flexgate 0.11.0__tar.gz → 0.11.2__tar.gz

{flexgate-0.11.0 → flexgate-0.11.2}/CLAUDE.md

@@ -95,7 +95,7 @@ OpenAI-style client → POST /v1/images/generations
 | `config.py` | Pydantic-like dataclasses (`GatewayConfig`, `ProviderConfig`, `RouteConfig`, `ScheduleEntry`), YAML load/save, `TIER_PATTERNS` regex map, `is_placeholder_key` |
 | `router.py` | `resolve(config, model)` — schedule-first then default routes, first regex match wins; model aliases normalized before matching |
 | `proxy.py` | `handle_request()` — httpx async proxy, per-key fallback retry loop, SSE streaming + JSON pass-through |
-| `usage.py` | `flexgate status` usage queries — per-platform adapters (MiniMax coding_plan API, Kimi Code usages API, GLM coding-plan/team quota APIs, LiteLLM `/key/info`, DeepSeek balance API) plus a minimal chat probe fallback; every adapter also fills a normalized `QuotaInfo` (plan + 5h/weekly `WindowQuota`: remaining percent, reset epoch) used by the unified summary and `--json`; unreported platform data ("?") stays None → displayed as 0% / no countdown |
+| `usage.py` | `flexgate status` usage queries — per-platform adapters (MiniMax coding_plan API, Kimi Code usages API, GLM coding-plan/team quota APIs, LiteLLM `/key/info`, DeepSeek balance API) plus a minimal chat probe fallback; every adapter also fills a normalized `QuotaInfo` (plan + 5h/weekly `WindowQuota`: remaining percent, reset epoch; plus a `balance` text for prepaid credit) used by the unified summary and `--json`; unreported platform data ("?") stays None → displayed as 0% / no countdown |
 | `server.py` | Starlette app creation, `POST /v1/messages` + `POST /v1/images/{generations,edits}` endpoints, `SIGUSR1` lifespan reload |
 | `images.py` | OpenAI Images API → MiniMax `/v1/image_generation` translation (`image-01`, size→width/height clamped to 512–2048 mult of 8, `data.image_base64[]`→`data[].b64_json`); edits endpoint returns 501; unknown image model names get 501 `model_not_supported` |
 | `fallback.py` | Catch-all route: merged Anthropic/OpenAI-shaped JSON errors for every unserved path — 404 `route_not_found` (lists valid routes), 405 `method_not_allowed`, 501 `endpoint_not_supported` per recognized API family with an actionable reason |
@@ -114,8 +114,8 @@ OpenAI-style client → POST /v1/images/generations
 - **Model alias normalization** (`router.py`): before matching, bare aliases (`sonnet`/`opus`/`haiku`/`default`, case-insensitive, optional `[plan]` suffix) map to `claude-<tier>` prefixes (`default` → sonnet, Claude Code's default tier), and legacy numbered names (`claude-3-7-sonnet-latest`) reduce to their tier prefix; everything else passes through unchanged, so custom patterns and the catch-all behave exactly as before.
 - **Model resolution & `available_models` fallback**: A route may omit `model`; `router.resolve()` then falls back to the provider's first `available_models` entry, so `model_override` handed to the proxy is always a concrete name. `config._parse_routes` rejects routes that omit `model` on a provider with no `available_models` — so adding a provider without models requires an explicit `model` on every route using it.
 - **Proxy rewrite contract** (`proxy.py`): the upstream request gets the provider's `x-api-key` plus a fixed header set, and the JSON `model` field is rewritten only when the route set an override. Streaming responses are forwarded as raw bytes (`aiter_bytes`), never parsed.
-- **Key fallback** (`proxy.py`): a provider's `api_keys` list holds one or more keys for the same upstream (entries are key strings or `{key, note}` mappings — plus optional `organization`/`project` on GLM team keys, used only by the usage check; `note` is a free-form label shown by `flexgate status` and in fallback logs). Multiple accounts on one upstream — e.g. several MiniMax subscriptions — belong in ONE provider, not separate ones. GLM coding plans are the documented exception: z.ai (api.z.ai) and zhipu (open.bigmodel.cn) share one account system — the same keys work on both endpoints — so the same key may legitimately sit under a `zai` and a `zhipu` provider (or `zai-team`/`zhipu-team`) to keep a fallback endpoint. Each *route* has an active-key pointer (`active_key` in YAML, 1-based; `RouteConfig.key_index` 0-based, default first key): requests on that route start from the pointed key, and on a retryable failure (HTTP 401/402/403/429/500/502/503/529, connect error, or timeout) the next key is tried and the pointer advances with it (in memory only, wrapping around — it is not written back to the file); every key is tried at most once and a full circle of failures returns the last error. The pointer is set interactively via `flexgate config edit` → "api keys" (pick a route, with live per-key usage) or by editing the route's `active_key` directly. `resolve()` (`router.py`) returns the matched `RouteConfig` so the proxy can read/advance its pointer. Streaming requests can only fall back before the upstream returns 200. `_regular_proxy`/`_stream_proxy` return `(response, retryable)`; `handle_request` owns the loop and returns the last error when all keys fail. The legacy `api_key` + `fallback_keys` schema is still parsed (`config._parse_api_keys`); migration v3 → v4 rewrites it to `api_keys`. `ProviderConfig.api_key` remains as a first-key property (getter/setter) for the settings import code.
