tripocli v2.0.0：Tripo AI 3D 生成与加工 CLI（面向 Agent 与开发者，纯 Python 标准库、零第三方依赖）

覆盖文生 3D、图生 3D、多视图、批量生成，以及重拓扑、绑定、动画、智能分割、
纹理、PBR、智能 UV、部件补全、格式导出与下载。

代码组织：一个功能一个模块（tripo/features/），入口薄壳 tripocli.py。

tripo/features/uv.py

@@ -0,0 +1,81 @@
+"""智能 UV — unfold UVs for a low-poly model.
+
+    tripocli uv -p <project_id> --wait
+
+实测流程（在页面上真操作抓到的请求）：
+
+    POST /v2/studio/operation/uv_edit/generate
+         {project_id, current_operator_id, action:"generate"}   → 生成 UV 候选（新 operator）
+    POST /v2/studio/operation/uv_edit/apply
+         {project_id, current_operator_id, candidate_operator_id}  → 保存/应用候选
+    POST /v2/studio/operation/uv_edit/context
+         {project_id, current_operator_id}                       → 查询当前处于哪一步
+
+所以这是**两步握手**：先 generate 出候选，再 apply 保存。只调 generate 不 apply，
+上下文会停在「待保存」，再调 generate 就会报 `2046 UV Edit action no longer matches the context`。
+
+另外服务端要求网格是低模（约 ≤8 万三角面），否则需要先 `tripocli remesh`。
+"""
+
+from . import _common
+
+NAME = "uv"
+HELP = "智能 UV：为低模展开 UV（需 ≤8 万三角面）"
+
+
+def add_arguments(parser):
+    _common.add_op_options(parser, "应用 UV 后的模型保存到该路径（可选）")
+    parser.add_argument("--action", default="generate", help="UV 操作动作（默认 generate）")
+    parser.add_argument("--generate-only", action="store_true", help="只生成候选，不自动保存应用")
+
+
+def run(ctx):
+    args = ctx.args
+    client = ctx.client
+
+    operator = client.project_operator(args.project)
+    current_operator_id = operator.get("operator_id")
+    if not current_operator_id:
+        raise ValueError(f"无法读取项目 {args.project} 的 operator_id")
+
+    generate_body = {
+        "project_id": args.project,
+        "current_operator_id": current_operator_id,
+        "action": args.action,
+    }
+    model_version = _common.project_version(client, args.project, args.model)
+    if model_version:
+        generate_body["model_version"] = model_version
+
+    ctx.out.line(f"🚀 生成 UV 候选 (current_operator_id={current_operator_id})...")
+    generated = client.get_data("POST", "/v2/studio/operation/uv_edit/generate", body=generate_body) or {}
+    candidate_id = generated.get("candidate_operator_id") or generated.get("operator_id")
+    if not candidate_id:
+        raise RuntimeError(f"generate 未返回候选 operator: {generated}")
+
+    if args.wait and candidate_id:
+        # The generate step produces a candidate; wait for it, but never hang on it —
+        # if it does not report a terminal state we still go on to apply.
+        try:
+            _common.wait_and_report(ctx, candidate_id, timeout=180)
+        except Exception as error:
+            ctx.out.line(f"（候选未在 180s 内就绪，继续尝试保存: {error}）")
+
+    if args.generate_only:
+        ctx.out.data(
+            {"generate": generated, "candidate_operator_id": candidate_id},
+            human=lambda: ctx.out.line(f"✅ 已生成候选（未保存）: {candidate_id}"),
+        )
+        return {"generate": generated, "candidate_operator_id": candidate_id}
+
+    apply_body = {
+        "project_id": args.project,
+        "current_operator_id": current_operator_id,
+        "candidate_operator_id": candidate_id,
+    }
+    ctx.out.line("💾 应用（保存）UV 候选...")
+    applied = client.get_data("POST", "/v2/studio/operation/uv_edit/apply", body=apply_body) or {}
+
+    return _common.run_operator(
+        ctx, applied, output=args.output, wait=args.wait, label="智能 UV"
+    )