `disable_thinking` the `deepseek`
|
||||
adapter sets `extra_body.thinking.type` to `disabled`.
|
||||
|
||||
**kimicode** demonstrates `build_headers`: the cloud validates the calling
|
||||
app, so you HMAC-sign timestamp+URL+body with `meta.app_secret` and add
|
||||
`X-App-Sign`-style headers. Configure `sources[].meta.{app_id, app_secret,
|
||||
app_agent}`.
|
||||
|
||||
## Desktop GUI (Electron)
|
||||
|
||||
Optional standalone product for desktop users on your internal network: an
|
||||
**embedded ModelRouter core** (Clash Verge-style): one-click start/stop, system
|
||||
tray, autostart, silent launch, and a pixel-for-pixel embedded full WebUI
|
||||
(Status/Chat/Keys/Priority/Sources/Adapters — no login needed). Server users
|
||||
keep using the plain Go binary.
|
||||
|
||||
### Two installer kinds: Headless and Desktop
|
||||
|
||||
Release installers ship in two flavors for different audiences:
|
||||
|
||||
- **Headless (server)**: the plain core binary (no Electron), config-driven,
|
||||
for servers / containers / systemd / unattended runs. Ships as `deb` + `rpm`
|
||||
(systemd unit with memory tuning + example config) and `tar.gz` (binary +
|
||||
`config.example.yaml` + `adapters/` + README), Windows as `.zip`.
|
||||
- **Desktop (desktop app)**: the Electron GUI installer with an embedded core,
|
||||
for personal daily desktop use.
|
||||
|
||||
### GUI vs. plain backend
|
||||
|
||||
| Scenario | Pick | Why |
|
||||
| ---- | ---- | ---- |
|
||||
| Server / intranet gateway / unattended long-running | **Headless** (single binary) | ~8 MB binary; RSS scales with the number of sources (~10 MB for one, ~32-35 MB for 16 — see [Memory footprint](#memory-footprint-measured-and-tuned)), zero-dep single process — drop it into systemd or any container, remote admin |
|
||||
| Personal desktop daily use / multi-device intranet sharing | **Desktop** (GUI) | no-login embedded WebUI, tray one-click, autostart, silent background — for non-CLI users |
|
||||
| Windows desktop | **Desktop** | plain backend needs manual service registration; GUI ships native tray/autostart |
|
||||
| CI one-shot 3-platform installers | **Desktop packaging scripts** | `make gui-*` emits deb / AppImage / NSIS, drops straight into a release pipeline |
|
||||
|
||||
Both share the exact same `llmsproxy` core (LuaJIT build) — configs and
|
||||
adapters are fully compatible, freely interchangeable.
|
||||
|
||||
### Build & run
|
||||
Any OpenAI SDK works: point `base_url` at the gateway and use an authorized gateway
|
||||
key as `api_key`.
|
||||
|
||||
## Documentation
|
||||
|
||||
The detailed documentation is currently written in Chinese.
|
||||
|
||||
| Topic | Document |
|
||||
|---|---|
|
||||
| Implementation and the full feature list | [docs/features.md](docs/features.md) |
|
||||
| Configuration: `config.yaml`, model discovery, multiple keys | [docs/configuration.md](docs/configuration.md) |
|
||||
| Routing: model routing, AUTO chain, cooldown, sources vs adapters, source templates | [docs/routing.md](docs/routing.md) |
|
||||
| Adapters: Lua protocol and built-in adapters | [docs/adapters.md](docs/adapters.md) · [docs/lua-adapters-en.md](docs/lua-adapters-en.md) (English) |
|
||||
| Plugins: SDK contract, hooks, UI extensions, HTTP endpoints | [docs/plugins.md](docs/plugins.md) |
|
||||
| Desktop app: Headless / Desktop artifacts, build and run | [docs/gui.md](docs/gui.md) |
|
||||
| Performance: measured memory use, tuning, thinking budget | [docs/performance.md](docs/performance.md) |
|
||||
| WebUI usage | [docs/webui.md](docs/webui.md) |
|
||||
| Development: layout, tests, branch and release discipline | [docs/development.md](docs/development.md) · [docs/git-workflow-en.md](docs/git-workflow-en.md) (English) |
|
||||
| Deployment and migration | [docs/deployment.md](docs/deployment.md) |
|
||||
|
||||
## Renamed from llmsproxy
|
||||
|
||||
The product name has been ModelRouter for a while; the binary, config directory,
|
||||
service name and environment variables are now `modelrouter` as well:
|
||||
|
||||
| Before | After |
|
||||
|---|---|
|
||||
| `llmsproxy` (binary) | `modelrouter` |
|
||||
| `/etc/llmsproxy/` | `/etc/modelrouter/` |
|
||||
| `llmsproxy.service` | `modelrouter.service` |
|
||||
| `LLMSPROXY_*` | `MODELROUTER_*` |
|
||||
|
||||
Existing installs migrate with a dry-run-by-default script (copies, never moves,
|
||||
and can roll back):
|
||||
|
||||
```bash
|
||||
cd cmd/gui
|
||||
npm install
|
||||
make gui # or npm run dev — dev run (needs make build first for bin/llmsproxy)
|
||||
make gui-deb # local deb (share with other Linux users)
|
||||
make gui-dist # deb + rpm + AppImage (linux; rpm needs system rpmbuild)
|
||||
make core-dist # core deb + rpm + tar.gz (server edition; needs nfpm)
|
||||
make gui-win # Windows NSIS (host mingw + wine)
|
||||
make gui-win-docker # Windows NSIS, fully dockerized (wine in image; no host mingw/node/wine)
|
||||
sudo ./packaging/migrate-to-modelrouter.sh # preview
|
||||
sudo ./packaging/migrate-to-modelrouter.sh --apply # apply
|
||||
sudo ./packaging/migrate-to-modelrouter.sh --rollback --apply
|
||||
```
|
||||
|
||||
Artifacts land in `cmd/build/gui-dist/`: `ModelRouter-<ver>.AppImage`,
|
||||
`modelrouter-gui_<ver>_amd64.deb`, `ModelRouter Setup <ver>.exe` (win); core
|
||||
packages land in `cmd/build/dist/`: `llmsproxy_<ver>_amd64.deb`,
|
||||
`llmsproxy-<ver>.x86_64.rpm`. Both target families run
|
||||
`packaging/verify-dist.sh` afterwards — a size floor that rejects degenerate
|
||||
installers (1.3.0 shipped a 264 KB NSIS stub because the host wine was broken
|
||||
and nothing caught it). At release time artifacts are renamed into the
|
||||
`ModelRouter-Desktop-*` (GUI) and `ModelRouter-Headless-*` (plain backend)
|
||||
families before upload to GitCode Releases.
|
||||
Old paths and old environment variable names are still accepted on the read side,
|
||||
so an upgrade cannot fail merely because of ordering. `deploy.sh` refuses to run
|
||||
when it finds the old `/etc/llmsproxy` without a new config, because that would
|
||||
deploy a gateway with no key and no sources.
|
||||
|
||||
#### Windows packaging (dockerized, recommended)
|
||||
|
||||
`make gui-win-docker` does cross-compile + NSIS **entirely in docker**; the host
|
||||
needs **no mingw, no wine, no node**. Wine (with the i386 runtime) lives inside
|
||||
the image, fixing the broken-host-wine failure that produced the 264 KB stub
|
||||
installer.
|
||||
|
||||
- Image: `cmd/gui/docker/win-builder/Dockerfile` (`golang:1.25` + mingw-w64 +
|
||||
node + wine32/wine64 + prebuilt LuaJIT for Windows: `lua51.dll` + import lib).
|
||||
LuaJIT source comes from the gitcode mirror (reachable in CN) with github
|
||||
fallback; Go modules go through `https://goproxy.cn`.
|
||||
- One-shot release: `cmd/gui/scripts/dist-win-docker.sh` — core build + NSIS
|
||||
packaging run inside the container (`USE_SYSTEM_WINE=true` uses the image's
|
||||
wine); electron-builder's ~400 MB toolchain cache lives in a docker volume so
|
||||
later builds reuse it.
|
||||
- The `modelrouter/win-builder` image is auto-built on first run (~5-10 min)
|
||||
and reused afterwards.