-- **Usage inspection** (`usage.py`): `flexgate status` maps `base_url` to a platform adapter (`_ADAPTERS`): MiniMax → `{origin}/v1/api/openplatform/coding_plan/remains` (only the "general" text-model quota entry is used; its counts are usually 0/0, so usage is read from the `*_remaining_percent` fields), Kimi Code → `{base}/v1/usages` (undocumented, same endpoint as the CLI's `/usage`; reports weekly quota, 5-hour window and parallel limit), GLM (z.ai / bigmodel.cn, one account system — same keys on both endpoints) → adapter picked per key entry: without `organization`/`project` it is a personal coding-plan key → `{origin}/api/monitor/usage/quota/limit` (undocumented but used by z.ai's own plugin; both endpoints serve it), with them it is a team key → the same endpoint with `?type=2` plus `bigmodel-organization`/`bigmodel-project` headers and the team subscription detail for plan name/expiry (both endpoints serve it too), USTC/LiteLLM → `{origin}/key/info`, DeepSeek → `{origin}/user/balance`. Hosts in `_PROBE_ONLY_MARKERS` (e.g. xiaomimimo.com — no key-based usage API) and unknown platforms get the minimal chat probe: `POST /v1/messages` with input `"hi"`, `max_tokens=128`. A failed adapter call also degrades to the probe. Any key whose usage query errored — a hard failure (adapter AND probe), or an adapter error merely recovered by the probe — is recorded in `~/.flexgate/usage-cache.json` (sha256 fingerprint of `base_url|key`, never key material — scoped by endpoint so the same key under two providers caches independently; entry keeps the `ok` flag and output lines) and skipped by later `status`/`usage` runs, shown as `cached (… ago)` / `cached failure (… ago)` with a hint; `flexgate usage --force` rechecks them and a clean success clears the entry. Placeholder keys are never cached.
+- **Key fallback** (`proxy.py`): a provider's `api_keys` list holds one or more keys for the same upstream (entries are key strings or `{key, note}` mappings — plus optional `organization`/`project` on GLM team keys, used only by the usage check; `note` is a free-form label shown by `flexgate status` and in fallback logs). Multiple accounts on one upstream — e.g. several MiniMax subscriptions — belong in ONE provider, not separate ones. GLM coding plans are the documented exception: z.ai (api.z.ai) and zhipu (open.bigmodel.cn) share one account system — the same keys work on both endpoints, so a provider's `base_url` can be swapped freely when one endpoint is unstable (convention: one provider per key set, `zhipu`/`zhipu-team` on bigmodel). Each *route* has an active-key pointer (`active_key` in YAML, 1-based; `RouteConfig.key_index` 0-based, default first key): requests on that route start from the pointed key, and on a retryable failure (HTTP 401/402/403/429/500/502/503/529, connect error, or timeout) the next key is tried and the pointer advances with it (in memory only, wrapping around — it is not written back to the file); every key is tried at most once and a full circle of failures returns the last error. The pointer is set interactively via `flexgate config edit` → "api keys" (pick a route, with live per-key usage) or by editing the route's `active_key` directly. `resolve()` (`router.py`) returns the matched `RouteConfig` so the proxy can read/advance its pointer. Streaming requests can only fall back before the upstream returns 200. `_regular_proxy`/`_stream_proxy` return `(response, retryable)`; `handle_request` owns the loop and returns the last error when all keys fail. The legacy `api_key` + `fallback_keys` schema is still parsed (`config._parse_api_keys`); migration v3 → v4 rewrites it to `api_keys`. `ProviderConfig.api_key` remains as a first-key property (getter/setter) for the settings import code.
+- **Usage inspection** (`usage.py`): `flexgate status` maps `base_url` to a platform adapter (`_ADAPTERS`): MiniMax → `{origin}/v1/api/openplatform/coding_plan/remains` (only the "general" text-model quota entry is used; its counts are usually 0/0, so usage is read from the `*_remaining_percent` fields), Kimi Code → `{base}/v1/usages` (undocumented, same endpoint as the CLI's `/usage`; reports weekly quota, 5-hour window and parallel limit), GLM (z.ai / bigmodel.cn, one account system — same keys on both endpoints) → adapter picked per key entry: without `organization`/`project` it is a personal coding-plan key → `{origin}/api/monitor/usage/quota/limit` (undocumented but used by z.ai's own plugin; both endpoints serve it), with them it is a team key → the same endpoint with `?type=2` plus `bigmodel-organization`/`bigmodel-project` headers and the team subscription detail for plan name/expiry (both endpoints serve it too), USTC/LiteLLM → `{origin}/key/info`, DeepSeek → `{origin}/user/balance` (prepaid total per currency normalized into `QuotaInfo.balance`, shown next to the key in the unified summary). Hosts in `_PROBE_ONLY_MARKERS` (e.g. xiaomimimo.com — no key-based usage API) and unknown platforms get the minimal chat probe: `POST /v1/messages` with input `"hi"`, `max_tokens=128`. A failed adapter call also degrades to the probe. Any key whose usage query errored — a hard failure (adapter AND probe), or an adapter error merely recovered by the probe — is recorded in `~/.flexgate/usage-cache.json` (sha256 fingerprint of `base_url|key`, never key material — scoped by endpoint so the same key under two providers caches independently; entry keeps the `ok` flag and output lines) and skipped by later `status`/`usage` runs, shown as `cached (… ago)` / `cached failure (… ago)` with a hint; `flexgate usage --force` rechecks them and a clean success clears the entry. Placeholder keys are never cached.
 - **Multimodal degradation**: `MULTIMODAL_MODELS` (currently `{"MiniMax-M3", "glm-4.6v"}`) is the allowlist. Requests carrying image blocks aimed at any other model have images stripped and a `[flexgate]` text note injected into both the outgoing request and the returned response, so non-multimodal backends don't 4xx.
 - **Schedule-based overrides**: Optional time windows (e.g. 22:00-06:00) override default routes. Overnight wrap is supported.
 - **Service-first lifecycle**: `flexgate.service` is the sole persistent runtime. systemd owns restart, boot startup, logs, and process state.

