diff --git a/DEPLOY.md b/DEPLOY.md
index 6a17e6f..cb6a882 100644
--- a/DEPLOY.md
+++ b/DEPLOY.md
@@ -53,15 +53,34 @@ provided by the AutoDL image, python at `/root/miniconda3`.
   has no systemd / boot hook, so the autostart lives in `~/.bashrc` and only fires
   when you open an **interactive shell (SSH / JupyterLab terminal)**.
 
-Talking to supervisord (socket at `~/bgfilter-supervisor.sock`, config at
-`~/supervisord.conf`):
+**Managing the service** (supervisorctl lives next to python; socket
+`~/bgfilter-supervisor.sock`, supervisord config `~/supervisord.conf`):
 
 ```bash
-/root/miniconda3/bin/supervisorctl -c ~/supervisord.conf status bgfilter   # status
-/root/miniconda3/bin/supervisorctl -c ~/supervisord.conf restart bgfilter  # restart
-tail -f ~/autodl-tmp/bgfilter.log                                           # logs (default ~/autodl-tmp)
+S="/root/miniconda3/bin/supervisorctl -c $HOME/supervisord.conf"
+$S status bgfilter      # state + uptime
+$S restart bgfilter     # restart (e.g. after editing configs/*.yaml)
+$S stop bgfilter        # stop the service (supervisord keeps running)
+$S start bgfilter       # start it again
+$S shutdown             # stop everything, incl. supervisord itself
+
+tail -f ~/autodl-tmp/bgfilter.log   # service log (dir set by --log-dir)
+tail -f ~/supervisord.log           # supervisord's own log (when the service won't start)
 ```
 
+If supervisord itself is not running — fresh container restart with no shell
+opened yet, or a stale-socket error — start it the same way the `~/.bashrc`
+autostart does:
+
+```bash
+rm -f ~/bgfilter-supervisor.sock ~/supervisord.pid
+/root/miniconda3/bin/supervisord -c ~/supervisord.conf
+```
+
+Changes to `~/supervisord.conf` itself (port, log path, …) are applied by
+re-running the deploy script — it regenerates the file and does
+`reread + update + restart`.
+
 > After editing `configs/*.yaml` (e.g. enabling `cross_check.reuse_as_seg`), run
 > the `restart bgfilter` above so the service reloads config — config is read only
 > at startup; there is no per-request switch for it.
@@ -218,39 +237,26 @@ Override the base directory with `BGFILTER_WEIGHTS_DIR`. `models/` is gitignored
 
 ### Cross-check veto (on by default)
 
-The pipeline runs a second, high-resolution matting model
-(`ZhengPeng7/BiRefNet_HR-matting`, ~425 MB) as a "second opinion" that clears
-background residue trapped between hair strands. It is **ON by default** (both CLI
-and HTTP service). **You must provision this model** (Option A or B above):
+The pipeline runs a second matting model (`ZhengPeng7/BiRefNet_HR-matting`,
+~425 MB) as an independent "second opinion" — what it does and why is in
+[README.md](README.md) → *Pipelines*. Deploy-side facts:
 
-- ⚠️ **On an offline server (`HF_HUB_OFFLINE=1`) it must be present before the
-  first request** — otherwise every default request fails, since it cannot
-  download the model at runtime.
-- Turn it off service-wide with `BGFILTER_CROSS_CHECK=0`, or override per request
-  with the `cross_check` form field (`true`/`false`).
-- The service preloads it at startup with the rest of the warm set.
-- Optional: `cross_check.reuse_as_seg: true` in the config (default **off**) makes
-  that same forward double as the segmentation mask on the birefnet backend —
-  replacing the primary seg model instead of adding to it (~20 s/image faster on
-  CPU, ~1 GB less VRAM; the redundant primary segmenter is then skipped at
-  startup). Off by default so the veto stays an independent signal.
+- It is **ON by default** (CLI and HTTP service), so **provision this model**
+  (Option A or B above). ⚠️ On an offline server (`HF_HUB_OFFLINE=1`) it must be
+  present before the first request — otherwise every default request fails.
+- Turn it off service-wide with `BGFILTER_CROSS_CHECK=0`, or per request with the
+  `cross_check` form field; the service preloads it at startup when enabled.
+- `cross_check.reuse_as_seg: true` in the config (default **off**) makes that
+  forward double as the segmentation mask: one less model loaded, ~20 s/image
+  faster on CPU, ~1 GB less VRAM. Config-only — restart to change it.
 
