coding-proxy 0.2.1a2__tar.gz → 0.2.2__tar.gz

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/CHANGELOG.md

@@ -3,20 +3,16 @@
 本文件基于 [Keep a Changelog](https://keepachangelog.com/zh-CN/) 规范维护，版本号遵循 [语义化版本](https://semver.org/lang/zh-CN/)。
 
 ## [Unreleased]
-## [v0.2.1](https://github.com/ThreeFish-AI/coding-proxy/releases/tag/v0.2.1a1) — 2026-04-11
 
-- fix(antigravity): 修复 Google OAuth token 刷新后 scope 校验过严导致 403 的问题;
-- refactor(request-normalizer): 移除跨供应商 tool_result 重定位死代码，docstring 对齐实际剥离行为;
+## [v0.2.2](https://github.com/ThreeFish-AI/coding-proxy/releases/tag/v0.2.2) — 2026-04-13
 
+- feat(reset): CLI reset 命令新增 -v/--vendor 参数，支持运行时 N-tier 链路重排序（逗号分隔的 vendor 列表）;
+- fix(logging): 修复 uvicorn.error 日志在文件中重复打印的问题;
 
-- **请求规范化死代码清理**：移除 `request_normalizer` 中永远不可达的 `tool_result` 重定位逻辑（Phase 1 剥离已覆盖全部场景，Phase 2 重定位因方向错误无法启用）；更新 docstring 使之与实际行为（剥离）一致；
+## [v0.2.1](https://github.com/ThreeFish-AI/coding-proxy/releases/tag/v0.2.1) — 2026-04-11
 
-### 🐛 Bug 修复
-
-- **fix(antigravity)**: 新增 Google Cloud Code **v1internal 协议支持**，彻底解决 `ACCESS_TOKEN_SCOPE_INSUFFICIENT` (403) 问题。
-  - **根因**：此前调用标准 Generative Language API (`generativelanguage.googleapis.com`)，该端点对 OAuth scope 校验严格；参考项目 Antigravity-Manager 实际使用的是 Cloud Code v1internal 内部 API (`cloudcode-pa.googleapis.com/v1internal`)，接受相同凭证但协议格式不同
-  - **修复**：新增 `project_id` 配置字段，当配置 `project_id` 且 `base_url` 含 `v1internal` 时自动启用 v1internal 模式：请求体包裹在 v1internal 信封中（含 `project`/`requestId`/`model`/`userAgent` 字段），注入客户端指纹 Headers（`x-client-name`/`x-client-version`/Chrome-Electron UA），端点 URL 适配 v1internal 格式
-  - **附带改进**：`_acquire()` scope 校验保持 warning 降级（与 Antigravity-Manager 行为对齐）；`_mark_scope_error_if_needed()` 增强诊断日志，提示用户切换 v1internal 协议
+- feat(logging): 实现日志双写（控制台 + 本地文件），日志文件支持 5MB 自动轮转及 gzip 压缩备份；ModelCall 日志降级为 DEBUG 级别；
+- feat(circuit-breaker): 补全熔断器状态转换日志的 vendor 上下文信息;
 
 ## [v0.2.0](https://github.com/ThreeFish-AI/coding-proxy/releases/tag/v0.2.0) — 2026-04-09
 

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: coding-proxy
-Version: 0.2.1a2
+Version: 0.2.2
 Summary: A High-Availability, Transparent, and Smart Multi-Vendor Proxy for Claude Code. Support Claude Plans, GitHub Copilot, Google Antigravity, ZAI/GLM, MiniMax, Qwen, Xiaomi, Kimi, Doubao...
 Project-URL: Source Code, https://github.com/ThreeFish-AI/coding-proxy
 Project-URL: User Guide, https://github.com/ThreeFish-AI/coding-proxy/blob/master/docs/user-guide.md
@@ -56,7 +56,7 @@ When you're deeply immersed in your coding "zone" with **Claude Code** (or any A
 