|
||||
- After packaging, `packaging/verify-dist.sh` runs as a gate: an exe under
|
||||
5 MB (or a deb/rpm/7z under 10 MB) fails the build — a broken installer can
|
||||
never reach a Release again.
|
||||
|
||||
```bash
|
||||
# Full Windows release (rebuild core + NSIS):
|
||||
make gui-win-docker
|
||||
```
|
||||
|
||||
### Features
|
||||
|
||||
- **Embedded core**: auto-spawns the bundled `llmsproxy` (LuaJIT build);
|
||||
config/keys/adapters live in `<userData>/profile/`. First run generates a
|
||||
random admin key written into `keys` (not seed) and self-injects it — WebUI
|
||||
needs no login and never nags about changing the initial key. Port changeable
|
||||
in settings (default 8787).
|
||||
- **System tray**: core status dot, autostart/silent toggles, start/stop/restart
|
||||
core, quit.
|
||||
- **Silent launch**: `--silent` flag or setting — starts to the tray only;
|
||||
first run always shows the window.
|
||||
- **Autostart**: `setLoginItemSettings` on Windows/macOS; writes
|
||||
`~/.config/autostart/modelrouter-gui.desktop` on Linux, appending `--silent`
|
||||
when silent mode is on.
|
||||
The plugin SDK follows the same naming: the only public entry point is
|
||||
`modelrouter` (`llmsproxy` remains a compatibility alias). See
|
||||
[docs/plugins.md](docs/plugins.md).
|
||||
|
||||
## Layout
|
||||
|
||||
```
|
||||
cmd/llmsproxy # entry point
|
||||
internal/config # YAML config loading/validation
|
||||
internal/lua # LuaJIT VM + worker pool + AdapterCache + built-ins (embed)
|
||||
```text
|
||||
cmd/modelrouter # entry point (pure Go server)
|
||||
cmd/gui # Electron desktop shell (embeds the same core)
|
||||
internal/config # config parsing, sealed storage, runtime file
|
||||
internal/core # scheduling, kernel actions, state and audit
|
||||
internal/lua # LuaJIT VM + elastic worker pool + built-in adapters (embed)
|
||||
internal/lua/plugins # optional plugin examples (not shipped in binary/archives)
|
||||
internal/provider # Provider(HTTP) + Registry(routing)
|
||||
internal/gateway # OpenAI-compatible HTTP + auth + SDK/streaming + WebUI
|
||||
internal/types # unified format & OpenAI wire types
|
||||
internal/gateway # OpenAI-compatible HTTP service + auth + streaming + WebUI
|
||||
internal/types # unified format and OpenAI wire types
|
||||
packaging/ # deb/rpm/tar packaging, systemd unit, migration script
|
||||
docs/ # topic documentation
|
||||
```
|
||||
|
||||
## Development & collaboration
|
||||
|
||||
- **Git branching workflow**: main / feature / release-branch conventions in
|
||||
**[docs/git-workflow-en.md](docs/git-workflow-en.md)**
|
||||
([中文](docs/git-workflow.md)). The line: `main` is always deployable →
|
||||
features merge into `main` → cut `release/vX.Y.Z` and tag → release → hotfixes
|
||||
land on the release branch and are cherry-picked back into `main`.
|
||||
- **Lua adapter protocol**: see
|
||||
**[docs/lua-adapters-en.md](docs/lua-adapters-en.md)**
|
||||
([中文](docs/lua-adapters.md)).
|
||||
|
||||
## Tests
|
||||
## Development and tests
|
||||
|
||||
```bash
|
||||
go test -tags luajit ./...
|
||||
go test -tags luajit ./... # everything
|
||||
go test -tags luajit ./internal/lua # one package (required for LuaJIT changes)
|
||||
```
|
||||
|
||||
Covers: config validation, adapter load/transform, signature hooks, gateway
|
||||
auth, SDK round trip, SSE streaming, model routing, multimodal passthrough and
|
||||
disable_thinking.
|
||||
- Branch and release discipline: [docs/git-workflow-en.md](docs/git-workflow-en.md).
|
||||
- Plugins live in `plugin_dir`. The gateway never creates, overwrites or deletes
|
||||
plugin files, and a desktop upgrade never touches them either.
|
||||
|
||||
@ -1,5 +1,5 @@
|
||||
// ModelRouter Desktop GUI — renderer
|
||||
// Clash-Verge-style: the content area is a full-screen iframe loading the
|
||||
// Desktop shell: the content area is a full-screen iframe loading the
|
||||
// real WebUI 1:1 (status / chat / keys / priority / sources / adapters).
|
||||
// The shell only adds the titlebar, settings overlay and tray integration.
|
||||
const state = {
|
||||
|
||||
@ -83,7 +83,7 @@
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<!-- side rail: autostart / silent toggles (Clash-Verge style) -->
|
||||
<!-- side rail: autostart / silent toggles (desktop-shell style) -->
|
||||
<aside class="rail" id="rail">
|
||||
<div class="rail-gap"></div>
|
||||
<button
|
||||
|
||||
@ -1,5 +1,5 @@
|
||||
/* =====================================================================
|
||||
ModelRouter Desktop GUI — Clash-Verge-style shell
|
||||
ModelRouter Desktop GUI — desktop-shell-style shell
|
||||
Content area = full-screen iframe embedding the real WebUI 1:1.
|
||||
Palette mirrors the WebUI (sakura/frost/glassmorphism).
|
||||
===================================================================== */
|
||||
|
||||
60
docs/adapters.md