{flexgate-0.11.0 → flexgate-0.11.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: flexgate
-Version: 0.11.0
+Version: 0.11.2
 Summary: Flexible local API gateway for Claude Code — route models to different providers
 Project-URL: Homepage, https://github.com/Agony5757/flexible-gateway
 Project-URL: Repository, https://github.com/Agony5757/flexible-gateway
@@ -204,7 +204,7 @@ Usage (timeout 15s per key)
 | GLM 个人版（api.z.ai / open.bigmodel.cn） | `GET /api/monitor/usage/quota/limit`（Authorization 直接带 key） | 未公开文档、但 z.ai 官方 coding 插件在用的接口，返回各窗口已用百分比与重置时间。z.ai 和智谱是同一账号体系，**两个端点都响应**；仅对个人版 coding plan 订阅有效，按量付费 key 退化为 probe |
 | GLM 团队版（api.z.ai / open.bigmodel.cn） | `GET /api/monitor/usage/quota/limit?type=2`，带 `bigmodel-organization`/`bigmodel-project` 头（取自 key 条目的 `organization`/`project` 字段） | 同一接口的团队版变体，返回 5 小时/每周窗口的积分用量；另查 `querySubscribeDetail` 显示套餐名与有效期，**两个端点都响应** |
 | USTC（api.llm.ustc.edu.cn，LiteLLM） | `GET /key/info`（Bearer 认证） | LiteLLM proxy 自带的 key 信息接口，返回 spend / max_budget / 限速等 |
-| DeepSeek（api.deepseek.com） | `GET /user/balance`（Bearer 认证） | 官方文档接口，返回预付费账户余额（总额 = 充值余额 + 未过期赠金，CNY/USD）；查询不消耗余额，明细见 `flexgate usage --verbose` |
+| DeepSeek（api.deepseek.com） | `GET /user/balance`（Bearer 认证） | 官方文档接口，返回预付费账户余额（总额 = 充值余额 + 未过期赠金，CNY/USD）；查询不消耗余额，统一摘要直接显示总额（如 `balance 489.61 CNY`），明细见 `flexgate usage --verbose` |
 | 小米 MiMo（token-plan-cn.xiaomimimo.com） | **无 key 可用的官方接口**（控制台内部接口需要浏览器 cookie，不采用） | 退化为 minimal probe |
 | 其他/未知平台 | minimal chat probe | 发一条输入 `"hi"`、`max_tokens=128` 的最小 `/v1/messages` 请求，验证 key 是否还能正常服务（会消耗极少量额度） |
 
@@ -353,8 +353,8 @@ server:
   port: 8765
 
 providers:
-  zai:                            # GLM 个人版 key，走 z.ai 端点
-    base_url: "https://api.z.ai/api/anthropic"
+  zhipu:                          # GLM 个人版 key（z.ai 端点同体系，base_url 可换）
+    base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
       - "your-glm-personal-api-key"
   minimax:
@@ -365,10 +365,6 @@ providers:
       - key: "your-minimax-api-key-2"
         note: "备用账号"
       - "your-minimax-api-key-3"   # 纯字符串写法（无备注）
-  zhipu:                          # 同一把 GLM 个人版 key 也可走智谱端点（同一账号体系）
-    base_url: "https://open.bigmodel.cn/api/anthropic"
-    api_keys:
-      - "your-glm-personal-api-key"
   zhipu-team:                     # GLM 团队版 key
     base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
@@ -376,13 +372,6 @@ providers:
         note: "队友名"
         organization: "org-XXXXXXXX"  # GLM 团队版 key：组织/项目 ID
         project: "proj_XXXXXXXX"      # （flexgate usage 查团队套餐额度用）
-  zai-team:                       # 团队版 key 同样可用于 z.ai 端点
-    base_url: "https://api.z.ai/api/anthropic"
-    api_keys:
-      - key: "your-glm-team-api-key"
-        note: "队友名"
-        organization: "org-XXXXXXXX"
-        project: "proj_XXXXXXXX"
 
 claude_settings:
   default_opus_model: "claude-opus-4-7"
@@ -397,12 +386,12 @@ claude_settings:
 #     end: "06:00"
 #     routes:
 #       - pattern: "^claude-sonnet"
-#         provider: zai
+#         provider: zhipu
 #         model: "glm-5.1"
 
 routes:                          # 从上到下匹配，首个命中生效
   - pattern: "^claude-opus"
-    provider: zai
+    provider: zhipu
     model: "glm-5.1"            # 可选，发给 provider 的实际模型名
   - pattern: "^claude-sonnet"
     provider: minimax

{flexgate-0.11.0 → flexgate-0.11.2}/README.md

@@ -187,7 +187,7 @@ Usage (timeout 15s per key)
 | GLM 个人版（api.z.ai / open.bigmodel.cn） | `GET /api/monitor/usage/quota/limit`（Authorization 直接带 key） | 未公开文档、但 z.ai 官方 coding 插件在用的接口，返回各窗口已用百分比与重置时间。z.ai 和智谱是同一账号体系，**两个端点都响应**；仅对个人版 coding plan 订阅有效，按量付费 key 退化为 probe |
 | GLM 团队版（api.z.ai / open.bigmodel.cn） | `GET /api/monitor/usage/quota/limit?type=2`，带 `bigmodel-organization`/`bigmodel-project` 头（取自 key 条目的 `organization`/`project` 字段） | 同一接口的团队版变体，返回 5 小时/每周窗口的积分用量；另查 `querySubscribeDetail` 显示套餐名与有效期，**两个端点都响应** |
 | USTC（api.llm.ustc.edu.cn，LiteLLM） | `GET /key/info`（Bearer 认证） | LiteLLM proxy 自带的 key 信息接口，返回 spend / max_budget / 限速等 |
-| DeepSeek（api.deepseek.com） | `GET /user/balance`（Bearer 认证） | 官方文档接口，返回预付费账户余额（总额 = 充值余额 + 未过期赠金，CNY/USD）；查询不消耗余额，明细见 `flexgate usage --verbose` |
+| DeepSeek（api.deepseek.com） | `GET /user/balance`（Bearer 认证） | 官方文档接口，返回预付费账户余额（总额 = 充值余额 + 未过期赠金，CNY/USD）；查询不消耗余额，统一摘要直接显示总额（如 `balance 489.61 CNY`），明细见 `flexgate usage --verbose` |
 | 小米 MiMo（token-plan-cn.xiaomimimo.com） | **无 key 可用的官方接口**（控制台内部接口需要浏览器 cookie，不采用） | 退化为 minimal probe |
 | 其他/未知平台 | minimal chat probe | 发一条输入 `"hi"`、`max_tokens=128` 的最小 `/v1/messages` 请求，验证 key 是否还能正常服务（会消耗极少量额度） |
 
@@ -336,8 +336,8 @@ server:
   port: 8765
 
 providers:
-  zai:                            # GLM 个人版 key，走 z.ai 端点
-    base_url: "https://api.z.ai/api/anthropic"
+  zhipu:                          # GLM 个人版 key（z.ai 端点同体系，base_url 可换）
+    base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
       - "your-glm-personal-api-key"
   minimax:
@@ -348,10 +348,6 @@ providers:
       - key: "your-minimax-api-key-2"
         note: "备用账号"
       - "your-minimax-api-key-3"   # 纯字符串写法（无备注）
-  zhipu:                          # 同一把 GLM 个人版 key 也可走智谱端点（同一账号体系）
-    base_url: "https://open.bigmodel.cn/api/anthropic"
-    api_keys:
-      - "your-glm-personal-api-key"
   zhipu-team:                     # GLM 团队版 key
     base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
@@ -359,13 +355,6 @@ providers:
         note: "队友名"
         organization: "org-XXXXXXXX"  # GLM 团队版 key：组织/项目 ID
         project: "proj_XXXXXXXX"      # （flexgate usage 查团队套餐额度用）
-  zai-team:                       # 团队版 key 同样可用于 z.ai 端点
-    base_url: "https://api.z.ai/api/anthropic"
-    api_keys:
-      - key: "your-glm-team-api-key"
-        note: "队友名"
-        organization: "org-XXXXXXXX"
-        project: "proj_XXXXXXXX"
 
 claude_settings:
   default_opus_model: "claude-opus-4-7"
@@ -380,12 +369,12 @@ claude_settings:
 #     end: "06:00"
 #     routes:
 #       - pattern: "^claude-sonnet"
-#         provider: zai
+#         provider: zhipu
 #         model: "glm-5.1"
 
 routes:                          # 从上到下匹配，首个命中生效
   - pattern: "^claude-opus"
-    provider: zai
+    provider: zhipu
     model: "glm-5.1"            # 可选，发给 provider 的实际模型名
   - pattern: "^claude-sonnet"
     provider: minimax