 ## 🌟 Core Features
 
-- **⛓️ N-tier Chained Failover**: Automatically downgrades from official Claude Plans, gracefully falling back to GitHub Copilot, then Google Antigravity, with Zhipu GLM acting as the ultimate safety net.
+- **⛓️ N-tier Chained Failover**: Autonomous descending sequence, supporting Claude's official plans, as well as Coding Plans from GitHub Copilot, Z AI, MiniMax, Alibaba Qwen, Xiaomi, Kimi, Doubao, etc.
 - **🛡️ Smart Resilience & Quota Guardians**: Every single vendor node comes fully armed with an independent **Circuit Breaker** and **Quota Guard** to proactively dodge avalanches without breaking a sweat.
 - **👻 Phantom-like Transparency**: **100% transparent** to the client! No code tweaks required. Overwrite `ANTHROPIC_BASE_URL` with a single line, and you're good to go.
 - **🔄 Universal Alchemy (Formats & Models)**: Native support for two-way request/streaming (SSE) translations between Anthropic ←→ Gemini. Plus, auto/DIY model name mapping (e.g., effortlessly morphing `claude-*` into `glm-*`).

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/README.md

@@ -29,7 +29,7 @@ When you're deeply immersed in your coding "zone" with **Claude Code** (or any A
 
 ## 🌟 Core Features
 
-- **⛓️ N-tier Chained Failover**: Automatically downgrades from official Claude Plans, gracefully falling back to GitHub Copilot, then Google Antigravity, with Zhipu GLM acting as the ultimate safety net.
+- **⛓️ N-tier Chained Failover**: Autonomous descending sequence, supporting Claude's official plans, as well as Coding Plans from GitHub Copilot, Z AI, MiniMax, Alibaba Qwen, Xiaomi, Kimi, Doubao, etc.
 - **🛡️ Smart Resilience & Quota Guardians**: Every single vendor node comes fully armed with an independent **Circuit Breaker** and **Quota Guard** to proactively dodge avalanches without breaking a sweat.
 - **👻 Phantom-like Transparency**: **100% transparent** to the client! No code tweaks required. Overwrite `ANTHROPIC_BASE_URL` with a single line, and you're good to go.
 - **🔄 Universal Alchemy (Formats & Models)**: Native support for two-way request/streaming (SSE) translations between Anthropic ←→ Gemini. Plus, auto/DIY model name mapping (e.g., effortlessly morphing `claude-*` into `glm-*`).

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/docs/zh-CN/README.md

@@ -29,7 +29,7 @@
 
 ## 🌟 核心特性 (Core Features)
 
-- **⛓️ N-tier 链式故障转移 (Failover)**：自动从 Claude 官方 Plans 依次降级至 GitHub Copilot、Google Antigravity，最后由智谱 GLM 兜底。
+- **⛓️ N-tier 链式故障转移 (Failover)**：自主降序序列，支持 Claude 官方 Plans，以及 GitHub Copilot、智谱、MiniMax、阿里千问、小米、Kimi、豆包等的 Coding Plan。
 - **🛡️ 智能弹性与容灾守卫**：每个供应商节点独立配备 **熔断器 (Circuit Breaker)** 与 **配额守卫 (Quota Guard)**，防雪崩、主动避险。
 - **👻 透明无感代理机制**：对客户端 **100% 透明**！无需修改任何代码，仅需一行配置覆盖 `ANTHROPIC_BASE_URL` 即可接入。
 - **🔄 跨模型与全格式转换**：原生支持 Anthropic ←→ Gemini 的请求与流式响应（SSE）双向转换，并支持自动/自助映射模型名称（如 `claude-*` 至 `glm-*`）。

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "coding-proxy"
-version = "0.2.1a2"
+version = "0.2.2"
 description = "A High-Availability, Transparent, and Smart Multi-Vendor Proxy for Claude Code. Support Claude Plans, GitHub Copilot, Google Antigravity, ZAI/GLM, MiniMax, Qwen, Xiaomi, Kimi, Doubao..."
 readme = "README.md"
 requires-python = ">=3.12"