Normal file
60
docs/adapters.md
Normal file
@ -0,0 +1,60 @@
|
||||
# 适配器协议
|
||||
|
||||
> 本文档由 README 拆分而来：Lua 适配器协议与内置适配器清单。
|
||||
> 编写适配器的完整参考见 [lua-adapters.md](lua-adapters.md)。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## Lua 适配器协议
|
||||
|
||||
完整 API 见 **[Lua 适配器 API 文档](docs/lua-adapters.md)**（[English](docs/lua-adapters-en.md)）。
|
||||
|
||||
每个适配器是一个返回 table 的 Lua 脚本（`internal/lua/adapters/<name>.lua`），
|
||||
加载时可被 `adapter_dir` 下的同名脚本覆盖——**改 Lua 脚本无需重编译**，
|
||||
重启即生效；通过 WebUI 上传的适配器与在线编辑的源配置则即时生效。
|
||||
|
||||
```lua
|
||||
return {
|
||||
name = "mysrc",
|
||||
version = "1.0.0",
|
||||
endpoint = "/chat/completions", -- 上游路径（可被 source.endpoint 覆盖）
|
||||
headers = { ["X-Static"] = "v" }, -- 静态默认请求头（build_headers 缺省时使用）
|
||||
|
||||
-- 请求转换：把 OpenAI 格式 req 转成上游原生格式，返回字符串
|
||||
transform_request = function(raw_json) ... end,
|
||||
|
||||
-- 响应转换：把上游原生响应转成统一格式字符串
|
||||
-- { content, reasoning_content, finish_reason, token_usage{...}, tool_calls[{...}] }
|
||||
transform_response = function(raw_json) ... end,
|
||||
|
||||
-- 流式分块转换：把上游 SSE data 转成 { content, done, ... }，返回 "" 则跳过
|
||||
transform_stream_chunk = function(raw_chunk) ... end,
|
||||
|
||||
-- [可选] 动态请求头/签名钩子
|
||||
-- meta = { url, method, body, api_key, timestamp, source={ name, meta={...} } }
|
||||
build_headers = function(meta) return { ["X-App-Sign"] = sign } end,
|
||||
|
||||
-- [可选] 出向序列化（契约 v2）：上游不是 OpenAI 协议时接管写给客户端的字节。
|
||||
-- 不声明则沿用内核的 OpenAI 兜底；声明即完全取代，内核不再混合字段。
|
||||
serialize_response = function(result) return body_str end,
|
||||
serialize_open_preamble = function(result) return bytes end,
|
||||
serialize_chunk = function(delta) return bytes end,
|
||||
serialize_close = function(result) return bytes end,
|
||||
}
|
||||
```
|
||||
|
||||
内置信号辅助：`hmac_sha256_hex(key, data)`、`sha256_hex(data)`、`base64_encode(s)`、
|
||||
`tohex(s)`、`json.encode/decode`、`log(level, msg)`。
|
||||
|
||||
### 内置适配器
|
||||
|
||||
`openai` `deepseek` `anthropic` `gemini` `github` `groq` `mistral` `ollama` `kimicode`
|
||||
`qwen38` `sensenova` `trae`。
|
||||
|
||||
`anthropic`/`gemini`/`ollama` 适配器内置多模态转换（`image_url` → 各自上游格式）；
|
||||
推理开关由中立的 `thinking` 字段表达，`deepseek` 适配器把它映射为
|
||||
`thinking.type = "disabled"`。
|
||||
|
||||
**kimicode** 是展示 `build_headers` 的样例：云端校验调用方 app，需要按
|
||||
`meta.app_secret` 对时间戳+URL+请求体哈希做 HMAC 签名并附 `X-App-Sign` 等头。
|
||||
配好 `sources[].meta.{app_id, app_secret, app_agent}` 即可。
|
||||
|
||||
183
docs/configuration.md
Normal file
183
docs/configuration.md
Normal file
@ -0,0 +1,183 @@
|
||||
# 配置
|
||||
|
||||
> 本文档由 README 拆分而来：网关配置、模型自动探测、多密钥。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## 配置文件
|
||||
|
||||
首次运行在 `-config` 路径自动生成默认配置（默认 `./config.yaml`），核心字段：
|
||||
|
||||
```yaml
|
||||
listen: 127.0.0.1:8080 # 网关监听地址（建议绑内网/回环）
|
||||
gateway_keys: [sk-gw-<随机>] # 初始 admin 密钥（每次安装随机生成，仅首启写入运行时存储用）
|
||||
default_model: AUTO # model 无法路由时自动按优先级链选源
|
||||
adapter_dir: adapters # Lua 适配目录；目录不存在时首启创建并 seed 内置，存在则只读
|
||||
runtime_file: runtime.json # WebUI 编辑的源/密钥/AUTO 链持久化到此文件
|
||||
|
||||
sources:
|
||||
- name: deepseek
|
||||
base_url: https://api.deepseek.com
|
||||
api_key: sk-... # 也可用 api_key_env: SOME_ENV 引用环境变量（不落盘明文）
|
||||
adapter: deepseek
|
||||
max_concurrent: 8
|
||||
models:
|
||||
- id: deepseek-v4-flash
|
||||
priority: 100 # YAML 源首次启动会 seed 进 AUTO 链的初值
|
||||
kind: chat
|
||||
- id: deepseek-v4-pro
|
||||
priority: 60
|
||||
kind: chat
|
||||
# 静态请求头（比适配器默认优先）
|
||||
headers: { X-Tenant: prod }
|
||||
# 传给 Lua build_headers 的透传元数据
|
||||
meta: { app_id: x, app_secret: y }
|
||||
temperature: 0.7
|
||||
max_tokens: 4096
|
||||
timeout: 120s # 请求超时，默认 120s
|
||||
```
|
||||
|
||||
|
||||
## 自动探测模型列表
|
||||
|
||||
不想手抄模型 id 时，只填 URL 和 key，让网关去扫：
|
||||
|
||||
```bash
|
||||
# 只探测，不写任何东西
|
||||
curl -X POST http://127.0.0.1:8080/api/sources/discover \
|
||||
-H "Authorization: Bearer sk-gw-…" -H "Content-Type: application/json" \
|
||||
-d '{"base_url":"https://api.vendor.com/v1","api_key":"sk-…"}'
|
||||
# → {"ok":true,"base_url":"…","endpoint":"","models":["gpt-x",…],
|
||||
# "image_models":["gpt-image-2"],"models_path":"/models","latency_ms":1387,…}
|
||||
|
||||
# 探测后直接落库
|
||||
curl -X POST http://127.0.0.1:8080/api/sources/discover/apply \
|
||||
-H "Authorization: Bearer sk-gw-…" -H "Content-Type: application/json" \
|
||||
-d '{"base_url":"https://api.vendor.com/v1","api_key":"sk-…","name":"vendor"}'
|
||||
```
|
||||
|
||||
WebUI 源编辑框里也有「自动探测模型」按钮：填好 URL 和 key 点一下，列表自动
|
||||
填好，仍需点保存才生效。
|
||||
|
||||
几点值得知道：
|
||||
|
||||
- **URL 形态不用猜。** 厂商对「`/v1` 放 base 还是放 path」没有共识，探测会
|
||||
依次尝试若干候选并返回答对的那个，同时给出归一化后的 `base_url` +
|
||||
`endpoint`（`endpoint` 平时留空，因为 openai adapter 自带 `/chat/completions`）。
|
||||
- **`image_endpoint` 是另一套规则。** 网关的图像默认路径是**绝对**的
|
||||
`/v1/images/generations`，而 chat 默认是相对的。所以 `base_url` 带 `/v1` 时
|
||||
必须显式给 `image_endpoint: /images/generations`，否则图像请求 404。
|
||||
探测会替你填好。
|
||||
- **图像模型只看上游自己声明的 modality**，绝不从模型名猜。`gpt-image-2`
|
||||
这类 id 也出现在同时提供对话服务的目录上，猜错会让 AUTO 链选中一个只支持
|
||||
图像生成的模型，其后每一层都被它拖垮。
|
||||
- **不会逐模型探活。** 34 个模型的目录跑一轮要花钱且耗时数分钟，`/models`
|
||||
是足够便宜的信号；模型是否真能应答留给正常请求去发现。
|
||||
- **对已存在的源，必须显式 `replace_models: true`** 才会覆盖模型列表。重新
|
||||
探测顺手抹掉为 AUTO 链手工调过的 priority，是最难排查的一类数据丢失。
|
||||
- **`api_key` 留空表示沿用已存值**，不会把凭据清空。
|
||||
|
||||
|
||||
`config.yaml`、`runtime.json`、`master.key`、`adapters/` 均不提交版本库（见
|
||||
`.gitignore`），请随备份带走——`master.key` 丢失后运行时密文无法解密。
|
||||
|
||||
运行文件生成在 `-config` 所在目录：`adapter_dir` / `runtime_file` 自动指向
|
||||
配置文件旁边的 `adapters/` 与 `runtime.json`，从任意工作目录启动都可用。
|
||||
|
||||
运行时文件（`runtime_file`）中的敏感字段自动加密：
|
||||
|
||||
- 加密算法 AES-256-GCM，格式 `enc:v1:<base64>`。
|
||||
- 主密钥来源：环境变量 `MODELROUTER_MASTER_KEY`（64 位 hex）；否则读取
|
||||
`runtime_file` 同目录的 `master.key`；都不存在时首次启动自动生成 `master.key`（0600）。
|
||||
- 升级时旧明文文件自动兼容：首次运行正常读取，任何 UI 保存操作触发全文件加密迁移。
|
||||
- 注意：**master.key 丢失后密文无法解密**，请随配置一起备份；切勿提交到版本库。
|
||||
|
||||
## 网关密钥（多密钥）
|
||||
|
||||
网关鉴权采用「多密钥 + 角色 + 模型范围」架构，密钥持久化在 `runtime_file` 的
|
||||
`keys` 字段（加密存储）：
|
||||
|
||||
- `gateway_keys` 配置只是**初始 admin 密钥种子**：首次启动迁移为运行时 admin
|
||||
key，之后不再参与鉴权管理。
|
||||
- **重要：首次启动后请在 WebUI「密钥」页更换管理员密钥**——初始密钥明文写在
|
||||
`config.yaml` 里（每次安装随机生成），继续使用存在被盗风险；用新密钥登录后删除初始密钥。
|
||||
- WebUI **密钥页**可创建/删除密钥；每个密钥可指定 `admin`（管理全部）或
|
||||
`user`（仅看自己的 key）角色，并配置**模型范围**（模型 + 源 + token 配额 +
|
||||
重置周期）。
|
||||
- 客户端用任意一个已授权的密钥明文作为 Bearer（`Authorization: Bearer <key>`）。
|
||||
- 删除密钥即从运行时存储移除，立即失效。
|
||||
|
||||
### 密钥用量配额
|
||||
|
||||
**配额按模型单独设置**：每个密钥的 `models[]` 里，每一条模型范围各自带一份
|
||||
token 配额与请求数配额。一个模型用满只影响该模型，同一密钥的其它模型照常工作。
|
||||
|
||||
```yaml
|
||||
keys:
|
||||
- key: sk-gw-<hex>
|
||||
role: user
|
||||
name: agent-alice
|
||||
models:
|
||||
- model: deepseek-v4-flash
|
||||
token_quota: 1000000 # 本周期内该 key 用这个模型的 token 预算
|
||||
req_quota: 20000 # 本周期内的请求次数
|
||||
period: nhour # "" | hour | week | month | nhour
|
||||
hours: 6 # 仅 nhour
|
||||
- model: AUTO
|
||||
token_quota: 5000000 # AUTO 也是一条独立配额
|
||||
period: day # ← 这种写法会被拒绝（词表只有 hour/week/month/nhour）
|
||||
- model: kimi-k3 # 未列配额 = 无限
|
||||
```
|
||||
|
||||
- **没有「整钥总配额」**：这是刻意的设计。整钥总额会让「把 A 模型的额度挪给
|
||||
B 模型」变成一次全局重分配；按模型独立计费则每个模型各自可控，运维可以
|
||||
看出哪个模型吃掉了预算。
|
||||
- `period` 词表：空 = 永不过期（累计总量），`hour` / `week` / `month` = 固定窗口，
|
||||
`nhour` + `hours` = 自定义小时数。**拼错的周期在写入时就被拒**，不会静默变成
|
||||
永不过期。
|
||||
- 配额严格按密钥隔离，且**同一密钥内按模型隔离**：A 密钥用满 `m1` 不会消耗
|
||||
B 密钥的额度，同一密钥的 `m2` 也不受影响。
|
||||
- 配额统计含聊天、流式、生图，跨重启从审计日志回放（保留 40 天，覆盖最长的
|
||||
month 窗口）。
|
||||
- 配额耗尽返回 **429 + `Retry-After`**（`rate_limit_exceeded`），消息里点名是哪个
|
||||
模型用满了，客户端可等窗口重置后自动恢复；模型越权才是 403。
|
||||
**admin 密钥永不受配额限制**（其 scope 上的配额也不强制），避免把管理员
|
||||
锁在门外。
|
||||
- 窗口用量按整点小时分桶统计，实际释放比配置窗口最多晚 1 小时（配额宁可晚释放
|
||||
也不超发）。
|
||||
- `PUT /api/keys/{key}` 提交 `models` 即同时提交它们的配额（配额就是 scope 的一部分，
|
||||
不存在会与模型列表脱节的第二份预算）。显式 `0` = 解除该模型的限制。
|
||||
|
||||
#### 配额拒绝 vs 容量拒绝：两种「拒绝」含义不同
|
||||
|
||||
配额在**入口**检查，不占用任何上游槽位，因此是廉价拒绝；容量不足则要走
|
||||
`busyWait` 有界等待后才返回 503。实测（容量 8、0.6s/请求、100 并发）：
|
||||
|
||||
| 情况 | 状态码 | 延迟 | 是否打到上游 |
|
||||
|---|---|---|---|
|
||||
| 配额用尽（`req_quota` 50 → 100 并发） | `429 rate_limit_exceeded` + `Retry-After` | **19–21ms** | 否 |
|
||||
| 容量打满（同一拓扑，未限配额） | `503 upstream_error` | ~2.6s（`busyWait`） | 超出部分否 |
|
||||
|
||||
客户端因此可以区分「等窗口重置」与「等上游腾出容量」：前者只需等待，后者
|
||||
通常该降并发或换源。
|
||||
|
||||
#### 配额的开销
|
||||
|
||||
配额桶按 (密钥, 模型, 整点小时) 分桶保留 40 天，实测（AMD 7840HS）：
|
||||
|
||||
- 窗口查询按窗口长度而非保留总量扫描：1h 窗口 49ns、24h 窗口 55ns、
|
||||
30d 窗口 3.9μs（此前全扫保留总量，960 桶时 5.9μs）。
|
||||
- 记录一条请求：283ns、3 分配（与引入配额前相同，分配来自 ring buffer）。
|
||||
- 内存：生产形态（7 密钥 × 8 模型 × 2 源 × 满 40 天 retention）约 **3.7MB**。
|
||||
按源 pin 的 `source::model` 桶**惰性创建**——只有当某条配额真的 pin 了
|
||||
某个源时才维护，否则每条记录多写一份桶，在 20 密钥 × 8 模型 × 3 源下会
|
||||
多占 18MB。
|
||||
代价：配置 pinned 配额之前的历史用量无法事后按源拆分，所以 **pinned 配额的
|
||||
首个窗口可能少算**。
|
||||
|
||||
```bash
|
||||
# 配额耗尽时客户端看到（点名了具体模型）
|
||||
HTTP/1.1 429 Too Many Requests
|
||||
Retry-After: 2100
|
||||
{"error":{"type":"rate_limit_exceeded","message":"token quota exceeded for \"deepseek-v4-flash\" (5000000/5000000)"}}
|
||||
```
|
||||
|
||||
137
docs/deployment.md
Normal file
137
docs/deployment.md
Normal file
@ -0,0 +1,137 @@
|
||||
# 部署与迁移
|
||||
|
||||
> 本文档面向运维：单机 systemd 部署、发行包安装、目录与权限、从 `llmsproxy`
|
||||
> 改名后的迁移与回滚。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## 1. 单机部署（推荐）
|
||||
|
||||
仓库根目录的 `deploy.sh` 负责：构建 → 预检配置 → 备份 → 原子替换二进制 →
|
||||
重启服务 → 后置校验。它**不**修改源码，也**不**假设全新机器。
|
||||
|
||||
```bash
|
||||
sudo ./deploy.sh # 默认读取仓库内 config.yaml
|
||||
sudo ./deploy.sh --config other.yaml # 投放并部署另一份配置
|
||||
```
|
||||
|
||||
部署目标（改名后）：
|
||||
|
||||
| 项 | 路径 |
|
||||
|---|---|
|
||||
| 二进制 | `/usr/local/bin/modelrouter` |
|
||||
| 配置 | `/etc/modelrouter/config.yaml` |
|
||||
| 适配器 | `/etc/modelrouter/adapters/` |
|
||||
| 插件（可选） | `/etc/modelrouter/plugins/`（需在配置里设 `plugin_dir`） |