{flexgate-0.11.0 → flexgate-0.11.2}/config.yaml.template

@@ -24,32 +24,16 @@ providers:
     available_models:
       - "MiniMax-M3"
   # GLM Coding Plan：z.ai（api.z.ai）和智谱（open.bigmodel.cn）是同一账号
-  # 体系，同一把 key 在两个端点都可用，因此同一把 key 可以同时配在 zai 和
-  # zhipu 两个 provider 下（例如留一个备用端点）。provider 按端点选择；
+  # 体系，同一把 key 在两个端点都可用（哪个端点不稳就换 base_url）。
   # usage 检查按 key 条目区分：未配 organization/project 的是个人版
   # coding plan key，配了的是团队版 key。
-  zai:
-    base_url: "https://api.z.ai/api/anthropic"
-    api_keys:
-      - "your-glm-personal-api-key"
-    available_models:
-      - "glm-5.3"      # 路由省略 model 时的回退模型
-      - "glm-4.6v"
-  zai-team:
-    base_url: "https://api.z.ai/api/anthropic"
-    api_keys:
-      - key: "your-glm-team-api-key"
-        note: "队友名"
-        organization: "org-XXXXXXXX"    # GLM 团队版：组织 ID
-        project: "proj_XXXXXXXX"        # GLM 团队版：项目 ID
-    available_models:
-      - "glm-5.3"
   zhipu:
     base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
       - "your-glm-personal-api-key"
     available_models:
-      - "glm-5.3"
+      - "glm-5.3"      # 路由省略 model 时的回退模型
+      - "glm-4.6v"
   zhipu-team:
     base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
@@ -85,7 +69,7 @@ claude_settings:
 #     end: "24:00"
 #     routes:
 #       - pattern: "^claude-opus"
-#         provider: zai
+#         provider: zhipu
 #         model: "glm-5.3"
 #       - pattern: "^claude-sonnet"
 #         provider: xiaomi

{flexgate-0.11.0 → flexgate-0.11.2}/docs/cli.md

@@ -72,6 +72,9 @@ flexgate service uninstall           # 停止、禁用并删除 unit
 - `service reload` 和 `config set/edit` 会在仅路由变化时发送 SIGUSR1；
   如果 endpoint 变化，则先检查再 restart。若只改 host、仍复用当前 port，
   为避免误停服务会要求先执行 `service stop`，再执行 `service start`。
+- 网关进程自身每 2 秒监视 config.yaml：无论文件被谁修改（`flexgate config`、
+  编辑器、`flexgate sync pull`），路由变更都会在数秒内自动热重载生效，
+  不依赖编辑端能否触达 systemd；解析失败的半成品文件保留旧配置继续服务。
 - 查看日志：`flexgate log`（`-f` 跟随、`--since`/`--grep` 过滤、`-r` 只看路由决策行）。
 
 ## 状态总览与用量查询
@@ -211,7 +214,8 @@ Flexgate config  —  ~/.flexgate/config.yaml
   （与 `flexgate status` 相同），并标注当前 active 的 key。
 - model 列表包含：`available_models` 中的各个模型、「使用 provider 默认
   （首个可用模型，不写死 model）」、以及「自定义模型…」（手动输入）。
-- 按 `s` 保存（并向运行中的网关发送 SIGUSR1 热重载，**无需重启即生效**），
+- 按 `s` 保存（**无需重启即生效**：网关监视配置文件，数秒内自动热重载；
+  SIGUSR1 仍可用作立即触发），
   按 `q` 退出（有未保存改动时会询问 "Config changed — activate now?"：
   选 Yes 立即保存并热重载生效，选 No 放弃改动）；子菜单中按 `Esc`/`←`
   返回上一级。

{flexgate-0.11.0 → flexgate-0.11.2}/docs/configuration.md

@@ -25,8 +25,8 @@ server:
   port: 8765
 
 providers:
-  zai:                            # GLM 个人版 key，走 z.ai 端点
-    base_url: "https://api.z.ai/api/anthropic"
+  zhipu:                          # GLM 个人版 key（z.ai 端点同体系，base_url 可换）
+    base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
       - "your-glm-personal-api-key"
   minimax:
@@ -37,10 +37,6 @@ providers:
       - key: "your-minimax-api-key-2"
         note: "备用账号"
       - "your-minimax-api-key-3"   # 纯字符串写法（无备注）
-  zhipu:                          # 同一把 GLM 个人版 key 也可走智谱端点（同一账号体系）
-    base_url: "https://open.bigmodel.cn/api/anthropic"
-    api_keys:
-      - "your-glm-personal-api-key"
   zhipu-team:                     # GLM 团队版 key
     base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
@@ -48,13 +44,6 @@ providers:
         note: "队友名"
         organization: "org-XXXXXXXX"  # GLM 团队版 key：组织/项目 ID
         project: "proj_XXXXXXXX"      # （flexgate usage 查团队套餐额度用）
-  zai-team:                       # 团队版 key 同样可用于 z.ai 端点
-    base_url: "https://api.z.ai/api/anthropic"
-    api_keys:
-      - key: "your-glm-team-api-key"
-        note: "队友名"
-        organization: "org-XXXXXXXX"
-        project: "proj_XXXXXXXX"
 
 claude_settings:
   default_opus_model: "claude-opus-4-7"
@@ -69,12 +58,12 @@ claude_settings:
 #     end: "06:00"
 #     routes:
 #       - pattern: "^claude-sonnet"
-#         provider: zai
+#         provider: zhipu
 #         model: "glm-5.1"
 
 routes:                          # 从上到下匹配，首个命中生效
   - pattern: "^claude-opus"
-    provider: zai
+    provider: zhipu
     model: "glm-5.1"            # 可选，发给 provider 的实际模型名
   - pattern: "^claude-sonnet"
     provider: minimax

{flexgate-0.11.0 → flexgate-0.11.2}/docs/development.md

@@ -10,7 +10,7 @@
 | `flexgate/router.py` | `resolve(config, model)` — 先定时路由后默认路由，首个正则命中生效；匹配前归一化模型别名（`sonnet`/`claude-3-7-sonnet-latest` → `^claude-<tier>`） |
 | `flexgate/proxy.py` | `handle_request()` — httpx 异步代理，per-key fallback 重试，SSE 流式 + JSON 透传；记录每个请求由哪把 key 服务 |
 | `flexgate/usage.py` | `flexgate status` / `flexgate usage` 的用量查询 — 各平台适配器 + minimal chat probe 兜底 + 查询失败缓存（`usage --force` 强制重查） |
-| `flexgate/server.py` | Starlette 应用，`POST /v1/messages` + `POST /v1/images/{generations,edits}` 端点，SIGUSR1 热重载 |
+| `flexgate/server.py` | Starlette 应用，`POST /v1/messages` + `POST /v1/images/{generations,edits}` 端点，SIGUSR1 热重载 + 配置文件 watcher 自动重载 |
 | `flexgate/images.py` | OpenAI Images API → MiniMax `/v1/image_generation` 翻译（模型名映射、size 钳制、`b64_json` 透传；edits 返回 501） |
 | `flexgate/fallback.py` | 兜底路由：所有未服务路径返回 Anthropic/OpenAI 双兼容的结构化 JSON 错误（404 `route_not_found` / 405 / 501 `endpoint_not_supported`） |
 | `flexgate/main.py` | 启动引导：加载配置 → 创建 app → 跑 uvicorn |

{flexgate-0.11.0 → flexgate-0.11.2}/docs/usage.md

@@ -16,7 +16,7 @@
 | GLM 个人版（api.z.ai / open.bigmodel.cn） | `GET /api/monitor/usage/quota/limit`（Authorization 直接带 key） | 未公开文档、但 z.ai 官方 coding 插件在用的接口，返回各窗口已用百分比与重置时间。z.ai 和智谱（bigmodel.cn）是同一账号体系，**两个端点都响应**此接口。仅对 GLM Coding Plan 个人订阅有效：按量付费 key 或团队 key 会返回"当前用户不存在coding plan"并退化为 probe（团队 key 需在条目上配 `organization`/`project`） |
 | GLM 团队版（api.z.ai / open.bigmodel.cn） | `GET /api/monitor/usage/quota/limit?type=2`，带 `bigmodel-organization`/`bigmodel-project` 头（取自 key 条目的 `organization`/`project` 字段） | 同一 monitor 接口的团队版变体，返回 5 小时/每周窗口的积分用量（绝对值 + 百分比 + 重置时间）；另查 `querySubscribeDetail` 显示套餐名与有效期。**两个端点都响应**；key 条目未配 `organization`/`project` 时按个人版 key 处理 |
 | USTC（api.llm.ustc.edu.cn，LiteLLM） | `GET /key/info`（Bearer 认证） | LiteLLM proxy 自带的 key 信息接口，返回 spend / max_budget / 限速等 |
-| DeepSeek（api.deepseek.com） | `GET /user/balance`（Bearer 认证） | 官方文档接口，返回预付费账户余额（总额 = 充值余额 + 未过期赠金，按 CNY/USD 分别列出）；查询本身不消耗余额。余额不属于窗口配额，统一摘要中只显示方法名，具体金额见 `--verbose` / `--json --verbose` |
+| DeepSeek（api.deepseek.com） | `GET /user/balance`（Bearer 认证） | 官方文档接口，返回预付费账户余额（总额 = 充值余额 + 未过期赠金，按 CNY/USD 分别列出）；查询本身不消耗余额。余额不属于窗口配额，统一摘要在 key 旁直接显示总额（如 `balance 489.61 CNY`），充值/赠金明细见 `--verbose` |
 | 小米 MiMo（token-plan-cn.xiaomimimo.com） | **无 key 可用的官方接口**（控制台内部接口需要浏览器 cookie，不采用） | 退化为 minimal probe |
 | 其他/未知平台 | minimal chat probe | 发一条输入 `"hi"`、`max_tokens=128` 的最小 `/v1/messages` 请求，验证 key 是否还能正常服务（会消耗极少量额度） |
 
@@ -31,9 +31,11 @@
 
 ```text
 Usage (timeout 15s per key)
-  zai
+  zhipu
     ✓ #1 cbc4***sL7n [main]  plan pro   5h 88% · 4h 36m   weekly 81% · 6d 16h 15m
     ✓ #2 302b***RLBh [gsj]   plan max   5h 100% · —   weekly 99% · 6d 17h 46m
+  deepseek
+    ✓ #1 sk-c***9e13 [official]  balance 489.61 CNY
   ustc
     ✗ #1 sk-G***jJYg  LiteLLM /key/info failed (ConnectError: ); probe: ConnectError:
 ```
@@ -45,8 +47,8 @@ Usage (timeout 15s per key)
   - `dhm`（默认）：`4d 4h 23m` / `2h 47m` / `23m`
   - `hours`：`52.1h`
   - `minutes`：`3125m`
-- 无窗口配额的 key（probe / 缓存失败 / 仅余额型接口如 DeepSeek、LiteLLM）
-  只显示一行方法或错误摘要。
+- 无窗口配额的 key 只显示一行文本：probe / 缓存失败 / LiteLLM 显示方法
+  或错误摘要；余额型平台（DeepSeek）显示 `balance 489.61 CNY`。
 
 ### "?" 值的语义
 
@@ -122,6 +124,8 @@ Usage (timeout 15s per key)
 
 - 字段语义与统一摘要一致：余额未知 → `remaining_percent: 0.0`；
   重置时间未知 → `remaining_seconds`/`reset_at` 为 `null`。
+- DeepSeek 的余额在顶层 `balance` 字段（如 `"487.77 CNY"`，多币种以
+  ` · ` 连接）。
 - 平台额外信息保留在 `window_hours`（如 5.0）和 `used`（原始用量文本）。
 - key 一律掩码显示，永不输出明文。
 

{flexgate-0.11.0 → flexgate-0.11.2}/flexgate/__init__.py

@@ -1,3 +1,3 @@
 """Flexgate — Flexible API Gateway for Claude Code."""
 
-__version__ = "0.11.0"
+__version__ = "0.11.2"

{flexgate-0.11.0 → flexgate-0.11.2}/flexgate/cli.py

@@ -313,7 +313,12 @@ def _print_usage_summary(usage: dict, time_format: str) -> None:
             if quota is not None and (quota.plan or quota.five_hour or quota.weekly):
                 rows.append((r, key, quota.plan or "—", quota.five_hour, quota.weekly, None))
             else:
-                detail = r.method if r.ok else (r.lines[0] if r.lines else r.method)
+                if quota is not None and quota.balance:
+                    detail = f"balance {quota.balance}"
+                elif r.ok:
+                    detail = r.method
+                else:
+                    detail = r.lines[0] if r.lines else r.method
                 rows.append((r, key, None, None, None, detail))
         key_w = max(len(row[1]) for row in rows)
         plan_w = max((len(row[2]) for row in rows if row[2] is not None), default=0)

{flexgate-0.11.0 → flexgate-0.11.2}/flexgate/config.py

@@ -386,39 +386,23 @@ providers:
     available_models:
       - "MiniMax-M3"
   # GLM Coding Plan keys: z.ai (api.z.ai) and zhipu (open.bigmodel.cn) are
-  # one account system — the same keys work on both endpoints, so a key may
-  # be listed under a zai and a zhipu provider at the same time (e.g. to
-  # keep a fallback endpoint). Pick the provider by endpoint; the usage
+  # one account system — the same keys work on both endpoints, so either
+  # base_url can be used (swap it if one endpoint is unstable). The usage
   # adapter is picked per key entry: without organization/project it is a
   # personal coding-plan key, with them a team key.
-  zai:
-    base_url: "https://api.z.ai/api/anthropic"
-    api_keys:
-      - "your-glm-personal-api-key"
-    available_models:
-      - "glm-5.3"      # used as fallback when a route omits `model`
-      - "glm-4.6v"
-  zai-team:
-    base_url: "https://api.z.ai/api/anthropic"
-    api_keys:
-      - key: "your-glm-team-api-key"
-        note: "teammate name"
-        organization: "org-XXXXXXXX"   # GLM team IDs (see comment above)
-        project: "proj_XXXXXXXX"
-    available_models:
-      - "glm-5.3"
   zhipu:
     base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
       - "your-glm-personal-api-key"
     available_models:
-      - "glm-5.3"
+      - "glm-5.3"      # used as fallback when a route omits `model`
+      - "glm-4.6v"
   zhipu-team:
     base_url: "https://open.bigmodel.cn/api/anthropic"
     api_keys:
       - key: "your-glm-team-api-key"
         note: "teammate name"
-        organization: "org-XXXXXXXX"
+        organization: "org-XXXXXXXX"   # GLM team IDs (see comment above)
         project: "proj_XXXXXXXX"
     available_models:
       - "glm-5.3"
@@ -449,7 +433,7 @@ claude_settings:
 #     end: "24:00"
 #     routes:
 #       - pattern: "^claude-opus"
-#         provider: zai
+#         provider: zhipu
 #         model: "glm-5.3"
 #       - pattern: "^claude-sonnet"
 #         provider: xiaomi

{flexgate-0.11.0 → flexgate-0.11.2}/flexgate/server.py

@@ -1,7 +1,9 @@
 from __future__ import annotations
 
+import asyncio
 import json
 import logging
+import os
 import signal
 import time
 from collections.abc import AsyncGenerator
@@ -62,6 +64,56 @@ async def messages(request: Request) -> JSONResponse:
     return resp
 
 
+def _load_new_config(config_path: str) -> GatewayConfig | None:
+    try:
+        return load_config(config_path)
+    except Exception as e:
+        logger.error("Failed to reload config: %s", e)
+        return None
+
+
+# The gateway watches its config file so edits take effect within seconds no
+# matter how the file was changed (flexgate config, an editor, sync). The
+# CLI-side SIGUSR1 reload depends on the editing process reaching the systemd
+# user bus, which is not always available (SSH sessions, agent shells).
+CONFIG_WATCH_INTERVAL = 2.0
+
+
+async def _watch_config(app, config_path: str | None) -> None:
+    last_mtime: float | None = None
+    if config_path:
+        try:
+            last_mtime = os.path.getmtime(config_path)
+        except OSError:
+            pass
+    while True:
+        await asyncio.sleep(CONFIG_WATCH_INTERVAL)
+        try:
+            if not config_path:
+                continue
+            try:
+                mtime = os.path.getmtime(config_path)
+            except OSError:
+                continue
+            if last_mtime is None:
+                last_mtime = mtime
+                continue
+            if mtime == last_mtime:
+                continue
+            new_config = _load_new_config(config_path)
+            if new_config is not None:
+                app.state.config = new_config
+                logger.info("Config auto-reloaded (file changed)")
+            # Advance the marker even on failure so a broken file logs once,
+            # not every poll; the corrected save rewrites the file (new mtime)
+            # and reloads again.
+            last_mtime = mtime
+        except asyncio.CancelledError:
+            raise
+        except Exception:
+            logger.exception("Config watcher error")
+
+
 def create_app(config: GatewayConfig, config_path: str | None = None) -> Starlette:
     client = httpx.AsyncClient(
         timeout=httpx.Timeout(connect=10, read=3600, write=10, pool=10),
@@ -82,16 +134,19 @@ def create_app(config: GatewayConfig, config_path: str | None = None) -> Starlet
     async def lifespan(app: Starlette) -> AsyncGenerator[None, None]:
         app.state.config = config
         app.state.client = client
+        watcher = asyncio.create_task(_watch_config(app, config_path))
 
         def _handle_sigusr1(signum, frame):
-            try:
-                app.state.config = load_config(config_path)
+            new_config = _load_new_config(config_path)
+            if new_config is not None:
+                app.state.config = new_config
                 logger.info("Config reloaded via SIGUSR1")
-            except Exception as e:
-                logger.error("Failed to reload config: %s", e)
 
         signal.signal(signal.SIGUSR1, _handle_sigusr1)
-        yield
+        try:
+            yield
+        finally:
+            watcher.cancel()
         await client.aclose()
 
     return Starlette(routes=routes, lifespan=lifespan)

{flexgate-0.11.0 → flexgate-0.11.2}/flexgate/usage.py

@@ -29,16 +29,18 @@ queries it with the provider's own API key:
 * DeepSeek  → GET {origin}/user/balance
               (officially documented endpoint; reports the prepaid
               pay-as-you-go balance — topped-up plus granted — per
-              currency. A balance is not a windowed quota, so like
-              LiteLLM it stays in the raw lines only)
+              currency. A balance is not a windowed quota, so there are
+              no 5h/weekly windows; the total per currency is normalized
+              into ``QuotaInfo.balance`` and shown in the unified summary)
 * anything else (e.g. Xiaomi MiMo, which has no key-based usage API)
             → minimal chat probe: POST /v1/messages with input "hi" and
               max_tokens=128 to verify the key can still serve requests
 
 Alongside the human-readable ``lines``, every adapter fills a normalized
 ``QuotaInfo`` (plan name + 5-hour/weekly ``WindowQuota`` with remaining
-percent and reset epoch) so the CLI can render one unified summary and
-``--json`` consumers get structured fields. Data a platform did not report
+percent and reset epoch, plus a ``balance`` text for prepaid credit) so
+the CLI can render one unified summary and ``--json`` consumers get
+structured fields. Data a platform did not report
 (the "?" cases) stays None: an unknown balance is displayed as 0% and an
 unknown reset time means no countdown is active.
 """
@@ -96,6 +98,10 @@ class QuotaInfo:
     plan: str | None = None
     five_hour: WindowQuota | None = None
     weekly: WindowQuota | None = None
+    # Pay-as-you-go balance text (e.g. "489.61 CNY") — DeepSeek-style prepaid
+    # credit, not a windowed quota; shown in the unified summary next to the
+    # key instead of the method name.
+    balance: str | None = None
 
 
 @dataclass
@@ -507,6 +513,7 @@ async def _usage_deepseek(
     if err:
         return [], None, err
     lines: list[str] = []
+    balances: list[str] = []
     if not data.get("is_available"):
         lines.append("NO balance available for API calls")
     for info in data.get("balance_infos", []):
@@ -517,7 +524,9 @@ async def _usage_deepseek(
             f"(topped-up {info.get('topped_up_balance', '?')}, "
             f"granted {info.get('granted_balance', '?')})"
         )
-    return lines or ["no balance data"], None, None
+        balances.append(f"{info.get('total_balance', '?')} {info.get('currency', '?')}")
+    quota = QuotaInfo(balance=" · ".join(balances)) if data.get("is_available") and balances else None
+    return lines or ["no balance data"], quota, None
 
 
 async def _usage_kimi(
@@ -816,6 +825,7 @@ def result_to_dict(result: UsageResult, *, verbose: bool = False, now: float | N
         "method": result.method,
         "adapter_error": result.adapter_error,
         "plan": quota.plan if quota else None,
+        "balance": quota.balance if quota else None,
         "five_hour": window(quota.five_hour) if quota else None,
         "weekly": window(quota.weekly) if quota else None,
     }