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/cli/__init__.py

@@ -216,16 +216,50 @@ async def _run_usage(
 @app.command()
 def reset(
     port: int = typer.Option(8046, "--port", "-p", help="代理服务端口"),
+    vendor: str | None = typer.Option(
+        None,
+        "--vendor",
+        "-v",
+        help="提升/重排序 vendor 优先级（单个或逗号分隔多个）",
+    ),
 ) -> None:
-    """重置所有层级的熔断器和配额守卫（恢复使用最高优先级供应商）."""
+    """重置所有层级的熔断器和配额守卫.
+
+    可通过 -v 指定运行时 N-tier 链路重排序：
+
+    \b
+      -v zhipu               提升 zhipu 到最高优先级
+      -v zhipu,anthropic     替换整个 N-tier 链路顺序
+    """
     import httpx
 
+    # 构建请求 body
+    json_body: dict | None = None
+    if vendor:
+        parts = [v.strip() for v in vendor.split(",") if v.strip()]
+        if parts:
+            json_body = {"vendors": parts}
+
     try:
-        resp = httpx.post(f"http://127.0.0.1:{port}/api/reset", timeout=5)
+        resp = httpx.post(
+            f"http://127.0.0.1:{port}/api/reset",
+            json=json_body,
+            timeout=5,
+        )
         if resp.status_code == 200:
+            data = resp.json()
             console.print("[green]所有层级的熔断器和配额守卫已重置[/green]")
+            tier_order = data.get("tier_order")
+            if tier_order:
+                order_str = " → ".join(tier_order)
+                console.print(f"[cyan]当前链路顺序:[/] {order_str}")
         else:
-            console.print(f"[red]重置失败: {resp.status_code}[/red]")
+            try:
+                err = resp.json()
+                msg = err.get("error", {}).get("message", resp.text)
+            except Exception:
+                msg = resp.text
+            console.print(f"[red]重置失败: {msg}[/red]")
     except httpx.ConnectError:
         console.print("[red]代理服务未运行[/red]")
 

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/config/config.default.yaml

@@ -111,7 +111,7 @@ vendors:
     # 不配置 circuit_breaker → 自动成为终端层，不触发向下故障转移
     circuit_breaker:
       failure_threshold: 3
-      recovery_timeout_seconds: 300
+      recovery_timeout_seconds: 30
       success_threshold: 2
     quota_guard:
       enabled: true # 启用后按 Premium Requests 配额管理
@@ -421,6 +421,11 @@ pricing:
     input_cost_per_mtok: ¥0.80
     output_cost_per_mtok: ¥2.00
     cache_read_cost_per_mtok: ¥0.16
+  - vendor: zhipu
+    model: glm-4.7 # 待区分长短上下文定价
+    input_cost_per_mtok: ¥2.00
+    output_cost_per_mtok: ¥8.00
+    cache_read_cost_per_mtok: ¥0.40
   - vendor: zhipu
     model: glm-5v-turbo # 待区分长短上下文定价
     input_cost_per_mtok: ¥5.00

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/logging/__init__.py

@@ -1,6 +1,6 @@
 """日志模块.
 
-提供 uvicorn 兼容的 dictConfig 构建、JSON 结构化格式化器、
+提供 uvicorn 兼容的 dictConfig 构建、文件日志字符串格式化器、
 以及 gzip 压缩轮转支持。
 """
 