|
||||
| 主密钥 | `/etc/modelrouter/master.key`（0600） |
|
||||
| systemd unit | `/etc/systemd/system/modelrouter.service` |
|
||||
|
||||
`deploy.sh` 会在以下情况**中止**而不是"看起来成功"：
|
||||
|
||||
- 部署后二进制起不来、或新配置 `-check` 不通过；
|
||||
- 凭据无法解封（`master.key` 丢失 / 与配置不匹配）；
|
||||
- 检测到改名前的 `/etc/llmsproxy`，但新配置尚未建立（否则会部署出一个
|
||||
没有 key、没有源的空实例，所有请求都被拒）。
|
||||
|
||||
## 2. 发行包安装
|
||||
|
||||
三个产物由 `packaging/core-dist.sh` 生成到 `cmd/build/dist/`：
|
||||
|
||||
```text
|
||||
modelrouter_<ver>_amd64.deb
|
||||
modelrouter-<ver>.x86_64.rpm
|
||||
modelrouter-<ver>-linux-amd64.tar.gz
|
||||
```
|
||||
|
||||
deb / rpm 安装后：
|
||||
|
||||
```bash
|
||||
sudo systemctl enable --now modelrouter
|
||||
sudo journalctl -u modelrouter -f
|
||||
```
|
||||
|
||||
tar 包内只有：`modelrouter`、`config.example.yaml`、`modelrouter.service`、
|
||||
`migrate-to-modelrouter.sh`、`adapters/`。
|
||||
|
||||
**发行包不含业务插件**（`channels` / `billing` / `quota` / `usage` 等）。
|
||||
它们保留在仓库中作为可选扩展，需要时手工放入 `plugin_dir`，网关不会自动安装、
|
||||
覆盖或删除。这是有意的：插件属于运维资产，生命周期不与内核绑定。
|
||||
|
||||
首次启动会在 `/etc/modelrouter/config.yaml` 生成一份**随机 admin key** 的配置，
|
||||
key 打印在启动日志里。发行包不预置任何可用 key（历史上曾因此泄露过一个真实的
|
||||
`sk-gw-local-0001`）。
|
||||
|
||||
## 3. 目录与权限
|
||||
|
||||
- `/etc/modelrouter` 是唯一的读写根（systemd unit 用 `ProtectSystem=strict` +
|
||||
`ReadWritePaths=/etc/modelrouter` 固化这一点）。
|
||||
- `master.key` 必须 0600。当前 unit 以 root 运行：降权到普通用户需要先把
|
||||
`master.key` 交给该用户并统一目录属主，否则服务会"带着明文落盘继续跑"，
|
||||
失败只留一行日志——这是独立变更，不要和加固混在一起。
|
||||
- `plugin_dir` 由运维维护；插件自带的状态文件（如 `.channels-store.json`）也在其中。
|
||||
|
||||
## 4. 从 llmsproxy 迁移
|
||||
|
||||
二进制、配置目录、服务名、环境变量已统一为 `modelrouter`：
|
||||
|
||||
| 旧 | 新 |
|
||||
|---|---|
|
||||
| `/usr/local/bin/llmsproxy` | `/usr/local/bin/modelrouter` |
|
||||
| `/etc/llmsproxy/` | `/etc/modelrouter/` |
|
||||
| `llmsproxy.service` | `modelrouter.service` |
|
||||
| `LLMSPROXY_PLUGIN_DIR` | `MODELROUTER_PLUGIN_DIR` |
|
||||
| `LLMSPROXY_XDISPLAY_BASE` | `MODELROUTER_XDISPLAY_BASE` |
|
||||
| `LLMS_PROXY_MASTER_KEY` | `MODELROUTER_MASTER_KEY` |
|
||||
|
||||
迁移脚本**默认 dry-run**，复制而非移动，旧目录永不删除：
|
||||
|
||||
```bash
|
||||
sudo ./packaging/migrate-to-modelrouter.sh # 预演（逐条打印将做什么）
|
||||
sudo ./packaging/migrate-to-modelrouter.sh --apply # 执行
|
||||
```
|
||||
|
||||
它做五件事：复制状态目录并改写配置里的旧绝对路径、复核 `master.key` 0600、
|
||||
安装新二进制（旧文件保留）、由旧 unit 生成新 unit（改写 ExecStart /
|
||||
WorkingDirectory / ReadWritePaths）、复制 `/usr/share` 资源；最后跑
|
||||
`modelrouter -check` 并 `systemctl daemon-reload`。
|
||||
|
||||
确认新服务可用后，再手工清理旧布局（脚本不会删）：
|
||||
|
||||
```bash
|
||||
sudo systemctl disable --now llmsproxy.service
|
||||
sudo rm -f /etc/systemd/system/llmsproxy.service
|
||||
sudo rm -rf /usr/local/bin/llmsproxy /etc/llmsproxy /usr/share/llmsproxy
|
||||
```
|
||||
|
||||
回滚（不需要旧文件被删）：
|
||||
|
||||
```bash
|
||||
sudo ./packaging/migrate-to-modelrouter.sh --rollback --apply
|
||||
```
|
||||
|
||||
**读取侧兼容**：旧配置路径与旧环境变量仍被接受，所以"先换二进制、后改 unit"
|
||||
这种顺序不会立刻失效。但写入永远落在新路径，迁移完成后不要两台布局混用。
|
||||
|
||||
## 5. 部署后验证清单
|
||||
|
||||
```bash
|
||||
# 1) 进程与监听
|
||||
systemctl status modelrouter --no-pager
|
||||
ss -ltnp | grep modelrouter
|
||||
|
||||
# 2) 鉴权生效（无 key 必须 401）
|
||||
curl -s -o /dev/null -w '%{http_code}\n' http://127.0.0.1:8080/v1/models
|
||||
|
||||
# 3) 带 key 能列出模型
|
||||
curl -s -H "Authorization: Bearer $KEY" http://127.0.0.1:8080/v1/models | head -c 400
|
||||
|
||||
# 4) 真实请求走通（证明 LuaJIT 适配器路径未被打断）
|
||||
curl -s -H "Authorization: Bearer $KEY" \
|
||||
-d '{"model":"AUTO","messages":[{"role":"user","content":"hi"}]}' \
|
||||
http://127.0.0.1:8080/v1/chat/completions | head -c 400
|
||||
|
||||
# 5) 仓库文件与生产文件一致（部署脚本的收尾检查）
|
||||
sha256sum /usr/local/bin/modelrouter build/modelrouter
|
||||
```
|
||||
|
||||
第 4 步失败但第 3 步成功，通常是适配器或上游凭据问题，不是部署问题：
|
||||
查看 `journalctl -u modelrouter -n 100`。
|
||||
34
docs/development.md
Normal file
34
docs/development.md
Normal file
@ -0,0 +1,34 @@
|
||||
# 开发与目录结构
|
||||
|
||||
> 本文档由 README 拆分而来：目录职责、开发协作、测试。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## 目录
|
||||
|
||||
```
|
||||
cmd/modelrouter # 入口（纯 Go 服务端）
|
||||
cmd/gui # Electron 桌面 GUI（内嵌核心 + 托盘 + 打包）
|
||||
internal/config # YAML 配置加载/校验
|
||||
internal/lua # LuaJIT VM + worker 池 + AdapterCache + 内置适配器 (embed)
|
||||
internal/provider # Provider(HTTP) + Registry(路由)
|
||||
internal/gateway # OpenAI 兼容 HTTP 服务 + 鉴权 + SDK/流式 + WebUI
|
||||
internal/types # 统一格式 & OpenAI wire 类型
|
||||
```
|
||||
|
||||
## 开发协作
|
||||
|
||||
- **Git 分支工作流**：主分支 / 特性分支 / 发布分支的规范见
|
||||
**[docs/git-workflow.md](docs/git-workflow.md)**（[English](docs/git-workflow-en.md)）。
|
||||
主线：`main` 永可部署 → 特性合入 `main` → 切 `release/vX.Y.Z` 打 tag 发布 →
|
||||
hotfix 提交发布分支并 cherry-pick 回 `main`。
|
||||
- **Lua 适配器协议**：见 **[docs/lua-adapters.md](docs/lua-adapters.md)**
|
||||
（[English](docs/lua-adapters-en.md)）。
|
||||
|
||||
## 测试
|
||||
|
||||
```bash
|
||||
go test -tags luajit ./...
|
||||
```
|
||||
|
||||
覆盖：配置校验、适配器加载/变换、签名钩子、Gateway 鉴权、SDK 结算、SSE 流式、
|
||||
模型路由、多模态透传与 thinking 预算。
|
||||
48
docs/features.md
Normal file
48
docs/features.md
Normal file
@ -0,0 +1,48 @@
|
||||
# 实现与完整特性
|
||||
|
||||
> 本文档由 README 拆分而来：架构来源与完整特性清单（README 只保留摘要）。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## 实现
|
||||
|
||||
统一 OpenAI 兼容网关：把多个上游 LLM 源（DeepSeek、Qijiar、OpenAI、Anthropic、Gemini、
|
||||
Groq、Mistral、Ollama、KimiCode…）通过 **Lua 适配器** 做协议转换，对内网暴露
|
||||
一个标准 `OpenAI Chat Completions` 接口（`/v1/chat/completions` + `/v1/models`），
|
||||
支持单次结算、SSE 流式、**AUTO 模型路由**与**多模态**透传。
|
||||
|
||||
从 [HomeAgent](https://gitcode.com/JianFeeeee/HomeAgent) 的多源 LLM 适配层
|
||||
（`internal/agent/api/provider.go` + `internal/lua/adapters/*`）抽离而来并独立演进。
|
||||
|
||||
## 特性
|
||||
|
||||
- **多源**：一个进程内配置任意多个上游源，按请求的 `model` 自动路由。
|
||||
- **AUTO 模式**：`default_model: AUTO` 时按优先级页保存的 AUTO 链档位逐档调度并发请求。
|
||||
- **统一输出**：所有源对外都是 OpenAI 格式（含 `reasoning_content`、`tool_calls`、`usage`）。
|
||||
- **生图**：`POST /v1/images/generations`，`kind: image` 的模型独立路由。
|
||||
- **多模态**：`content` 数组（`image_url` 等）在多源间无损透传，Anthropic/Gemini/Ollama 自动转换。
|
||||
- **LuaJIT VM**：基于 golua 绑定的 LuaJIT，每个适配器独立 VM + worker 池，安全并发。
|
||||
池是 **弹性** 的：`Σ max_concurrent` 只是上限而非预分配，状态按需创建、空闲时自动回收，
|
||||
因此空载网关几乎不持有 Lua 状态（启动 0 个）。
|
||||
- 扩容步长由适配器最大并发决定：`clamp(ceil(max/8), 1, 8)`，且仅在**真实并发争用**
|
||||
（所有现有状态均已占用）时批量预热，顺序流量始终只用 1 个状态。
|
||||
- 缩容步长由当前连接数决定：`clamp(ceil(冗余/(1+占用)), 1, 冗余)`——无连接时一轮收到
|
||||
常驻下限，忙时每轮只释放 1 个，保护热路径。WebUI「适配器」页展示实时池指标。
|
||||
- **thinking 预算**：请求里的推理开关/预算归一为中立的 `thinking:{enabled,budget_tokens}`，
|
||||
适配器映射到各自协议（Anthropic `thinking.budget_tokens`、Gemini `thinkingConfig`、
|
||||
DeepSeek `thinking.type="disabled"`）。
|
||||
- **Lua 适配协议**：每个源挂一个 `.lua` 适配器，完成 `transform_request` /
|
||||
`transform_response` / `transform_stream_chunk` 双向转换，协议差异全在 Lua 层。
|
||||
- **签名 / 请求头钩子**：适配器可定义 `build_headers(meta)`，在 Go 发 HTTP 前
|
||||
动态注入/签名请求头——用于云端 API 校验调用方 app（如 KimiCode 只放行特定
|
||||
agent）。提供 `hmac_sha256_hex` / `sha256_hex` / `base64_encode` 等签名辅助。
|
||||
- **WebUI**：内置管理界面，可在线查看/新增/编辑上游源与模型、配置 AUTO 优先级链、
|
||||
管理密钥模型范围，写入运行时文件持久化。
|
||||
- **密钥加密存储**：运行时文件中的上游 `api_key`、自定义请求头值、网关 key 均以
|
||||
AES-256-GCM 加密落盘（`master.key` 独立 0600，或 `MODELROUTER_MASTER_KEY`）。
|
||||
- **实时源探测**：状态页探测各源可达性（`GET {base}/models`，失败回退最小请求），
|
||||
不污染正常调度的退避状态，错误信息在 UI 可悬停查看。
|
||||
- **鉴权**：网关自身用 `gateway_keys` 校验客户端 Bearer key；与上游各自的 key 相互独立。
|
||||
- **流式**：SSE `chat.completion.chunk`，含角色首包与 `[DONE]` 收尾。
|
||||
- **全链路审计**：HTTP 访问日志、登录/配置变更审计、请求记录 CSV 导出、密钥用量统计 CSV 导出
|
||||
- **AES-256-GCM 加密存储**：运行时文件敏感字段加密落盘，master.key 0600 权限，支持环境变量注入主密钥
|
||||
|
||||
86
docs/gui.md
Normal file
86
docs/gui.md
Normal file
@ -0,0 +1,86 @@
|
||||
# 桌面 GUI（Electron）
|
||||
|
||||
> 本文档由 README 拆分而来：Headless 与 Desktop 两种产物、构建与运行。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## 桌面 GUI（Electron）
|
||||
|
||||
可选独立产物，面向内网桌面端用户：**内嵌 ModelRouter 核心**（桌面壳形态），
|
||||
一键启停、系统托盘常驻、开机自启、静默启动，内容区一比一内嵌完整 WebUI
|
||||
（状态/对话/密钥/优先级/源/适配器 六个页面，免登录）。服务器用户继续使用纯 Go 二进制。
|
||||
|
||||
### 两种安装包：Headless 与 Desktop
|
||||
|
||||
发布到 Release 的安装包分为两类，面向不同用户：
|
||||
|
||||
- **Headless（服务器版）**：纯核心二进制（无 Electron），配置驱动，适合服务器、
|
||||
容器、systemd 等无人值守场景。发布包含 `deb` + `rpm`（含 systemd unit、内存调优
|
||||
与示例配置）与 `tar.gz`（二进制 + `config.example.yaml` + `adapters/` + README），
|
||||
Windows 为 `.zip`。
|
||||
- **Desktop（桌面版）**：Electron GUI 安装包，内嵌核心，适合个人桌面日常使用。
|
||||
|
||||
### 选择 GUI 还是纯后端
|
||||
|
||||
| 场景 | 推荐 | 理由 |
|
||||
| ------ | ------ | ------ |
|
||||
| 服务器 / 内网网关 / 无人值守常驻 | **Headless**（单二进制） | 极轻量（二进制 ~8MB；RSS 随源数量，单源 ~10MB、16 源 ~32–35MB，见[内存占用](#内存占用实测与调优)），零依赖单进程，直接跑在 systemd/任意容器里，远程管理 |
|
||||
| 个人桌面日常使用 / 多设备内网共享 | **Desktop**（GUI） | 免登录内嵌 WebUI、系统托盘一键启停、开机自启、静默后台，适合不懂命令行的使用者 |
|
||||
| Windows 桌面 | **Desktop** | 纯后端在 Windows 上需自行注册服务，GUI 提供原生托盘/自启体验 |
|
||||
| CI 一键出三平台安装包 | **Desktop 打包脚本** | `make gui-*` 系列出 deb/AppImage/NSIS，可进发行流水线 |
|
||||
|
||||
两者完全同源：GUI 内嵌的就是纯后端的同一份 `modelrouter` 核心（LuaJIT 版），
|
||||
配置/适配器格式完全一致，可随时互换。
|
||||
|
||||
### 构建与运行
|
||||
|
||||
```bash
|
||||
cd cmd/gui
|
||||
npm install
|
||||
make gui # 或 npm run dev —— 开发运行（需先 make build 生成 bin/modelrouter）
|
||||
make gui-deb # 本机构建 deb（分享给其它 Linux 用户）
|
||||
make gui-dist # 构建 deb + rpm + AppImage（linux，rpm 需系统 rpmbuild）
|
||||
make core-dist # 构建核心 deb + rpm + tar.gz（服务器版，需 nfpm）
|
||||
make gui-win # 构建 Windows nsis 安装包（本机 mingw + wine）
|
||||
make gui-win-docker # 构建 Windows nsis 安装包（全 docker 自足，含 wine）
|
||||
make gui-dist-dir # 仅解包产物（调试用，不产安装包）
|
||||
```
|
||||
|
||||
产物在 `cmd/build/gui-dist/`：`ModelRouter-<ver>.AppImage`、
|
||||
`modelrouter-gui_<ver>_amd64.deb`、`ModelRouter Setup <ver>.exe`（win）；
|
||||
核心包在 `cmd/build/dist/`：`modelrouter_<ver>_amd64.deb`、`modelrouter-<ver>.x86_64.rpm`。
|
||||
rpm 需系统 `rpmbuild`。每次打包后都会跑 `packaging/verify-dist.sh` 产物体检
|
||||
（大小下限门禁——1.3.0 曾因宿主 wine 损坏产出 264KB 的残缺 exe 却没有被拦住）。
|
||||
发布时会把产物重命名为 `ModelRouter-Desktop-*`（GUI）与 `ModelRouter-Headless-*`（纯后端）
|
||||
两类上传到 GitCode Release。
|
||||
|
||||
#### Windows 打包（docker 固化方案，推荐）
|
||||
|
||||
`make gui-win-docker` 一步完成：交叉编译核心 + NSIS 打包**全在 docker 内**，宿主机
|
||||
**无需 mingw、无需 wine、无需 node**。镜像内置 wine（含 i386 运行时），修复了
|
||||
此前宿主 wine 损坏（缺 syswow64 → `c0000135`）导致产出 264KB 残缺 exe 的问题。
|
||||
|
||||
- 构建镜像：`cmd/gui/docker/win-builder/Dockerfile`（`golang:1.25` + mingw-w64 +
|
||||
node + wine32/wine64 + 预编译 LuaJIT windows 版 `lua51.dll` 与 import lib）。
|
||||
LuaJIT 源码取自 gitcode 镜像（国内可达），github 兜底；Go 依赖走 `https://goproxy.cn`。
|
||||
- 一键发布：`cmd/gui/scripts/dist-win-docker.sh` —— 核心编译 + electron-builder
|
||||
NSIS 全部在容器内执行（`USE_SYSTEM_WINE=true` 走镜像内的 wine）；electron-builder
|
||||
的 ~400MB 工具链缓存挂到 docker volume，二次构建增量复用。
|
||||
- 首次运行自动构建 `modelrouter/win-builder` 镜像（约 5–10 分钟）；已存在则直接复用。
|
||||
- 打包完成后自动跑 `packaging/verify-dist.sh` 门禁：exe 低于 5MB / deb·rpm·7z 低于
|
||||
10MB 即失败——残缺安装包不可能再溜进 Release。
|
||||
|
||||
```bash
|
||||
# 全 Windows 发布（核心重编 + NSIS）：
|
||||
make gui-win-docker
|
||||
```
|
||||
|
||||
### 特性
|
||||
|
||||
- **内嵌核心**：自动拉起捆绑的 `modelrouter`（LuaJIT 版），配置/密钥/adapter 存于
|
||||
用户数据目录 `<userData>/profile/`，首次运行自动生成随机 admin key 写入 `keys`
|
||||
（非 seed）并自注入，WebUI 免登录、不会提示"更换初始密钥"。端口可在设置中更改（默认 8787）。
|
||||
- **系统托盘**：内核状态点、开机自启/静默启动开关、启停/重启内核、退出。
|
||||
- **静默启动**：`--silent` 参数或设置项，启动仅驻托盘不弹窗；首次运行始终显示窗口。
|
||||
- **开机自启**：Windows/macOS 用 `setLoginItemSettings`；Linux 写
|
||||
`~/.config/autostart/modelrouter-gui.desktop`，静默开启时自动附加 `--silent`。
|
||||
|
||||
60
docs/performance.md
Normal file
60
docs/performance.md
Normal file
@ -0,0 +1,60 @@
|
||||
# 性能与内存
|
||||
|
||||
> 本文档由 README 拆分而来：实测内存占用、调优手段、thinking 预算。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## 内存占用（实测与调优）
|
||||
|
||||
内存不是一个定数：它随**配置的源数量**增长（每个源一个 `http.Transport` 与连接池、
|
||||
一组健康状态），与运行时长无关。本机实测（Linux x86_64，12 核）：
|
||||
|
||||
| 部署形态 | 启动 RSS | 稳态 RSS |
|
||||
|---|---|---|
|
||||
| 1 源 / 1 适配器（最小配置） | ~4 MB | ~10 MB |
|
||||
| 1 源 + 29 MB 历史审计日志 | ~19 MB | ~20 MB |
|
||||
| **16 源 / 13 适配器 / 59 模型（本机生产）** | ~28 MB | **~32–35 MB** |
|
||||
|
||||
> 历史参考：本项优化前同一生产配置为 **~105 MB**。降幅来自三处：审计日志的**原始记录**
|
||||
> 不再常驻内存（约 25 MB；统计聚合仍扫全量，但扫完即释放）、Lua 状态池不再单调增长、
|
||||
> 以及下面两个运行时开关。
|
||||
|
||||
内存构成（生产实例分段测量，`/proc/<pid>/smaps`）：
|
||||
|
||||
| 区域 | RSS | 说明 |
|
||||
|---|---|---|
|
||||
| Go 堆 | ~14 MB | provider/registry/scheduler 结构 + 连接池缓冲 |
|
||||
| 其他匹名（线程栈 / LuaJIT chunk / runtime） | ~12 MB | 与线程数、已加载适配器数相关 |
|
||||
| 二进制 text+rodata | ~8 MB | 映射的可执行文件页（只读、可被内核回收） |
|
||||
| Go runtime 预留 | ~4 MB | VSZ 上看到的 2 GB+ 是地址空间预留，不占物理内存 |
|
||||
| 共享库 | ~3 MB | libc / libluajit / libm |
|
||||
|
||||
**两个推荐的部署开关**（只适用于环境变量，无需改代码）：
|
||||
|
||||
```ini
|
||||
# /etc/systemd/system/modelrouter.service
|
||||
Environment=GOGC=50
|
||||
Environment=MALLOC_ARENA_MAX=2
|
||||
```
|
||||
|
||||
- `MALLOC_ARENA_MAX=2`：LuaJIT 的分配走 cgo → glibc malloc，glibc 默认允许 `8×nproc`
|
||||
个 per-thread arena，每个碰到 malloc 的 OS 线程会占用一个（各约 1 MB，且**不归还给系统**）。
|
||||
实测从 8–15 个 arena（约 7–12 MB）降到 **0**。
|
||||
- `GOGC=50`：把 Go 堆增长目标减半。**单独使用无效**（省下的堆会立即被更多 glibc
|
||||
arena 吃掉，实测 20.3 → 21.5 MB 反而变大），必须与 `MALLOC_ARENA_MAX` 配合，
|
||||
两者同时开启才降 **~19%**。网关是 I/O 密集型（本机 9 小时仅消耗 1min10s CPU），
|
||||
多出的 GC 周期与它的空闲 CPU 相比可忽略。
|
||||
|
||||
> 桌面版（Electron）已在 `cmd/gui/main.js` 里默认为内嵌核心注入这两个开关；手动 export 同名
|
||||
> 环境变量可覆盖。如果你用自己的 systemd unit / 容器，建议照上面加上。
|
||||
|
||||
可选：`GOMEMLIMIT=48MiB` 作软上限，实测再省 ~1 MB；代价是逐近上限时 GC 转激进，
|
||||
温和场景不必开。
|
||||
|
||||
## thinking 预算
|
||||
|
||||
请求体带 `"thinking": {"enabled": false}` 时关闭推理；`"thinking": {"enabled": true,
|
||||
"budget_tokens": 4096}` 指定预算。内核只定义这个**中立**语义字段，映射到具体协议由
|
||||
适配器负责（Anthropic `thinking.budget_tokens`、Gemini `thinkingConfig.thinkingBudget`、
|
||||
DeepSeek `thinking.type = "disabled"`）。内核不再向适配器泄漏 `disable_thinking` /
|
||||
`extra_body` 这类私有字段。
|
||||
|
||||
111
docs/routing.md
Normal file
111
docs/routing.md
Normal file
@ -0,0 +1,111 @@
|
||||
# 路由与调度
|
||||
|
||||
> 本文档由 README 拆分而来：模型路由、AUTO 链、冷却与半冷却、源与适配器、源模板。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## 模型路由
|
||||
|
||||
`/v1/chat/completions` 的 `model` 解析顺序：
|
||||
|
||||
1. `source/model` 或 `source:model` 前缀 → 指定源；
|
||||
2. 精确匹配某个源的 `model`；
|
||||
3. 配置 `default_model: AUTO` 时 → 走优先级页保存的 AUTO 链（见下）；
|
||||
4. 否则回落（返回错误）。
|
||||
|
||||
任何 OpenAI 客户端，只要 `model` 设为某个源的 `name/任意名`，即可锁定走该源；
|
||||
设为 `AUTO`（或网关配了 `default_model: AUTO`）即自动按优先级链选源。
|
||||
|
||||
## AUTO 链（优先级页）
|
||||
|
||||
AUTO 调度**只**由优先级页保存的规则决定，源配置里的 `models[].priority` 数字不再
|
||||
参与调度、也不再显示。**聊天与生图是两套独立的链**，在优先级页面通过「聊天 / 生图」
|
||||
按钮切换编辑：
|
||||
|
||||
- **聊天链**（`runtime_file` 的 `auto` 字段）：服务 `POST /v1/chat/completions`
|
||||
的 `model: AUTO` 请求；只接受 `kind: chat` 模型，配置在链里的生图槽位会被忽略。
|
||||
- **生图链**（`runtime_file` 的 `auto_image` 字段）：服务 `POST /v1/images/generations`
|
||||
的 `model: AUTO` 请求；只接受 `kind: image` 模型。未配置生图链时，AUTO 生图回落
|
||||
到「所有含生图模型的源、按注册顺序」的旧行为，兼容旧配置。
|
||||
|
||||
每条规则 = 一个「槽位」：`{ model, source, tier, token_quota, period, hours }`。
|
||||
|
||||
- `tier` 表示优先级档位：同一档的模型并排、共享该优先级；档位从上到下递减。
|
||||
- 同一模型可配置多个槽位（如 A 源低配 → B 源低配 → A 源高配 → B 源高配），按档位顺延。
|
||||
- `token_quota` > 0 时该槽位在重置周期内用满即顺延到下一槽位；`period` 支持
|
||||
`hour` / `week` / `month` / `nhour`（配合 `hours`），空 = 不限。
|
||||
- 同 tier 内按「偏好分」排序，失败槽位冷却后自动跳过；冷却 / 配额耗尽 / hard
|
||||
错误会顺延到下一 tier，全部失败时返回 503 并附上每档失败摘要。
|
||||
|
||||
## 冷却与半冷却探测（自愈调度）
|
||||
|
||||
槽位失败后进入指数退避冷却（5s → 10s → 20s …，上限 **5 分钟**）。冷却不是全时段
|
||||
硬闸：
|
||||
|
||||
- **前半段完全静默**：刚失败的槽位不接任何流量，避免对故障上游持续施压。
|
||||
- **后半段放行 1 个探测**：窗口过半后，同一 `(源, 模型)` 最多允许 **一个** 请求作为
|
||||
探测通过。探测槽数 = `clamp(max_concurrent / 10, 1, 2)`——`max_concurrent: 10`
|
||||
的源恰好放行 1 个探测请求。
|
||||
- **探测排在最后**：同 tier 内探测候选永远排在健康候选之后，只有没有正常槽可用时
|
||||
才承接真实流量，因此探测不会抢走可用容量。
|
||||
- **探测即真实请求**：成功则立即清零失败计数与冷却，槽位当场回到正常轮询——
|
||||
上游恢复/额度恢复后不必等满整个冷却窗口。失败则重开一个新窗口，下次探测顺延到
|
||||
新窗口的中点，不会连续重试。
|
||||
|
||||
分类冷却策略：
|
||||
|
||||
| 情况 | 冷却 | 说明 |
|
||||
|---|---|---|
|
||||
| 传输错误 / 5xx | 5s→10s→20s…，上限 5min | 指数退避 |
|
||||
| 401 / 403 凭据错误 | 10min | 换 key 后由下一次探测自动接回 |
|
||||
| 429 限流 | 30s 固定 | 「太快」不等于「坏了」，不进指数阶梯 |
|
||||
| 配额耗尽（`insufficient_quota` / `余额不足` 等） | 对齐配额窗口，上限 30min | **不进指数阶梯、不涨失败计数**，额度恢复即可用 |
|
||||
|
||||
WebUI「优先级」页的槽位健康标签会区分 `冷却` / `待探测` / `探测中`，悬停可看窗口
|
||||
起点、探测放行时刻与完全恢复时刻。
|
||||
|
||||
## 源（Source）与适配器（Adapter）的关系
|
||||
|
||||
- **源** 是到某个上游的连接描述：`name`、`base_url`、`api_key`、模型列表与优先级。
|
||||
- **适配器** 是协议转换逻辑（Lua 脚本）：把统一 OpenAI 格式请求转成上游原生格式，
|
||||
再把上游响应/流式分块转回统一格式。
|
||||
- 一个适配器可被多个源复用（如 `openai.lua` 同时服务多个 OpenAI 兼容站点）；
|
||||
同一个源也可以切换不同适配器（改 `adapter` 字段即可）。
|
||||
- 源决定“连谁、暴露哪些模型”，适配器决定“怎么对话”——二者在 `sources[]` 条目中
|
||||
通过 `adapter` 字段关联。
|
||||
|
||||
加载流程（`internal/core` 的 `Core` 负责装配）：
|
||||
|
||||
1. 启动时 `lua.NewVM(adapter_dir)` 加载全部适配器：内置适配器（编译期 embed）+
|
||||
`adapter_dir` 下的同名覆盖文件；内置文件写在代码内，覆盖文件要求更高优先级。
|
||||
2. `config.Load` 读取 `config.yaml`，`config.NewStore(runtime_file)` 读 WebUI 改动的
|
||||
运行时源，二者按名称合并成完整源列表。
|
||||
3. `rebuildRegistry` 为每个源创建 `provider.Provider`（持有目标适配器），并按源的
|
||||
并发上限配置适配器 worker 池**上限**（不预分配，见下）。
|
||||
4. 请求进来时 `Registry` 按 `model` 路由到 Provider，Provider 调适配器
|
||||
`transform_request` → HTTP 发送 → `transform_response` / `transform_stream_chunk`。
|
||||
|
||||
WebUI 上的"新增/编辑源"、"上传 Lua 适配器"、"改 AUTO 优先级链"与"密钥模型范围"
|
||||
都即时生效（写入运行时文件或 `adapter_dir` 后重新装配，无需重启）；直接编辑
|
||||
`config.yaml` / `adapter_dir` 下的文件则需要重启进程才会重新加载。
|
||||
|
||||
在 WebUI 删除源/适配器时：运行时（WebUI 创建）的源与 `.lua` 文件会真正移除；
|
||||
`config.yaml` 中定义的基源无法改写配置文件，采用删除标记隐藏（重启后仍隐藏），
|
||||
在 UI 里重新添加同名源即可恢复。
|
||||
|
||||
## 源模板（多 key 负载均衡）
|
||||
|
||||
多个上游 key 共享同一套配置（URL / 适配器 / 模型列表 / 并发 / rpm）时，逐一复制
|
||||
整段 source 太冗余。**源模板**保存除 `name` 和 `api_key` 以外的全部源字段，
|
||||
从模板一键创建多个仅 key 不同的源，天然支持多 key 负载均衡（每个展开源独立调度、
|
||||
独立冷却、独立健康状态）。
|
||||
|
||||
- WebUI「源」页右上角「**模板管理**」按钮：列出全部模板，可编辑 / 删除 / 新建。
|
||||
- 添加源弹窗右上角两个按钮：
|
||||
- **从模板创建** → 选模板 → 表单自动填充（name + key 仍需你填）；
|
||||
- **存为模板** → 把当前表单非 key/name 字段存为新模板。
|
||||
- 模板持久化在 `runtime_file` 的 `source_templates` 字段，与运行时源一样热管理，
|
||||
改完即时生效。模板本身不含密钥，可以安全共享 / 纳管。
|
||||
|
||||
模板不会自己变成源——它只是「配方」。真正承担流量的是从模板创建出来的、带 key
|
||||
的具体源。
|
||||
|
||||
23
docs/webui.md
Normal file
23
docs/webui.md
Normal file
@ -0,0 +1,23 @@
|
||||
# WebUI
|
||||
|
||||
> 本文档由 README 拆分而来。
|
||||
> 返回 [README](../README.md)。
|
||||
|
||||
## WebUI
|
||||
|
||||
内置管理界面（`GET /`），登录后可在浏览器完成：
|
||||
|
||||
- **状态页**：源在线状态（实时探测 + 悬停看错误）、模型/源/key 用量统计、请求记录，
|
||||
支持按时间范围导出 CSV；点击模型可生成 pin 到该模型的连接配置。
|
||||
- **对话页**：流式/非流式调试。
|
||||
- **密钥页**：创建/编辑网关 key，为每个 key 配模型范围（模型 + 源 + token 配额 + 周期），
|
||||
管理员管理全部 key，用户只看到自己的 key。每个模型砖块显示自己的配额徽标
|
||||
（如 `1.0K 77×·1h`，未设配额显示 `∞`），点开可编辑该模型的 token 配额、
|
||||
请求数配额与重置周期 —— 配额按模型独立生效，一个用满不影响同一 key 的其它模型。
|
||||
「我的密钥」页对用户展示本 key 的模型范围与各自配额。
|
||||
- **优先级页**：拖拽积木配置 AUTO 链档位。
|
||||
- **源页**：在线增删改上游源（API key 等敏感字段加密落盘）。
|
||||
- **适配器页**：上传 / 删除 Lua 适配器脚本。
|
||||
|
||||
改动写入 `runtime_file`（重启仍生效）。
|
||||
|
||||
Reference in New Issue
Block a user