Perfara 是一款开放、可私有化部署的移动端与 PC 端性能数据平台，面向团队提供与商业闭源方案同级的可视化与分析能力——可作为 Perfdog 等工具的免费替代思路：数据归你、部署归你、链路透明，适合研发、测试与性能专项长期落地。

iOS / PC / 全平台：列表、趋势与详情能力与上述 Demo 同源；使用同一看板入口，按项目与场景筛选即可。

• Android — CPU / 内存 / FPS / 卡顿 / 网络 / 温度 / 功耗等
• iOS & iOS GPU — 进程 CPU、内存、帧率、磁盘与线程相关指标、GPU 细分等
• PC — 与移动端同一套看板语义，便于跨端对比

采集端上传 JSON 后，云端统一解析、图表渲染与分享。

上图：在「趋势统计」中选定项目、平台、场景标记与时间范围后，可对齐同一场景的多轮测试，观察 FPS、CPU、内存等随时间的走势，支撑版本对比与劣化预警。

1. 云端看板 — 浏览器内完成列表、详情、图表与指标汇总，支持按 UID 管理数据范围。
2. 竞品比对 — 在比对页填入 两个 testid（分享令牌），即可并排对比关键指标，快速定位差异。
3. 趋势统计与自动防劣化 — 为用例打上场景标记后，在「趋势统计」中按场景筛选；每次同场景测试在时间轴上形成趋势，便于发现性能回退与波动。
4. 开放数据路径 — 支持自建服务、自主上传与导出；详情页可复制 testid 用于分享与集成。

• 采集端 / 客户端工具说明（飞书文档）
  https://my.feishu.cn/wiki/C7ClwzJvii4lCXk7e1AcWGEInSc?from=from_copylink

下载后按文档连接自有 Data / Web 服务即可上传；数据可在看板中浏览、管理与下载（具体以客户端与接口说明为准）。

1. 依赖：Python 3，安装 requirements.txt。

2. python需要大于python3.8：

3. 启动（二选一）

# 默认 0.0.0.0:8000
uvicorn backend.app:app --host 0.0.0.0 --port 8000

或使用项目根目录 serve.py（适合 Linux 一键启动，路径与数据库按项目根解析）：

# 修改端口示例（Windows / Linux 通用）[推荐] ，或者取serve中改端口和IP
set WEBPERF_PORT=8099
python serve.py

# Linux / macOS
export WEBPERF_PORT=8099
export WEBPERF_HOST=0.0.0.0
python3 serve.py

浏览器访问 http://<主机>:<端口>/；API 为同源 /api/*。更完整的接口与上传格式见仓库内 README.md。

将采集客户端解压或安装后，编辑：

resources/configout/config.js

将其中 开放 Data / Web 服务地址 改为你方部署的看板根 URL（与浏览器访问地址一致，含协议与端口），保存后重启客户端，即可向自有环境上报数据。

• 私有化 / 自建参考： https://github.com/codinghiker/perfara

Perfara is an open, self-hostable performance data platform for mobile and PC workloads—offering visualization and analysis comparable in spirit to commercial closed-source stacks. It serves as a free, transparent alternative to tools like Perfdog: your data, your deployment, your pipeline—ideal for engineering, QA, and long-term performance programs.

iOS / PC / all platforms — Listing, trends, and detail views share the same stack as the demo; open the dashboard URL and filter by project and scenario.

• Android — CPU, memory, FPS, jank, network, thermal, power, and more
• iOS & iOS GPU — Process CPU, memory, frame pacing, disk & scheduling-related metrics, GPU breakdowns
• PC — Same dashboard semantics as mobile for cross-platform benchmarking

Upload performance JSON to your server; the cloud parses, charts, and shares results.

Place these next to this file or under docs/:

Select project, platform, scenario tag, and date range under Trends to align multiple runs of the same scenario and track FPS, CPU, memory over time—ideal for release comparison and anti-regression.

1. Cloud dashboard — Lists, details, charts, and summaries in the browser; UID-scoped data access.
2. Competitor / A-B compare — Enter two testid tokens on the compare page for side-by-side metrics.
3. Trends & anti-degradation — Tag scenarios on each run; filter by scenario in Trends to see evolution and catch performance drift early.
4. Open data path — Self-host, upload, and export; copy testid from detail pages for sharing and automation.

• Client tooling & capture guide (Feishu doc)
  https://my.feishu.cn/wiki/C7ClwzJvii4lCXk7e1AcWGEInSc?from=from_copylink

Point the client at your own Data / Web endpoint after deployment. Browse, manage, and download data from the dashboard per your workflow.

1. Prerequisites: Python 3 and requirements.txt.

2. Run (pick one)

uvicorn backend.app:app --host 0.0.0.0 --port 8000

Or use serve.py at the repo root:

set WEBPERF_PORT=8099
python serve.py

export WEBPERF_PORT=8099
export WEBPERF_HOST=0.0.0.0
python3 serve.py

Open http://<host>:<port>/; APIs are same-origin under /api/*. See README.md in the repository for full API and upload schema.

Edit:

resources/configout/config.js

Set the public Data / Web base URL to your deployed dashboard root (scheme, host, and port). Restart the client to report to your environment.

• Reference / self-host: https://github.com/codinghiker/perfara

Perfara — Measure. Visualize. Stay in control.