@@ -12,7 +12,7 @@ import logging.handlers
 import os
 from pathlib import Path
 
-from .formatters import JsonFormatter
+from .formatters import FileFormatter
 
 # ── 常量 ────────────────────────────────────────────────────────
 
@@ -83,7 +83,7 @@ def build_log_config(
 
     双写行为：
         - 控制台：人类可读格式，级别由 ``level`` 参数控制（handler 级别过滤）
-        - 文件：JSON 结构化格式，固定 DEBUG 级别（捕获所有日志）
+        - 文件：字符串格式（与控制台风格一致），固定 DEBUG 级别（捕获所有日志）
         - 当 ``file_path`` 为 ``None`` 或空字符串时，退化为纯控制台模式（向后兼容）
     """
     config: dict = {
@@ -118,7 +118,11 @@ def build_log_config(
         },
         "loggers": {
             "uvicorn": {"handlers": ["default"], "level": level, "propagate": False},
-            "uvicorn.error": {"level": level},
+            "uvicorn.error": {
+                "handlers": ["default"],
+                "level": level,
+                "propagate": False,
+            },
             "uvicorn.access": {
                 "handlers": ["access"],
                 "level": "INFO",
@@ -138,16 +142,16 @@ def build_log_config(
         log_file = Path(file_path)
         log_file.parent.mkdir(parents=True, exist_ok=True)
 
-        # 注入 JSON formatter
-        config["formatters"]["json"] = {
-            "()": "coding.proxy.logging.formatters.JsonFormatter",
+        # 注入文件日志字符串 formatter
+        config["formatters"]["file_fmt"] = {
+            "()": "coding.proxy.logging.formatters.FileFormatter",
         }
 
         # 注入 RotatingFileHandler（gzip 压缩轮转）
         # 使用工厂函数（而非 class + namer/rotator kwargs），
         # 因为 dictConfig 不支持将 namer/rotator 作为构造参数传递
         config["handlers"]["file"] = {
-            "formatter": "json",
+            "formatter": "file_fmt",
             "()": "coding.proxy.logging._create_rotating_file_handler",
             "filename": str(log_file.resolve()),
             "maxBytes": max_bytes,
@@ -182,7 +186,7 @@ def build_log_config(
 
 __all__ = [
     "build_log_config",
-    "JsonFormatter",
+    "FileFormatter",
     "_gzip_namer",
     "_gzip_rotator",
 ]

coding_proxy-0.2.2/src/coding/proxy/logging/formatters.py

@@ -0,0 +1,27 @@
+"""文件日志格式化器（字符串输出）.
+
+为文件日志提供人类可读的字符串格式输出，
+与控制台输出风格一致，便于人工阅读和 grep 检索。
+"""
+
+from __future__ import annotations
+
+import logging
+
+
+class FileFormatter(logging.Formatter):
+    """将 LogRecord 格式化为单行可读字符串.
+
+    输出格式：``2026-04-11 16:51:13 INFO  ModelCall: vendor=zhipu ...``
+
+    设计要点：
+    - 时间戳使用 ``yyyy-MM-dd HH:mm:ss`` 格式（与控制台一致）
+    - 日志级别左对齐 5 字符宽度，保证多行对齐美观
+    - 无 ANSI 颜色码（文件输出不需要终端转义）
+    """
+
+    def __init__(self) -> None:
+        super().__init__(
+            fmt="%(asctime)s %(levelname)-5s %(message)s",
+            datefmt="%Y-%m-%d %H:%M:%S",
+        )

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/routing/circuit_breaker.py

@@ -32,11 +32,14 @@ class CircuitBreaker:
         recovery_timeout_seconds: int = 300,
         success_threshold: int = 2,
         max_recovery_seconds: int = 3600,
+        *,
+        vendor_name: str = "",
     ) -> None:
         self._failure_threshold = failure_threshold
         self._recovery_timeout = recovery_timeout_seconds
         self._success_threshold = success_threshold
         self._max_recovery = max_recovery_seconds
+        self._vendor_label = f" [{vendor_name}]" if vendor_name else ""
 