 ### Background mode (flat vs complex)
 
-The `/remove-background` `background_mode` form field selects the pipeline:
-
-- `flat` (**default**, also when omitted): the colour-keyed pipeline for a solid
-  flat background — unchanged behaviour.
-- `complex`: non-flat / scene background. No colour key: segmentation alone drives
-  the trimap and the colour-keyed stages (background auto-detect, hue split, chroma
-  suppression, despill) are bypassed. `screen_color` is ignored; expect flat
-  backgrounds to stay stronger on hair-level detail.
-
-No extra weights: `complex` reuses the already-provisioned BiRefNet segmenter +
-ViTMatte. The cross-check veto still works in `complex` (via a hue-free gate) but
-**defaults off** there — pass `cross_check=on` to enable it. The response echoes the
-mode used in the `X-BGFilter-Background-Mode: flat|complex` header. Full field/header
-reference is in `README.md` → *HTTP service*.
+`background_mode=complex` (non-flat / scene backgrounds) needs **no extra
+weights** — it reuses the already-provisioned BiRefNet segmenter + ViTMatte and
+bypasses the colour-keyed stages. The cross-check veto **defaults off** in
+complex mode (pass `cross_check=on` to enable it). Field semantics and response
+headers: [README.md](README.md) → *HTTP service*.
 
 ### Downloading behind a firewall (CN networks)
 
@@ -434,6 +440,21 @@ curl -H "X-Genarrative-Image-Token: <token>" \
   ViTMatte + one BiRefNet@1024, chunked attention). The **default lane** is slower —
   the cross-check adds a second HR-matting@2048 forward. Requests are serialized by
   the global lock, so plan for a handful of images/min per instance.
+- **Default lane, workstation CPU reference (Ryzen 9700X, native bf16, 32 GB;
+  cross-check on with `reuse_as_seg`, chunked attention):**
+
+  | input | precision | warm / image | peak memory |
+  |---|---|---|---|
+  | 1024×1536 | fp32 | ~45 s | — |
+  | 1024×1536 | bf16 | ~33 s | — |
+  | 2048×2048 | bf16 | ~53 s | 11.4 GB (batch), 8.1 GB (single image) |
+
+  Without `reuse_as_seg` add the primary BiRefNet@1024 forward (~+20 s/image).
+  The memory ceiling is the cross-check HR forward (fixed `input_size` 2048
+  regardless of image size). Before chunked attention the same 2048² bf16 run
+  peaked at 25.2 GB; fully un-optimized fp32 would need an estimated 55–60 GB.
+  CPUs without native bf16 (e.g. Zen 2) auto-fall back to fp32 — prefer
+  `--no-cross-check` there if memory or latency is tight.
 - For throughput: use a GPU (`BGFILTER_DEVICE=cuda`, install a CUDA torch build) —
   roughly an order of magnitude faster — and/or run multiple instances behind a load
   balancer (each with `--workers 1`).
diff --git a/DEPLOY_ZH.md b/DEPLOY_ZH.md
index c67ce46..922ca01 100644
--- a/DEPLOY_ZH.md
+++ b/DEPLOY_ZH.md
@@ -45,14 +45,32 @@ python 在 `/root/miniconda3`。
 - **容器重启后**服务不会自动起来——AutoDL 没有 systemd / 开机钩子，自启写在 `~/.bashrc` 里，
   需要**开一个交互式 shell（SSH / JupyterLab 终端）** 才会触发拉起。
 
-与 supervisord 打交道（socket 在 `~/bgfilter-supervisor.sock`，配置在 `~/supervisord.conf`）：
+**服务管理**（supervisorctl 与 python 同目录；socket 在 `~/bgfilter-supervisor.sock`，
+supervisord 配置在 `~/supervisord.conf`）：
 
 ```bash
-/root/miniconda3/bin/supervisorctl -c ~/supervisord.conf status bgfilter   # 状态
-/root/miniconda3/bin/supervisorctl -c ~/supervisord.conf restart bgfilter  # 重启
-tail -f ~/autodl-tmp/bgfilter.log                                           # 日志（默认 ~/autodl-tmp）
+S="/root/miniconda3/bin/supervisorctl -c $HOME/supervisord.conf"
+$S status bgfilter      # 状态 + 运行时长
+$S restart bgfilter     # 重启（如改了 configs/*.yaml 之后）
+$S stop bgfilter        # 停服务（supervisord 本身继续跑）
+$S start bgfilter       # 再启动
+$S shutdown             # 全部停掉，包括 supervisord 自己
+
+tail -f ~/autodl-tmp/bgfilter.log   # 服务日志（目录由 --log-dir 决定）
+tail -f ~/supervisord.log           # supervisord 自己的日志（服务起不来时看这个）
 ```
 
+若 supervisord 本身没在跑——容器刚重启还没开过 shell，或报 socket 残留错误——照
+`~/.bashrc` 自启的方式手动拉起：
+
+```bash
+rm -f ~/bgfilter-supervisor.sock ~/supervisord.pid
+/root/miniconda3/bin/supervisord -c ~/supervisord.conf
+```
+
+改 `~/supervisord.conf` 本身（端口、日志路径等）的方式是重跑部署脚本——它会重新生成
+该文件并做 `reread + update + restart`。
+
 > 改了 `configs/*.yaml`（比如开 `cross_check.reuse_as_seg`）后，用上面的 `restart bgfilter`
 > 让服务重新加载配置——配置只在启动时读入，没有请求级开关。
 
@@ -201,32 +219,22 @@ hf download ZhengPeng7/BiRefNet_HR-matting       --local-dir models/BiRefNet_HR-
 
 ### cross-check 否决（默认开启）
 
-管线会再跑一个高分辨率 matting 模型（`ZhengPeng7/BiRefNet_HR-matting`，约 425 MB）
-作为"第二意见"，清理发丝间残留的背景色。**CLI 和 HTTP 服务都默认开启**，因此**必须
-备好这个模型**（上面方案 A 或 B）：
+管线会再跑一个 matting 模型（`ZhengPeng7/BiRefNet_HR-matting`，约 425 MB）作为独立的
+"第二意见"——它做什么、为什么，见 [README_ZH.md](README_ZH.md) → *管线*。部署侧要点：
 
-- ⚠️ **离线服务器（`HF_HUB_OFFLINE=1`）必须在第一个请求前就备好它**，否则每个默认请求都会
-  失败——运行时它无法联网下载模型。
-- 服务级关闭：`BGFILTER_CROSS_CHECK=0`；或按请求覆盖：表单字段 `cross_check`（`true`/`false`）。
-- 服务启动时会随暖机集一起预载它。
-- 可选：配置里 `cross_check.reuse_as_seg: true`（默认**关**）可让这次前向在 birefnet
-  后端兼作分割掩码——**替换**而非叠加 primary 分割模型（CPU 快约 20 秒/张、GPU 省约
-  1 GB 显存；此时冗余的 primary 分割模型在启动时会被跳过）。默认关闭是为了让否决
-  保持为独立信号。
+- **CLI 和 HTTP 服务都默认开启**，因此**必须备好这个模型**（上面方案 A 或 B）。
+  ⚠️ 离线服务器（`HF_HUB_OFFLINE=1`）必须在第一个请求前就备好它，否则每个默认请求都失败。
+- 服务级关闭：`BGFILTER_CROSS_CHECK=0`；按请求覆盖：表单字段 `cross_check`。
+  开启时服务启动即预载。
+- 配置里 `cross_check.reuse_as_seg: true`（默认**关**）让这次前向兼作分割掩码：
+  少载一个模型、CPU 快约 20 秒/张、GPU 省约 1 GB 显存。仅配置文件可设——改动需重启。
 
 ### 背景模式（flat vs complex）
 
-`/remove-background` 的 `background_mode` 表单字段选择管线：
-
-- `flat`（**默认**，不传时也是它）：纯色背景的色键管线——行为不变。
-- `complex`：非纯色/实景背景。无色键：完全由分割器驱动 trimap，色键相关阶段（背景
-  自动探测、色相劈分、chroma 抑制、despill）全部跳过。`screen_color` 被忽略；发丝
-  级细节通常弱于纯色模式。
-
-不需要额外权重：`complex` 复用已部署的 BiRefNet 分割器 + ViTMatte。cross-check 否决
-在 `complex` 下仍可用（走无色相门控），但**默认关闭**——传 `cross_check=on` 才启用。
-响应头 `X-BGFilter-Background-Mode: flat|complex` 回显本次实际模式。完整的字段/响应头
-参考见 `README_ZH.md` → *HTTP 服务*。
+`background_mode=complex`（非纯色/实景背景）**不需要额外权重**——复用已部署的
+BiRefNet 分割器 + ViTMatte，跳过色键相关阶段。cross-check 否决在 complex 模式下
+**默认关闭**（传 `cross_check=on` 启用）。字段语义与响应头见
+[README_ZH.md](README_ZH.md) → *HTTP 服务*。
 
 ### 墙内下载（国内网络）
 
@@ -395,6 +403,20 @@ curl -H "X-Genarrative-Image-Token: <token>" \
 - CPU 吞吐：**`cpu-fast` lane** 约 **14–15 秒/张（1.5 MP）**（base ViTMatte + 一个
   BiRefNet@1024，分块注意力）。**默认 lane** 更慢——cross-check 会多跑一次 HR-matting@2048
   前向。请求被全局锁串行，单实例按每分钟几张估算。
+- **默认 lane，工作站 CPU 参考（Ryzen 9700X，原生 bf16，32 GB；cross-check 开且
+  `reuse_as_seg` 开，分块注意力）：**
+
+  | 输入 | 精度 | warm / 每图 | 峰值内存 |
+  |---|---|---|---|
+  | 1024×1536 | fp32 | ~45 s | — |
+  | 1024×1536 | bf16 | ~33 s | — |
+  | 2048×2048 | bf16 | ~53 s | 11.4 GB（批处理）、8.1 GB（单图） |
+
+  不开 `reuse_as_seg` 要加上主 BiRefNet@1024 前向（每图约 +20 s）。内存上限来自
+  cross-check 的 HR 前向（无论图多大都固定 `input_size` 2048）。分块注意力之前，
+  同一个 2048² bf16 run 峰值 25.2 GB；完全未优化的 fp32 估计要 55–60 GB。无原生
+  bf16 的 CPU（如 Zen 2）自动回落 fp32——这类机器上内存或延迟紧张时优先
+  `--no-cross-check`。
 - 要提吞吐：上 GPU（`BGFILTER_DEVICE=cuda`，装 CUDA 版 torch），大约快一个数量级；
   和/或多实例 + 负载均衡（每个都 `--workers 1`）。
 - **GPU 部署（实测，RTX 5070 Ti 16 GB，完整管线——cross-check 开，bf16）：**
diff --git a/README.md b/README.md
index 408775e..5d1209f 100644
--- a/README.md
+++ b/README.md
@@ -65,7 +65,7 @@ frame — detection **fails with an error**; pass `--screen-color` explicitly in
 
 To set it yourself, give a hex prior:
 
-```powershell
+```bash
 ... --screen-color "#CFEFFF"
 ```
 
@@ -76,26 +76,24 @@ auto-detected colour is used directly.)
 
 ## Environment
 
-Use the conda environment `lightML`.
+Any Python 3.11–3.13 environment (venv or conda) works:
 
-```powershell
-conda activate lightML
+```bash
 pip install -r requirements.txt
+python -m bgfilter.cli --help
 ```
 
-If the shell is not activated, call the environment Python directly:
-
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.cli --help
-```
+For a server install — CPU-wheel torch, offline weights, systemd — follow
+[DEPLOY.md](DEPLOY.md) instead; it covers the same dependencies with the
+production choices spelled out.
 
 ## Single Image
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.cli `
-  --input Samples\TestImage.png `
-  --output Outputs\TestImage_rgba.png `
-  --debug-dir Outputs\TestImage_debug
+```bash
+python -m bgfilter.cli \
+  --input Samples/TestImage.png \
+  --output Outputs/TestImage_rgba.png \
+  --debug-dir Outputs/TestImage_debug
 ```
 
 No `--config` is needed — the built-in defaults are identical to
@@ -112,34 +110,19 @@ auto-detect the flat background), and `--trimap-mode`.
 
 ## Batch
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.cli `
-  --input-dir Samples `
-  --output-dir Outputs `
-  --debug-dir Outputs\debug
+```bash
+python -m bgfilter.cli \
+  --input-dir Samples \
+  --output-dir Outputs \
+  --debug-dir Outputs/debug
 ```
 
-## Performance (CPU reference numbers)
+## Performance
 
-Measured on a Ryzen 9700X (Zen 5, native bf16), 32 GB RAM, with cross-check on,
-**`--cross-check-as-seg` enabled** and chunked global attention on. (Reuse is
-off by default on this branch; without it, add the primary BiRefNet@1024
-forward — roughly +20 s/image at these sizes.)
-
-| input | precision | warm / image | peak memory |
-|---|---|---|---|
-| 1024x1536 | fp32 | ~45 s | — |
-| 1024x1536 | bf16 | ~33 s | — |
-| 2048x2048 | bf16 | ~53 s | 11.4 GB (batch), 8.1 GB (single image) |
-
-The memory ceiling is the cross-check BiRefNet_HR forward (fixed
-`input_size` 2048 regardless of the image size) — genuine live activations
-of a full-resolution dense prediction net. Before chunked attention and the
-mimalloc fix, the same 2048x2048 bf16 run peaked at 25.2 GB; a fully
-un-optimized fp32 run would need an estimated 55-60 GB (ViTMatte's un-chunked
-N^2 attention alone ~39 GB). CPUs without native bf16 (e.g. Zen 2) auto-fall
-back to fp32 — on such machines prefer `--no-cross-check` if memory or
-latency is tight.
+Rough orders of magnitude: **~33–53 s/image** on a fast desktop CPU (bf16,
+cross-check on), **~1–2 s/image** on a modern GPU. All measured numbers — CPU
+lanes, GPU VRAM footprints, memory ceilings and the knobs that move them — live
+in [DEPLOY.md](DEPLOY.md) §11 (*Capacity & performance*).
 
 ## Chroma-alpha debug mode
 
@@ -147,12 +130,12 @@ latency is tight.
 alpha seed. Useful for fast inspection of chroma confidence, trimap, and despill.
 (Distinct from the chroma-only *pipeline* above, which still runs ViTMatte.)
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.cli `
-  --input-dir Samples `
-  --output-dir Outputs\chroma `
-  --debug-dir Outputs\chroma_debug `
-  --matting-method chroma `
+```bash
+python -m bgfilter.cli \
+  --input-dir Samples \
+  --output-dir Outputs/chroma \
+  --debug-dir Outputs/chroma_debug \
+  --matting-method chroma \
   --device cpu
 ```
 
@@ -267,39 +250,36 @@ curl -sS \
   http://127.0.0.1:18083/remove-background -o output.png
 ```
 
-Config comes from `BGFILTER_CONFIG` (defaults to `configs/default.yaml` when present);
-`BGFILTER_DEVICE` overrides both model and segmentation device; `BGFILTER_MAX_IMAGE_PIXELS`
-caps input size (default ~4MP → `413`); `BGFILTER_PRELOAD=1` loads the default models at
-startup; `BGFILTER_AUTH_TOKEN` gates the endpoint (see Authentication above).
+Runtime configuration (config file, device, size caps, preload, auth token,
+pipelining workers) is all via `BGFILTER_*` environment variables — the full table
+is [DEPLOY.md](DEPLOY.md) §6.
 
-Like the CLI, the service runs the **cross-model veto by default** (config
-default: on); its HR-matting model is preloaded with the rest of the warm set.
-Set `BGFILTER_CROSS_CHECK=0` to turn it off service-wide, or use the per-request
-`cross_check` form field to override either way; `/healthz` reports the default
-as `crossCheck`. With `cross_check.reuse_as_seg: true` in the config (off by
-default) the HR-matting forward replaces the primary seg forward instead of
-adding to it, and the then-redundant primary segmenter is skipped at startup.
+Like the CLI, the service runs the **cross-model veto by default**; `/healthz`
+reports the effective default as `crossCheck` and the per-request `cross_check`
+field overrides it either way (service-wide kill switch: `BGFILTER_CROSS_CHECK=0`).
+`cross_check.reuse_as_seg` (config-only, default off — see *Pipelines* above) has
+no per-request switch; changing it means editing the config and restarting.
 
 ## Quality Check
 
 The quality checker measures alpha validity and edge spill on semi-transparent edge
 pixels.
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.quality_cli `
-  Outputs\TestImage_rgba.png `
+```bash
+python -m bgfilter.quality_cli \
+  Outputs/TestImage_rgba.png \
   --max-edge-green-excess-p95 0.30
 ```
 
 Run the bundled sample smoke check:
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe scripts\smoke_samples.py `
-  --samples-dir Samples `
-  --output-dir Outputs\smoke_samples `
-  --config configs\default.yaml `
-  --device cpu `
-  --fallback-to-chroma-alpha `
+```bash
+python scripts/smoke_samples.py \
+  --samples-dir Samples \
+  --output-dir Outputs/smoke_samples \
+  --config configs/default.yaml \
+  --device cpu \
+  --fallback-to-chroma-alpha \
   --max-edge-green-excess-p95 0.30
 ```
 
@@ -310,26 +290,9 @@ D:\MiniConda\envs\lightML\python.exe scripts\smoke_samples.py `
   `anime-seg` (`skytnt/anime-seg`) is a plain ONNX download; `birefnet`
   (`ZhengPeng7/BiRefNet`) ships custom modelling code so it needs
   `trust_remote_code=True` plus `timm` / `einops` / `kornia`.
-- Downloading the weights behind a firewall — the reliable combination is the
-  `hf-mirror.com` mirror **with the local proxy bypassed** and **Xet disabled**:
-  ```powershell
-  $env:HF_ENDPOINT      = "https://hf-mirror.com"  # domestic mirror
-  $env:NO_PROXY         = "*"                       # bypass the proxy; the mirror is direct
-  $env:HF_HUB_DISABLE_XET = "1"                     # these repos are Xet-backed; force classic HTTP
-  ```
-  Two gotchas this avoids: (1) mirror + an overseas proxy makes the mirror's `resolve`
-  bounce back to `huggingface.co`, which recent `huggingface_hub` rejects with
-  `FileMetadataError`; (2) with `hf-xet` installed the Xet download path fails instantly.
-  Once the weights are cached, run offline with `HF_HUB_OFFLINE=1 TRANSFORMERS_OFFLINE=1`
-  (as the service does in production).
-- Bundling weights in the project — instead of the HF cache, drop each model into a
-  plain folder named after the repo basename under `models/`:
-  `models/vitmatte-base-composition-1k`, `models/BiRefNet`, `models/anime-seg`,
-  `models/BiRefNet_HR-matting` (cross-check second opinion, default on). The loader
-  prefers a matching local folder and falls back to the HF repo id / cache when absent, so
-  it is opt-in. Populate them with e.g. `hf download ZhengPeng7/BiRefNet --local-dir
-  models/BiRefNet`. Override the base directory with `BGFILTER_WEIGHTS_DIR`. `models/` is
-  gitignored.
+- **Provisioning weights** — offline HF cache, plain `models/` folders
+  (`BGFILTER_WEIGHTS_DIR`), the `scripts/fetch_weights.py` one-shot downloader,
+  and the CN-network mirror workaround — is covered in [DEPLOY.md](DEPLOY.md) §5.
 - Foreground colour estimation uses pymatting's `estimate_foreground_ml` to propagate
   clean foreground colour into semi-transparent edges before de-spill. Set
   `foreground.method: unmix` to fall back to the legacy heuristic.
diff --git a/README_ZH.md b/README_ZH.md
index 57f2099..236ae7f 100644
--- a/README_ZH.md
+++ b/README_ZH.md
@@ -54,7 +54,7 @@ RGB 输入
 
 要自己指定,给一个十六进制先验:
 
-```powershell
+```bash
 ... --screen-color "#CFEFFF"
 ```
 
@@ -64,26 +64,23 @@ RGB 输入
 
 ## 环境
 
-使用 conda 环境 `lightML`。
+任意 Python 3.11–3.13 环境（venv 或 conda）均可：
 
-```powershell
-conda activate lightML
+```bash
 pip install -r requirements.txt
+python -m bgfilter.cli --help
 ```
 
-若未激活 shell,直接调用该环境的 Python:
-
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.cli --help
-```
+服务器安装（CPU wheel 版 torch、离线权重、systemd）请直接按 [DEPLOY_ZH.md](DEPLOY_ZH.md)
+走——同样的依赖，但把生产环境的取舍都写明了。
 
 ## 单张图
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.cli `
-  --input Samples\TestImage.png `
-  --output Outputs\TestImage_rgba.png `
-  --debug-dir Outputs\TestImage_debug
+```bash
+python -m bgfilter.cli \
+  --input Samples/TestImage.png \
+  --output Outputs/TestImage_rgba.png \
+  --debug-dir Outputs/TestImage_debug
 ```
 
 无需 `--config`——内置默认值与 `configs/default.yaml` 完全一致。只有当你编辑了该文件
@@ -97,42 +94,30 @@ fp32;大输入还会默认对 ViTMatte 全局注意力按 query 分块——精
 
 ## 批量
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.cli `
-  --input-dir Samples `
-  --output-dir Outputs `
-  --debug-dir Outputs\debug
+```bash
+python -m bgfilter.cli \
+  --input-dir Samples \
+  --output-dir Outputs \
+  --debug-dir Outputs/debug
 ```
 
-## 性能(CPU 参考数据)
+## 性能
 
-在 Ryzen 9700X(Zen 5,原生 bf16)、32 GB 内存上实测,cross-check 开启、
-**`--cross-check-as-seg` 开启**、分块全局注意力开启。(本分支默认关闭复用;不开的话要
-加上主 BiRefNet@1024 前向——在这些尺寸下每图约 +20 s。)
-
-| 输入 | 精度 | warm / 每图 | 峰值内存 |
-|---|---|---|---|
-| 1024x1536 | fp32 | ~45 s | — |
-| 1024x1536 | bf16 | ~33 s | — |
-| 2048x2048 | bf16 | ~53 s | 11.4 GB(批处理)、8.1 GB(单图) |
-
-内存上限来自 cross-check 的 BiRefNet_HR 前向(无论图多大都固定 `input_size` 2048)——
-一个全分辨率密集预测网络的真实活跃激活。在分块注意力和 mimalloc 修复之前,同一个
-2048x2048 bf16 run 峰值 25.2 GB;完全未优化的 fp32 run 估计要 55-60 GB(单是 ViTMatte
-未分块的 N^2 注意力就 ~39 GB)。无原生 bf16 的 CPU(如 Zen 2)会自动回落 fp32——这类
-机器上若内存或延迟紧张,优先 `--no-cross-check`。
+量级参考：快速桌面 CPU（bf16、cross-check 开）约 **33–53 s/图**，现代 GPU 约
+**1–2 s/图**。全部实测数据——CPU 各 lane、GPU 显存占用、内存上限及相关旋钮——
+见 [DEPLOY_ZH.md](DEPLOY_ZH.md) §11（容量与性能）。
 
 ## Chroma-alpha 调试模式
 
 `--matting-method chroma` 跳过 ViTMatte,直接用 chroma 置信度作为 alpha 种子。适合快速
 检查 chroma 置信度、trimap 和 despill。(不同于上面的仅 Chroma *管线*,后者仍跑 ViTMatte。)
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.cli `
-  --input-dir Samples `
-  --output-dir Outputs\chroma `
-  --debug-dir Outputs\chroma_debug `
-  --matting-method chroma `
+```bash
+python -m bgfilter.cli \
+  --input-dir Samples \
+  --output-dir Outputs/chroma \
+  --debug-dir Outputs/chroma_debug \
+  --matting-method chroma \
   --device cpu
 ```
 
@@ -209,9 +194,9 @@ python -m uvicorn app:app --host 127.0.0.1 --port 18083 --workers 1
 | 码 | 何时 |
 | --- | --- |
 | `200` | 成功——响应体是 RGBA PNG。 |
-| `400` | 输入有误:图片无法解码,`screen_color`/`seg_model`/`cross_check`/`background_mode` 非法,或自动探测失败(请传 `screen_color`)。 |
+| `400` | 输入有误:图片无法解码,`file`/`image_url` 不是恰好其一,`image_url` 非法/fetch 失败,`screen_color`/`seg_model`/`cross_check`/`background_mode` 非法,或自动探测失败(请传 `screen_color`)。 |
 | `401` | 鉴权已开且 token 缺失或错误。 |
-| `413` | 图片超过 `BGFILTER_MAX_IMAGE_PIXELS`(默认约 4 MP)。 |
+| `413` | 图片超过 `BGFILTER_MAX_IMAGE_PIXELS`(默认约 4 MP),或 `image_url` 响应体超过 `BGFILTER_FETCH_MAX_BYTES`(默认约 25 MB)。 |
 | `500` | 推理失败。 |
 | `503` | 服务仍在加载模型(未就绪)。 |
 
@@ -237,38 +222,42 @@ curl -sS \
   -F "file=@scene.jpg" \
   -F "background_mode=complex" \
   http://127.0.0.1:18083/remove-background -o output.png
+
+# 用 URL 代替上传(服务端自行 fetch;image_url 与 file 二选一,不可同时给)
+curl -sS \
+  -F "image_url=https://example.com/input.png" \
+  -F "screen_color=#CFEFFF" \
+  -H "X-Genarrative-Image-Token: <token>" \
+  http://127.0.0.1:18083/remove-background -o output.png
 ```
 
-配置来自 `BGFILTER_CONFIG`(存在时默认 `configs/default.yaml`);`BGFILTER_DEVICE` 同时
-覆盖 model 和 segmentation 的 device;`BGFILTER_MAX_IMAGE_PIXELS` 限制输入尺寸
-(默认约 4MP → `413`);`BGFILTER_PRELOAD=1` 启动时预加载默认模型;`BGFILTER_AUTH_TOKEN`
-控制端点鉴权(见上方"鉴权")。
+运行时配置(配置文件、device、尺寸上限、预加载、鉴权 token、流水线 worker 数)全部
+通过 `BGFILTER_*` 环境变量——完整表格见 [DEPLOY_ZH.md](DEPLOY_ZH.md) §6。
 
-与 CLI 一样,服务**默认开启交叉模型否决**(配置默认:on);其 HR-matting 模型随暖机集
-一起预加载。用 `BGFILTER_CROSS_CHECK=0` 全服务关闭,或用按请求的 `cross_check` 表单
-字段任意覆盖;`/healthz` 以 `crossCheck` 报告默认值。若配置里 `cross_check.reuse_as_seg:
-true`(默认关闭),HR-matting 前向会替代主分割前向而非叠加,那个冗余的主分割器会在启动
-时跳过。
+与 CLI 一样,服务**默认开启交叉模型否决**;`/healthz` 以 `crossCheck` 报告生效的默认值,
+按请求的 `cross_check` 表单字段可任意覆盖(全服务关闭用 `BGFILTER_CROSS_CHECK=0`)。
+`cross_check.reuse_as_seg`(仅配置文件可设,默认关——见上方"管线")没有请求级开关;
+改它意味着改配置并重启服务。
 
 ## 质量检查
 
 质量检查器在半透明边缘像素上度量 alpha 有效性和边缘溢色。
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe -m bgfilter.quality_cli `
-  Outputs\TestImage_rgba.png `
+```bash
+python -m bgfilter.quality_cli \
+  Outputs/TestImage_rgba.png \
   --max-edge-green-excess-p95 0.30
 ```
 
 跑内置样本冒烟检查:
 
-```powershell
-D:\MiniConda\envs\lightML\python.exe scripts\smoke_samples.py `
-  --samples-dir Samples `
-  --output-dir Outputs\smoke_samples `
-  --config configs\default.yaml `
-  --device cpu `
-  --fallback-to-chroma-alpha `
+```bash
+python scripts/smoke_samples.py \
+  --samples-dir Samples \
+  --output-dir Outputs/smoke_samples \
+  --config configs/default.yaml \
+  --device cpu \
+  --fallback-to-chroma-alpha \
   --max-edge-green-excess-p95 0.30
 ```
 
@@ -278,22 +267,9 @@ D:\MiniConda\envs\lightML\python.exe scripts\smoke_samples.py `
 - ViTMatte 和分割权重在首次使用时从 Hugging Face 加载。`anime-seg`
   (`skytnt/anime-seg`)是普通 ONNX 下载;`birefnet`(`ZhengPeng7/BiRefNet`)自带
   自定义建模代码,故需 `trust_remote_code=True` 外加 `timm` / `einops` / `kornia`。
-- 防火墙后下载权重——可靠组合是 `hf-mirror.com` 镜像 **+ 绕过本地代理** **+ 禁用 Xet**:
-  ```powershell
-  $env:HF_ENDPOINT      = "https://hf-mirror.com"  # 国内镜像
-  $env:NO_PROXY         = "*"                       # 绕过代理;镜像是直连
-  $env:HF_HUB_DISABLE_XET = "1"                     # 这些仓库是 Xet 后端;强制走经典 HTTP
-  ```
-  规避两个坑:(1) 镜像 + 海外代理会让镜像的 `resolve` 弹回 `huggingface.co`,新版
-  `huggingface_hub` 会以 `FileMetadataError` 拒绝;(2) 装了 `hf-xet` 时 Xet 下载路径会
-  立即失败。权重缓存好后,用 `HF_HUB_OFFLINE=1 TRANSFORMERS_OFFLINE=1` 离线运行
-  (生产环境即如此)。
-- 把权重打包进项目——不走 HF 缓存,而是把每个模型放进 `models/` 下、以仓库 basename
-  命名的纯文件夹:`models/vitmatte-base-composition-1k`、`models/BiRefNet`、
-  `models/anime-seg`、`models/BiRefNet_HR-matting`(交叉否决第二意见,默认开启)。加载器
-  优先匹配本地文件夹,缺失时回落到 HF 仓库 id / 缓存,故为可选。用例如
-  `hf download ZhengPeng7/BiRefNet --local-dir models/BiRefNet` 填充。用
-  `BGFILTER_WEIGHTS_DIR` 改基目录。`models/` 已被 gitignore。
+- **权重准备**——离线 HF 缓存、`models/` 纯文件夹(`BGFILTER_WEIGHTS_DIR`)、一键下载
+  脚本 `scripts/fetch_weights.py`、以及国内网络的镜像 workaround——见
+  [DEPLOY_ZH.md](DEPLOY_ZH.md) §5。
 - 前景颜色估计用 pymatting 的 `estimate_foreground_ml`,在 de-spill 前把干净前景色
   传播进半透明边缘。设 `foreground.method: unmix` 回落到旧的启发式方法。
 - `docs/green_screen_matting_workflow.md` 是最初的第一阶段绿幕规格;本 README 反映当前