wonder
f0245e926d
Deploy Slides / build-and-deploy (push) Successful in 1m13s
- why-not-x.md: Engineering decision rationale (Go vs Python, SQLite vs PG, etc.) - request-journey.md: Full HTTP-to-pixel request tracing with timing breakdown - lessons.md: 4 real engineering pitfalls and how they were solved - Wire all 3 new pages into slides.md in logical positions
40 lines
1.3 KiB
Markdown
40 lines
1.3 KiB
Markdown
# 一次请求的完整旅程
|
||
|
||
从 HTTP 到像素 — 追踪一个生成请求的全链路
|
||
|
||
<div class="text-sm mt-2">
|
||
|
||
```
|
||
POST /api/v1/generate {prompt: "像素风骑士", assetType: "sprite", style: {artStyle: "pixel"}}
|
||
```
|
||
|
||
</div>
|
||
|
||
<v-clicks>
|
||
|
||
<Item title="① Middleware 层（~1ms）">
|
||
JWT 认证 → 解析 userID → 用户级限流检查（Redis Lua）→ 全局限流检查 → 放行
|
||
</Item>
|
||
|
||
<Item title="② Handler 层（~5ms）">
|
||
参数绑定 + 校验 → 去重检查 `hash(prompt+type+params)` → 创建 Task（DB 写入）→ 入队 TaskQueue → 返回 `taskId`
|
||
</Item>
|
||
|
||
<Item title="③ Pipeline 执行（~30s）">
|
||
goroutine 取出任务 → PromptOptimizer（风格合并 + LLM）→ AssetGenerator（文生图 API）→ QualitySupervisor（视觉模型质检，可能重试）→ FormatAdapter（精灵表拆分 + GIF）
|
||
</Item>
|
||
|
||
<Item title="④ 收尾（~2s）">
|
||
上传素材到七牛云 Kodo → 写入 Asset 表 → 更新 Task status=completed → WebSocket 推送完成通知
|
||
</Item>
|
||
|
||
</v-clicks>
|
||
|
||
<div v-click class="mt-2 text-center text-gray-400 text-sm">
|
||
全程异步：前端拿到 taskId 后通过 WebSocket 实时监听进度变化
|
||
</div>
|
||
|
||
<!--
|
||
这就是一个请求的完整生命周期。注意看耗时分布：管线占了 95% 的时间，这就是为什么要做异步。
|
||
-->
|