         self._state = CircuitState.CLOSED
         self._failure_count = 0
@@ -65,7 +68,13 @@ class CircuitBreaker:
                 self._success_count += 1
                 if self._success_count >= self._success_threshold:
                     self._transition_to(CircuitState.CLOSED)
-                    logger.info("Circuit breaker: HALF_OPEN → CLOSED (recovered)")
+                    logger.info(
+                        "Circuit breaker%s: HALF_OPEN → CLOSED "
+                        "(recovered, %d/%d consecutive successes)",
+                        self._vendor_label,
+                        self._success_count,
+                        self._success_threshold,
+                    )
             elif self._state == CircuitState.CLOSED:
                 # 正常状态下成功，无需操作
                 pass
@@ -94,7 +103,10 @@ class CircuitBreaker:
                 self._transition_to(CircuitState.OPEN)
                 self._backoff_recovery(hint_seconds=retry_after_seconds)
                 logger.warning(
-                    "Circuit breaker: HALF_OPEN → OPEN (recovery failed, next retry in %ds)",
+                    "Circuit breaker%s: HALF_OPEN → OPEN "
+                    "(recovery probe failed, backoff %ds → next retry in %ds)",
+                    self._vendor_label,
+                    self._current_recovery,
                     self._current_recovery,
                 )
             elif self._state == CircuitState.CLOSED:
@@ -117,14 +129,17 @@ class CircuitBreaker:
                         )
                     if force_open:
                         logger.warning(
-                            "Circuit breaker: CLOSED → OPEN "
-                            "(forced, rate-limited, next retry in %ds)",
+                            "Circuit breaker%s: CLOSED → OPEN "
+                            "(forced, rate-limited, retry-after=%ss → next retry in %ds)",
+                            self._vendor_label,
+                            retry_after_seconds or "N/A",
                             self._current_recovery,
                         )
                     else:
                         logger.warning(
-                            "Circuit breaker: CLOSED → OPEN "
+                            "Circuit breaker%s: CLOSED → OPEN "
                             "(%d consecutive failures, next retry in %ds)",
+                            self._vendor_label,
                             self._failure_count,
                             self._current_recovery,
                         )
@@ -134,7 +149,9 @@ class CircuitBreaker:
         with self._lock:
             self._transition_to(CircuitState.CLOSED)
             self._current_recovery = self._recovery_timeout
-            logger.info("Circuit breaker: manually reset to CLOSED")
+            logger.info(
+                "Circuit breaker%s: manually reset to CLOSED", self._vendor_label
+            )
 
     def get_info(self) -> dict:
         """获取熔断器状态信息."""
@@ -157,7 +174,13 @@ class CircuitBreaker:
         elapsed = time.monotonic() - self._last_failure_time
         if elapsed >= self._current_recovery:
             self._transition_to(CircuitState.HALF_OPEN)
-            logger.info("Circuit breaker: OPEN → HALF_OPEN (recovery timeout)")
+            elapsed_s = int(elapsed)
+            logger.info(
+                "Circuit breaker%s: OPEN → HALF_OPEN (recovery timeout, waited %ds/%ds)",
+                self._vendor_label,
+                elapsed_s,
+                self._current_recovery,
+            )
 
     def _transition_to(self, new_state: CircuitState) -> None:
         self._state = new_state

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/routing/executor.py

@@ -48,9 +48,13 @@ logger = logging.getLogger(__name__)
 
 
 def _log_http_error_detail(
-    tier_name: str, exc: Exception, *, is_stream: bool = False
+    tier_name: str,
+    exc: Exception,
+    *,
+    is_stream: bool = False,
+    tier: VendorTier | None = None,
 ) -> None:
-    """记录 HTTP 错误的详细信息（状态码 / 响应体摘要 / 异常类型）.
+    """记录 HTTP 错误的详细信息（状态码 / 响应体摘要 / 异常类型 / 熔断器快照）.
 
     替代原先单行 ``logger.warning("Tier %s stream failed: %s", ...)``，
     在非 200 响应时输出更丰富的诊断上下文，便于跟踪上游故障根因。
@@ -78,6 +82,14 @@ def _log_http_error_detail(
                 detail_parts.append(f"  error_msg={err.get('message', 'N/A')[:200]}")
     else:
         detail_parts.append(f"  message={str(exc)[:300]}")
+    # 熔断器状态快照
+    if tier and tier.circuit_breaker:
+        cb = tier.circuit_breaker
+        cb_info = cb.get_info()
+        detail_parts.append(
+            f"  circuit_breaker={cb_info['state']} "
+            f"(failures={cb_info['failure_count']}/{cb._failure_threshold})"
+        )
     logger.warning("\n".join(detail_parts))
 
 
@@ -268,6 +280,12 @@ class _RouteExecutor:
                 duration = int((time.monotonic() - start) * 1000)
                 model = body.get("model", "unknown")
                 model_served = usage.get("model_served") or tier.vendor.map_model(model)
+                if failed_tier_name is not None:
+                    logger.info(
+                        "Tier %s stream succeeded (took over from failed tier: %s)",
+                        tier.name,
+                        failed_tier_name,
+                    )
                 self._recorder.log_model_call(
                     vendor=tier.name,
                     model_requested=model,
@@ -310,7 +328,7 @@ class _RouteExecutor:
                 httpx.ConnectError,
                 httpx.ReadError,
             ) as exc:
-                _log_http_error_detail(tier.name, exc, is_stream=True)
+                _log_http_error_detail(tier.name, exc, is_stream=True, tier=tier)
                 (
                     should_continue,
                     failed_tier_name,
@@ -325,6 +343,7 @@ class _RouteExecutor:
                     request_body=body,
                 )
                 if should_continue:
+                    self._log_failover_transition(tier, exc, self._tiers, i)
                     continue
                 if is_last:
                     raise
@@ -399,6 +418,12 @@ class _RouteExecutor:
                     duration = int((time.monotonic() - start) * 1000)
                     model = body.get("model", "unknown")
                     model_served = resp.model_served or tier.vendor.map_model(model)
+                    if failed_tier_name is not None:
+                        logger.info(
+                            "Tier %s message succeeded (took over from failed tier: %s)",
+                            tier.name,
+                            failed_tier_name,
+                        )
                     self._recorder.log_model_call(
                         vendor=tier.name,
                         model_requested=model,
@@ -469,10 +494,15 @@ class _RouteExecutor:
                         rate_limit_deadline=compute_rate_limit_deadline(rl_info),
                     )
                     if not is_last:
+                        next_tier = (
+                            self._tiers[i + 1] if i + 1 < len(self._tiers) else None
+                        )
+                        next_info = f" → next: {next_tier.name}" if next_tier else ""
                         logger.warning(
-                            "Tier %s error %d, failing over",
+                            "Tier %s error %d, failing over%s",
                             tier.name,
                             resp.status_code,
+                            next_info,
                         )
                         failed_tier_name = tier.name
                         continue
@@ -513,7 +543,7 @@ class _RouteExecutor:
                 continue
 
             except (httpx.TimeoutException, httpx.ConnectError, httpx.ReadError) as exc:
-                _log_http_error_detail(tier.name, exc, is_stream=False)
+                _log_http_error_detail(tier.name, exc, is_stream=False, tier=tier)
                 tier.record_failure()
                 failed_tier_name = tier.name
                 if is_last:
@@ -694,6 +724,30 @@ class _RouteExecutor:
 
         return False, tier.name, exc
 
+    @staticmethod
+    def _log_failover_transition(
+        current_tier: VendorTier,
+        exc: Exception,
+        tiers: list[VendorTier],
+        current_index: int,
+    ) -> None:
+        """记录 vendor 轮转摘要日志（谁 → 谁，原因）."""
+        next_tier = tiers[current_index + 1] if current_index + 1 < len(tiers) else None
+        if next_tier is None:
+            return
+
+        # 提取错误摘要
+        reason = type(exc).__name__
+        if isinstance(exc, httpx.HTTPStatusError) and exc.response is not None:
+            reason = f"HTTP {exc.response.status_code}"
+
+        logger.info(
+            "Failover: %s → %s (reason: %s)",
+            current_tier.name,
+            next_tier.name,
+            reason,
+        )
+
     @staticmethod
     def _is_cap_error(resp: VendorResponse) -> bool:
         """判断是否为订阅用量上限错误."""

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/routing/router.py

@@ -68,6 +68,66 @@ class RequestRouter:
         """当前活跃供应商名称（由 Executor 在成功响应时写入）."""
         return self._active_vendor_name
 
+    # ── 运行时 N-tier 链路重排序 ─────────────────────────────
+
+    def get_vendor_names(self) -> list[str]:
+        """返回当前 tiers 的供应商名称列表（按优先级顺序）."""
+        return [t.name for t in self._tiers]
+
+    def reorder_tiers(self, vendor_names: list[str]) -> None:
+        """原地重排序 N-tier 链路.
+
+        使用切片赋值保持列表引用同一性，使 ``_RouteExecutor`` 立即可见。
+
+        Args:
+            vendor_names: 新的供应商名称顺序（必须包含所有当前 tier）。
+
+        Raises:
+            ValueError: 名称不存在、有重复、或未覆盖所有 tier。
+        """
+        name_to_tier = {t.name: t for t in self._tiers}
+        current_names = set(name_to_tier)
+
+        # 校验：重复
+        if len(vendor_names) != len(set(vendor_names)):
+            seen: set[str] = set()
+            dups = [n for n in vendor_names if n in seen or seen.add(n)]  # type: ignore[func-returns-value]
+            raise ValueError(f"vendor 名称重复: {', '.join(dups)}")
+
+        # 校验：名称存在性
+        unknown = [n for n in vendor_names if n not in current_names]
+        if unknown:
+            raise ValueError(
+                f"未知 vendor: {', '.join(unknown)}; "
+                f"可用: {', '.join(sorted(current_names))}"
+            )
+
+        # 校验：全量覆盖
+        provided = set(vendor_names)
+        if provided != current_names:
+            missing = current_names - provided
+            raise ValueError(f"缺少 vendor: {', '.join(sorted(missing))}")
+
+        self._tiers[:] = [name_to_tier[n] for n in vendor_names]
+
+    def promote_vendor(self, vendor_name: str) -> None:
+        """将指定 vendor 提升至最高优先级，其余保持相对顺序.
+
+        Args:
+            vendor_name: 要提升的供应商名称。
+
+        Raises:
+            ValueError: 名称不存在。
+        """
+        current_names = self.get_vendor_names()
+        if vendor_name not in current_names:
+            available = sorted(t.name for t in self._tiers)
+            raise ValueError(
+                f"未知 vendor: {vendor_name}; 可用: {', '.join(available)}"
+            )
+        new_order = [vendor_name] + [n for n in current_names if n != vendor_name]
+        self.reorder_tiers(new_order)
+
     # ── 公开路由接口（委托给 _RouteExecutor）───────────────
 
     async def route_stream(

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/routing/tier.py

@@ -155,11 +155,33 @@ class VendorTier:
         if not is_probe_scenario:
             return cb_allows and qg_allows and wqg_allows
 
+        # 构建探测上下文摘要
+        probe_context_parts: list[str] = []
+        if self.circuit_breaker:
+            cb_info = self.circuit_breaker.get_info()
+            probe_context_parts.append(
+                f"circuit_breaker={cb_info['state']}, "
+                f"failures={cb_info['failure_count']}"
+            )
+        if self._rate_limit_deadline > 0:
+            waited = int(time.monotonic() - self._rate_limit_deadline)
+            probe_context_parts.append(f"rate_limit_waited={waited}s")
+        probe_context = (
+            " (" + ", ".join(probe_context_parts) + ")" if probe_context_parts else ""
+        )
+
         # ── 第二层: Health Check 门控 ──
-        logger.info("Tier %s: probe scenario, running health check", self.name)
+        logger.info(
+            "Tier %s: probe scenario%s, running health check",
+            self.name,
+            probe_context,
+        )
         healthy = await self.vendor.check_health()
         if not healthy:
-            logger.warning("Tier %s: health check failed, staying degraded", self.name)
+            logger.warning(
+                "Tier %s: health check failed, staying degraded",
+                self.name,
+            )
             self.record_failure()
             return False
 

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/server/app.py

@@ -106,7 +106,9 @@ def create_app(config: ProxyConfig | None = None) -> FastAPI:
             vendor_cfg, config.failover, mapper, token_store
         )
         cb = (
-            _build_circuit_breaker(vendor_cfg.circuit_breaker)
+            _build_circuit_breaker(
+                vendor_cfg.circuit_breaker, vendor_name=vendor_cfg.vendor
+            )
             if vendor_cfg.circuit_breaker
             else None
         )

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/server/factory.py

@@ -91,13 +91,16 @@ def _find_copilot_vendor(router: Any) -> CopilotVendor | None:
     return None
 
 
-def _build_circuit_breaker(cfg: CircuitBreakerConfig) -> CircuitBreaker:
+def _build_circuit_breaker(
+    cfg: CircuitBreakerConfig, *, vendor_name: str = ""
+) -> CircuitBreaker:
     """从配置构建熔断器实例."""
     return CircuitBreaker(
         failure_threshold=cfg.failure_threshold,
         recovery_timeout_seconds=cfg.recovery_timeout_seconds,
         success_threshold=cfg.success_threshold,
         max_recovery_seconds=cfg.max_recovery_seconds,
+        vendor_name=vendor_name,
     )
 
 

{coding_proxy-0.2.1a2 → coding_proxy-0.2.2}/src/coding/proxy/server/routes.py

@@ -261,7 +261,40 @@ def register_admin_routes(app: Any, router: Any) -> None:
     """注册管理操作路由（重置等）."""
 
     @app.post("/api/reset")
-    async def reset_circuit() -> dict:
+    async def reset_circuit(request: Request) -> Response:
+        """重置所有层级的熔断器/配额守卫/rate limit.
+
+        可选 JSON body ``{"vendors": ["v1", "v2", ...]}`` 支持运行时重排序：
+        - 单个 vendor → 提升至最高优先级，其余保持相对顺序
+        - 多个 vendor → 替换整个 N-tier 链路顺序（需覆盖所有 vendor）
+        """
+        # 解析可选 body
+        vendor_names: list[str] | None = None
+        try:
+            body = await request.json()
+            if isinstance(body, dict):
+                raw = body.get("vendors")
+                if isinstance(raw, list) and raw:
+                    vendor_names = [str(v) for v in raw]
+        except Exception:
+            # 无 body 或非 JSON → 仅 reset（向后兼容）
+            pass
+
+        # 重排序（如果指定）
+        if vendor_names is not None:
+            try:
+                if len(vendor_names) == 1:
+                    router.promote_vendor(vendor_names[0])
+                else:
+                    router.reorder_tiers(vendor_names)
+            except ValueError as exc:
+                return json_error_response(
+                    400,
+                    error_type="invalid_request_error",
+                    message=str(exc),
+                )
+
+        # 全量 reset
         for tier in router.tiers:
             if tier.circuit_breaker:
                 tier.circuit_breaker.reset()
@@ -270,7 +303,16 @@ def register_admin_routes(app: Any, router: Any) -> None:
             if tier.weekly_quota_guard:
                 tier.weekly_quota_guard.reset()
             tier.reset_rate_limit()
-        return {"status": "ok"}
+
+        result: dict[str, Any] = {"status": "ok"}
+        if vendor_names is not None:
+            result["tier_order"] = router.get_vendor_names()
+
+        return Response(
+            content=json.dumps(result, ensure_ascii=False).encode(),
+            status_code=200,
+            media_type="application/json",
+        )
 
 
 def register_reauth_routes(app: Any, reauth_coordinator: Any) -> None: