fix(translator): correct funcName extraction and ensure proper handling of function response data in Antigravity Claude requests

Fix invalid thinking signature when proxying Claude via Antigravity

.dockerignore

@@ -28,3 +28,4 @@ bin/*
 .claude/*
 .vscode/*
 .serena/*
+.bmad/*

.gitignore

@@ -31,6 +31,7 @@ GEMINI.md
 .vscode/*
 .claude/*
 .serena/*
+.bmad/*
 
 # macOS
 .DS_Store

README.md

@@ -59,7 +59,7 @@ CLIProxyAPI includes integrated support for [Amp CLI](https://ampcode.com) and A
 - **Model mapping** to route unavailable models to alternatives (e.g., `claude-opus-4.5` → `claude-sonnet-4`)
 - Security-first design with localhost-only management endpoints
 
-**→ [Complete Amp CLI Integration Guide](docs/amp-cli-integration.md)**
+**→ [Complete Amp CLI Integration Guide](https://help.router-for.me/agent-client/amp-cli.html)**
 
 ## SDK Docs
 

README_CN.md

@@ -57,7 +57,7 @@ CLIProxyAPI 已内置对 [Amp CLI](https://ampcode.com) 和 Amp IDE 扩展的支
 - 智能模型回退与自动路由
 - 以安全为先的设计，管理端点仅限 localhost
 
-**→ [Amp CLI 完整集成指南](docs/amp-cli-integration_CN.md)**
+**→ [Amp CLI 完整集成指南](https://help.router-for.me/cn/agent-client/amp-cli.html)**
 
 ## SDK 文档
 

config.example.yaml

@@ -25,6 +25,9 @@ remote-management:
   # Disable the bundled management control panel asset download and HTTP route when true.
   disable-control-panel: false
 
+  # GitHub repository for the management control panel. Accepts a repository URL or releases API URL.
+  panel-github-repository: "https://github.com/router-for-me/Cli-Proxy-API-Management-Center"
+
 # Authentication directory (supports ~ for home directory)
 auth-dir: "~/.cli-proxy-api"
 
@@ -45,6 +48,9 @@ usage-statistics-enabled: false
 # Proxy URL. Supports socks5/http/https protocols. Example: socks5://user:pass@192.168.1.1:1080/
 proxy-url: ""
 
+# When true, unprefixed model requests only use credentials without a prefix (except when prefix == model name).
+force-model-prefix: false
+
 # Number of times to retry a request. Retries will occur if the HTTP response code is 403, 408, 500, 502, 503, or 504.
 request-retry: 3
 
@@ -62,6 +68,7 @@ ws-auth: false
 # Gemini API keys
 # gemini-api-key:
 #   - api-key: "AIzaSy...01"
+#     prefix: "test" # optional: require calls like "test/gemini-3-pro-preview" to target this credential
 #     base-url: "https://generativelanguage.googleapis.com"
 #     headers:
 #       X-Custom-Header: "custom-value"
@@ -76,6 +83,7 @@ ws-auth: false
 # Codex API keys
 # codex-api-key:
 #   - api-key: "sk-atSM..."
+#     prefix: "test" # optional: require calls like "test/gpt-5-codex" to target this credential
 #     base-url: "https://www.example.com" # use the custom codex API endpoint
 #     headers:
 #       X-Custom-Header: "custom-value"
@@ -90,6 +98,7 @@ ws-auth: false
 # claude-api-key:
 #   - api-key: "sk-atSM..." # use the official claude API key, no need to set the base url
 #   - api-key: "sk-atSM..."
+#     prefix: "test" # optional: require calls like "test/claude-sonnet-latest" to target this credential
 #     base-url: "https://www.example.com" # use the custom claude API endpoint
 #     headers:
 #       X-Custom-Header: "custom-value"
@@ -106,6 +115,7 @@ ws-auth: false
 # OpenAI compatibility providers
 # openai-compatibility:
 #   - name: "openrouter" # The name of the provider; it will be used in the user agent and other places.
+#     prefix: "test" # optional: require calls like "test/kimi-k2" to target this provider's credentials
 #     base-url: "https://openrouter.ai/api/v1" # The base URL of the provider.
 #     headers:
 #       X-Custom-Header: "custom-value"
@@ -120,6 +130,7 @@ ws-auth: false
 # Vertex API keys (Vertex-compatible endpoints, use API key + base URL)
 # vertex-api-key:
 #   - api-key: "vk-123..."                        # x-goog-api-key header
+#     prefix: "test"                              # optional: require calls like "test/vertex-pro" to target this credential
 #     base-url: "https://example.com/api"         # e.g. https://zenmux.ai/api
 #     proxy-url: "socks5://proxy.example.com:1080" # optional per-key proxy override
 #     headers:

docs/amp-cli-integration.md

@@ -1,443 +0,0 @@
-# Amp CLI Integration Guide
-
-This guide explains how to use CLIProxyAPI with Amp CLI and Amp IDE extensions, enabling you to use your existing Google/ChatGPT/Claude subscriptions (via OAuth) with Amp's CLI.
-
-## Table of Contents
-
-- [Overview](#overview)
-  - [Which Providers Should You Authenticate?](#which-providers-should-you-authenticate)
-- [Architecture](#architecture)
-- [Configuration](#configuration)
-  - [Model Mapping Configuration](#model-mapping-configuration)
-- [Setup](#setup)
-- [Usage](#usage)
-- [Troubleshooting](#troubleshooting)
-
-## Overview
-
-The Amp CLI integration adds specialized routing to support Amp's API patterns while maintaining full compatibility with all existing CLIProxyAPI features. This allows you to use both traditional CLIProxyAPI features and Amp CLI with the same proxy server.
-
-### Key Features
-
-- **Provider route aliases**: Maps Amp's `/api/provider/{provider}/v1...` patterns to CLIProxyAPI handlers
-- **Management proxy**: Forwards OAuth and account management requests to Amp's control plane
-- **Smart fallback**: Automatically routes unconfigured models to ampcode.com
-- **Model mapping**: Route unavailable models to alternatives you have access to (e.g., `claude-opus-4.5` → `claude-sonnet-4`)
-- **Secret management**: Configurable precedence (config > env > file) with 5-minute caching
-- **Security-first**: Management routes restricted to localhost by default
-- **Automatic gzip handling**: Decompresses responses from Amp upstream
-
-### What You Can Do
-
-- Use Amp CLI with your Google account (Gemini 3 Pro Preview, Gemini 2.5 Pro, Gemini 2.5 Flash)
-- Use Amp CLI with your ChatGPT Plus/Pro subscription (GPT-5, GPT-5 Codex models)
-- Use Amp CLI with your Claude Pro/Max subscription (Claude Sonnet 4.5, Opus 4.1)
-- Use Amp IDE extensions (VS Code, Cursor, Windsurf, etc.) with the same proxy
-- Run multiple CLI tools (Factory + Amp) through one proxy server
-- Route unconfigured models automatically through ampcode.com
-
-### Which Providers Should You Authenticate?
-
-**Important**: The providers you need to authenticate depend on which models and features your installed version of Amp currently uses. Amp employs different providers for various agent modes and specialized subagents:
-
-- **Smart mode**: Uses Google/Gemini models (Gemini 3 Pro)
-- **Rush mode**: Uses Anthropic/Claude models (Claude Haiku 4.5)
-- **Oracle subagent**: Uses OpenAI/GPT models (GPT-5 medium reasoning)
-- **Librarian subagent**: Uses Anthropic/Claude models (Claude Sonnet 4.5)
-- **Search subagent**: Uses Anthropic/Claude models (Claude Haiku 4.5)
-- **Review feature**: Uses Google/Gemini models (Gemini 2.5 Flash-Lite)
-
-For the most current information about which models Amp uses, see the **[Amp Models Documentation](https://ampcode.com/models)**.
-
-#### Fallback Behavior
-
-CLIProxyAPI uses a smart fallback system:
-
-1. **Provider authenticated locally** (`--login`, `--codex-login`, `--claude-login`):
-   - Requests use **your OAuth subscription** (ChatGPT Plus/Pro, Claude Pro/Max, Google account)
-   - You benefit from your subscription's included usage quotas
-   - No Amp credits consumed
-
-2. **Provider NOT authenticated locally**:
-   - Requests automatically forward to **ampcode.com**
-   - Uses Amp's backend provider connections
-   - **Requires Amp credits** if the provider is paid (OpenAI, Anthropic paid tiers)
-   - May result in errors if Amp credit balance is insufficient
-
-**Recommendation**: Authenticate all providers you have subscriptions for to maximize value and minimize Amp credit usage. If you don't have subscriptions to all providers Amp uses, ensure you have sufficient Amp credits available for fallback requests.
-
-## Architecture
-
-### Request Flow
-
-```
-Amp CLI/IDE
-  ↓
-  ├─ Provider API requests (/api/provider/{provider}/v1/...)
-  │   ↓
-  │   ├─ Model configured locally?
-  │   │   YES → Use local OAuth tokens (OpenAI/Claude/Gemini handlers)
-  │   │   NO  ↓
-  │   │       ├─ Model mapping configured?
-  │   │       │   YES → Rewrite model → Use local handler (free)
-  │   │       │   NO  → Forward to ampcode.com (uses Amp credits)
-  │   ↓
-  │   Response
-  │
-  └─ Management requests (/api/auth, /api/user, /api/threads, ...)
-      ↓
-      ├─ Localhost check (security)
-      ↓
-      └─ Reverse proxy to ampcode.com
-          ↓
-          Response (auto-decompressed if gzipped)
-```
-
-### Components
-
-The Amp integration is implemented as a modular routing module (`internal/api/modules/amp/`) with these components:
-
-1. **Route Aliases** (`routes.go`): Maps Amp-style paths to standard handlers
-2. **Reverse Proxy** (`proxy.go`): Forwards management requests to ampcode.com
-3. **Fallback Handler** (`fallback_handlers.go`): Routes unconfigured models to ampcode.com
-4. **Secret Management** (`secret.go`): Multi-source API key resolution with caching
-5. **Main Module** (`amp.go`): Orchestrates registration and configuration
-
-## Configuration
-
-### Basic Configuration
-
-Add these fields to your `config.yaml`:
-
-```yaml
-# Amp upstream control plane (required for management routes)
-amp-upstream-url: "https://ampcode.com"
-
-# Optional: Override API key (otherwise uses env or file)
-# amp-upstream-api-key: "your-amp-api-key"
-
-# Security: restrict management routes to localhost (recommended)
-amp-restrict-management-to-localhost: true
-```
-
-### Model Mapping Configuration
-
-When Amp CLI requests a model that you don't have access to, you can configure mappings to route those requests to alternative models that you DO have available. This avoids consuming Amp credits for models you could handle locally.
-
-```yaml
-# Route unavailable models to alternatives
-amp-model-mappings:
-  # Example: Route Claude Opus 4.5 requests to Claude Sonnet 4
-  - from: "claude-opus-4.5"
-    to: "claude-sonnet-4"
-  
-  # Example: Route GPT-5 requests to Gemini 2.5 Pro
-  - from: "gpt-5"
-    to: "gemini-2.5-pro"
-  
-  # Example: Map older model names to newer versions
-  - from: "claude-3-opus-20240229"
-    to: "claude-3-5-sonnet-20241022"
-```
-
-**How it works:**
-
-1. Amp CLI requests a model (e.g., `claude-opus-4.5`)
-2. CLIProxyAPI checks if a local provider is available for that model
-3. If not available, it checks the model mappings
-4. If a mapping exists, the request is rewritten to use the target model
-5. The request is then handled locally (free, using your OAuth subscription)
-
-**Benefits:**
-- **Save Amp credits**: Use your local subscriptions instead of forwarding to ampcode.com
-- **Hot-reload**: Mappings can be updated without restarting the proxy
-- **Structured logging**: Clear logs show when mappings are applied
-
-**Routing Decision Logs:**
-
-The proxy logs each routing decision with structured fields:
-
-```
-[AMP] Using local provider for model: gemini-2.5-pro          # Local provider (free)
-[AMP] Model mapped: claude-opus-4.5 -> claude-sonnet-4        # Mapping applied (free)
-[AMP] Forwarding to ampcode.com (uses Amp credits) - model_id: gpt-5  # Fallback (costs credits)
-```
-
-### Secret Resolution Precedence
-
-The Amp module resolves API keys using this precedence order:
-
-| Source | Key | Priority | Cache |
-|--------|-----|----------|-------|
-| Config file | `amp-upstream-api-key` | High | No |
-| Environment | `AMP_API_KEY` | Medium | No |
-| Amp secrets file | `~/.local/share/amp/secrets.json` | Low | 5 min |
-
-**Recommendation**: Use the Amp secrets file (lowest precedence) for normal usage. This file is automatically managed by `amp login`.
-
-### Security Settings
-
-**`amp-restrict-management-to-localhost`** (default: `true`)
-
-When enabled, management routes (`/api/auth`, `/api/user`, `/api/threads`, etc.) only accept connections from localhost (127.0.0.1, ::1). This prevents:
-- Drive-by browser attacks
-- Remote access to management endpoints
-- CORS-based attacks
-- Header spoofing attacks (e.g., `X-Forwarded-For: 127.0.0.1`)
-
-#### How It Works
-
-This restriction uses the **actual TCP connection address** (`RemoteAddr`), not HTTP headers like `X-Forwarded-For`. This prevents header spoofing attacks but has important implications:
-
-- ✅ **Works for direct connections**: Running CLIProxyAPI directly on your machine or server
-- ⚠️ **May not work behind reverse proxies**: If deploying behind nginx, Cloudflare, or other proxies, the connection will appear to come from the proxy's IP, not localhost
-
-#### Reverse Proxy Deployments
-
-If you need to run CLIProxyAPI behind a reverse proxy (nginx, Caddy, Cloudflare Tunnel, etc.):
-
-1. **Disable the localhost restriction**:
-   ```yaml
-   amp-restrict-management-to-localhost: false
-   ```
-
-2. **Use alternative security measures**:
-   - Firewall rules restricting access to management routes
-   - Proxy-level authentication (HTTP Basic Auth, OAuth)
-   - Network-level isolation (VPN, Tailscale, Cloudflare Access)
-   - Bind CLIProxyAPI to `127.0.0.1` only and access via SSH tunnel
-
-3. **Example nginx configuration** (blocks external access to management routes):
-   ```nginx
-   location /api/auth { deny all; }
-   location /api/user { deny all; }
-   location /api/threads { deny all; }
-   location /api/internal { deny all; }
-   ```
-
-**Important**: Only disable `amp-restrict-management-to-localhost` if you understand the security implications and have other protections in place.
-
-## Setup
-
-### 1. Configure CLIProxyAPI
-
-Create or edit `config.yaml`:
-
-```yaml
-port: 8317
-auth-dir: "~/.cli-proxy-api"
-
-# Amp integration
-amp-upstream-url: "https://ampcode.com"
-amp-restrict-management-to-localhost: true
-
-# Other standard settings...
-debug: false
-logging-to-file: true
-```
-
-### 2. Authenticate with Providers
-
-Run OAuth login for the providers you want to use:
-
-**Google Account (Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 3 Pro Preview):**
-```bash
-./cli-proxy-api --login
-```
-
-**ChatGPT Plus/Pro (GPT-5, GPT-5 Codex):**
-```bash
-./cli-proxy-api --codex-login
-```
-
-**Claude Pro/Max (Claude Sonnet 4.5, Opus 4.1):**
-```bash
-./cli-proxy-api --claude-login
-```
-
-Tokens are saved to:
-- Gemini: `~/.cli-proxy-api/gemini-<email>.json`
-- OpenAI Codex: `~/.cli-proxy-api/codex-<email>.json`
-- Claude: `~/.cli-proxy-api/claude-<email>.json`
-
-### 3. Start the Proxy
-
-```bash
-./cli-proxy-api --config config.yaml
-```
-
-Or run in background with tmux (recommended for remote servers):
-
-```bash
-tmux new-session -d -s proxy "./cli-proxy-api --config config.yaml"
-```
-
-### 4. Configure Amp CLI
-
-#### Option A: Settings File
-
-Edit `~/.config/amp/settings.json`:
-
-```json
-{
-  "amp.url": "http://localhost:8317"
-}
-```
-
-#### Option B: Environment Variable
-
-```bash
-export AMP_URL=http://localhost:8317
-```
-
-### 5. Login and Use Amp
-
-Login through the proxy (proxied to ampcode.com):
-
-```bash
-amp login
-```
-
-Use Amp as normal:
-
-```bash
-amp "Write a hello world program in Python"
-```
-
-### 6. (Optional) Configure Amp IDE Extension
-
-The proxy also works with Amp IDE extensions for VS Code, Cursor, Windsurf, etc.
-
-1. Open Amp extension settings in your IDE
-2. Set **Amp URL** to `http://localhost:8317`
-3. Login with your Amp account
-4. Start using Amp in your IDE
-
-Both CLI and IDE can use the proxy simultaneously.
-
-## Usage
-
-### Supported Routes
-
-#### Provider Aliases (Always Available)
-
-These routes work even without `amp-upstream-url` configured:
-
-- `/api/provider/openai/v1/chat/completions`
-- `/api/provider/openai/v1/responses`
-- `/api/provider/anthropic/v1/messages`
-- `/api/provider/google/v1beta/models/:action`
-
-Amp CLI calls these routes with your OAuth-authenticated models configured in CLIProxyAPI.
-
-#### Management Routes (Require `amp-upstream-url`)
-
-These routes are proxied to ampcode.com:
-
-- `/api/auth` - Authentication
-- `/api/user` - User profile
-- `/api/meta` - Metadata
-- `/api/threads` - Conversation threads
-- `/api/telemetry` - Usage telemetry
-- `/api/internal` - Internal APIs
-
-**Security**: Restricted to localhost by default.
-
-### Model Fallback Behavior
-
-When Amp requests a model:
-
-1. **Check local configuration**: Does CLIProxyAPI have OAuth tokens for this model's provider?
-2. **If YES**: Route to local handler (use your OAuth subscription)
-3. **If NO**: Check if a model mapping exists
-4. **If mapping exists**: Rewrite request to mapped model → Route to local handler (free)
-5. **If no mapping**: Forward to ampcode.com (uses Amp credits)
-
-This enables seamless mixed usage:
-- Models you've configured (Gemini, ChatGPT, Claude) → Your OAuth subscriptions
-- Models with mappings configured → Routed to alternative local models (free)
-- Models you haven't configured and have no mapping → Amp's default providers (uses credits)
-
-### Example API Calls
-
-**Chat completion with local OAuth:**
-```bash
-curl http://localhost:8317/api/provider/openai/v1/chat/completions \
-  -H "Content-Type: application/json" \
-  -d '{
-    "model": "gpt-5",
-    "messages": [{"role": "user", "content": "Hello"}]
-  }'
-```
-
-**Management endpoint (localhost only):**
-```bash
-curl http://localhost:8317/api/user
-```
-
-## Troubleshooting
-
-### Common Issues
-
-| Symptom | Likely Cause | Fix |
-|---------|--------------|-----|
-| 404 on `/api/provider/...` | Incorrect route path | Ensure exact path: `/api/provider/{provider}/v1...` |
-| 403 on `/api/user` | Non-localhost request | Run from same machine or disable `amp-restrict-management-to-localhost` (not recommended) |
-| 401/403 from provider | Missing/expired OAuth | Re-run `--codex-login` or `--claude-login` |
-| Amp gzip errors | Response decompression issue | Update to latest build; auto-decompression should handle this |
-| Models not using proxy | Wrong Amp URL | Verify `amp.url` setting or `AMP_URL` environment variable |
-| CORS errors | Protected management endpoint | Use CLI/terminal, not browser |
-
-### Diagnostics
-
-**Check proxy logs:**
-```bash
-# If logging-to-file: true
-tail -f logs/requests.log
-
-# If running in tmux
-tmux attach-session -t proxy
-```
-
-**Enable debug mode** (temporarily):
-```yaml
-debug: true
-```
-
-**Test basic connectivity:**
-```bash
-# Check if proxy is running
-curl http://localhost:8317/v1/models
-
-# Check Amp-specific route
-curl http://localhost:8317/api/provider/openai/v1/models
-```
-
-**Verify Amp configuration:**
-```bash
-# Check if Amp is using proxy
-amp config get amp.url
-
-# Or check environment
-echo $AMP_URL
-```
-
-### Security Checklist
-
-- ✅ Keep `amp-restrict-management-to-localhost: true` (default)
-- ✅ Don't expose proxy publicly (bind to localhost or use firewall/VPN)
-- ✅ Use the Amp secrets file (`~/.local/share/amp/secrets.json`) managed by `amp login`
-- ✅ Rotate OAuth tokens periodically by re-running login commands
-- ✅ Store config and auth-dir on encrypted disk if handling sensitive data
-- ✅ Keep proxy binary up to date for security fixes
-
-## Additional Resources
-
-- [CLIProxyAPI Main Documentation](https://help.router-for.me/)
-- [Amp CLI Official Manual](https://ampcode.com/manual)
-- [Management API Reference](https://help.router-for.me/management/api)
-- [SDK Documentation](sdk-usage.md)
-
-## Disclaimer
-
-This integration is for personal/educational use. Using reverse proxies or alternate API bases may violate provider Terms of Service. You are solely responsible for how you use this software. Accounts may be rate-limited, locked, or banned. No warranties. Use at your own risk.

docs/amp-cli-integration_CN.md

@@ -1,392 +0,0 @@
-# Amp CLI 集成指南
-
-本指南说明如何在 Amp CLI 和 Amp IDE 扩展中使用 CLIProxyAPI，通过 OAuth 让你能够把已有的 Google/ChatGPT/Claude 订阅与 Amp 的 CLI 一起使用。
-
-## 目录
-
-- [概述](#概述)
-  - [应该认证哪些服务提供商？](#应该认证哪些服务提供商)
-- [架构](#架构)
-- [配置](#配置)
-- [设置](#设置)
-- [用法](#用法)
-- [故障排查](#故障排查)
-
-## 概述
-
-Amp CLI 集成为 Amp 的 API 模式添加了专用路由，同时保持与现有 CLIProxyAPI 功能的完全兼容。这样你可以在同一个代理服务器上同时使用传统 CLIProxyAPI 功能和 Amp CLI。
-
-### 主要特性
-
-- **提供者路由别名**：将 Amp 的 `/api/provider/{provider}/v1...` 路径映射到 CLIProxyAPI 处理器
-- **管理代理**：将 OAuth 和账号管理请求转发到 Amp 控制平面
-- **智能回退**：自动将未配置的模型路由到 ampcode.com
-- **密钥管理**：可配置优先级（配置 > 环境变量 > 文件），缓存 5 分钟
-- **安全优先**：管理路由默认限制为 localhost
-- **自动 gzip 处理**：自动解压来自 Amp 上游的响应
-
-### 你可以做什么
-
-- 使用 Amp CLI 搭配你的 Google 账号（Gemini 3 Pro Preview、Gemini 2.5 Pro、Gemini 2.5 Flash）
-- 使用 Amp CLI 搭配你的 ChatGPT Plus/Pro 订阅（GPT-5、GPT-5 Codex 模型）
-- 使用 Amp CLI 搭配你的 Claude Pro/Max 订阅（Claude Sonnet 4.5、Opus 4.1）
-- 将 Amp IDE 扩展（VS Code、Cursor、Windsurf 等）与同一个代理一起使用
-- 通过一个代理同时运行多个 CLI 工具（Factory + Amp）
-- 将未配置的模型自动路由到 ampcode.com
-
-### 应该认证哪些服务提供商？
-
-**重要**：需要认证的提供商取决于你安装的 Amp 版本当前使用的模型和功能。Amp 的不同智能模式和子代理会使用不同的提供商：
-
-- **Smart 模式**：使用 Google/Gemini 模型（Gemini 3 Pro）
-- **Rush 模式**：使用 Anthropic/Claude 模型（Claude Haiku 4.5）
-- **Oracle 子代理**：使用 OpenAI/GPT 模型（GPT-5 medium reasoning）
-- **Librarian 子代理**：使用 Anthropic/Claude 模型（Claude Sonnet 4.5）
-- **Search 子代理**：使用 Anthropic/Claude 模型（Claude Haiku 4.5）
-- **Review 功能**：使用 Google/Gemini 模型（Gemini 2.5 Flash-Lite）
-
-有关 Amp 当前使用哪些模型的最新信息，请参阅 **[Amp 模型文档](https://ampcode.com/models)**。
-
-#### 回退行为
-
-CLIProxyAPI 采用智能回退机制：
-
-1. **本地已认证提供商**（`--login`、`--codex-login`、`--claude-login`）：
-   - 请求使用**你的 OAuth 订阅**（ChatGPT Plus/Pro、Claude Pro/Max、Google 账号）
-   - 享受订阅自带的额度
-   - 不消耗 Amp 额度
-
-2. **本地未认证提供商**：
-   - 请求自动转发到 **ampcode.com**
-   - 使用 Amp 的后端提供商连接
-   - 如果提供商是付费的（OpenAI、Anthropic 付费档），**需要消耗 Amp 额度**
-   - 若 Amp 额度不足，可能产生错误
-
-**建议**：对你有订阅的所有提供商都进行认证，以最大化价值并尽量减少 Amp 额度消耗。如果没有覆盖 Amp 使用的全部提供商，请确保为回退请求准备足够的 Amp 额度。
-
-## 架构
-
-### 请求流
-
-```
-Amp CLI/IDE
-  ↓
-  ├─ Provider API requests (/api/provider/{provider}/v1/...)
-  │   ↓
-  │   ├─ Model configured locally?
-  │   │   YES → Use local OAuth tokens (OpenAI/Claude/Gemini handlers)
-  │   │   NO  → Forward to ampcode.com (reverse proxy)
-  │   ↓
-  │   Response
-  │
-  └─ Management requests (/api/auth, /api/user, /api/threads, ...)
-      ↓
-      ├─ Localhost check (security)
-      ↓
-      └─ Reverse proxy to ampcode.com
-          ↓
-          Response (auto-decompressed if gzipped)
-```
-
-### 组件
-
-Amp 集成以模块化路由模块（`internal/api/modules/amp/`）实现，包含以下组件：
-
-1. **路由别名**（`routes.go`）：将 Amp 风格的路径映射到标准处理器
-2. **反向代理**（`proxy.go`）：将管理请求转发到 ampcode.com
-3. **回退处理器**（`fallback_handlers.go`）：将未配置的模型路由到 ampcode.com
-4. **密钥管理**（`secret.go`）：多来源 API 密钥解析并带缓存
-5. **主模块**（`amp.go`）：负责注册和配置
-
-## 配置
-
-### 基础配置
-
-在 `config.yaml` 中新增以下字段：
-
-```yaml
-# Amp 上游控制平面（管理路由必需）
-amp-upstream-url: "https://ampcode.com"
-
-# 可选：覆盖 API key（否则使用环境变量或文件）
-# amp-upstream-api-key: "your-amp-api-key"
-
-# 安全性：将管理路由限制为 localhost（推荐）
-amp-restrict-management-to-localhost: true
-```
-
-### 密钥解析优先级
-
-Amp 模块以如下优先级解析 API key：
-
-| 来源 | 键名 | 优先级 | 缓存 |
-|------|------|--------|------|
-| 配置文件 | `amp-upstream-api-key` | 高 | 无 |
-| 环境变量 | `AMP_API_KEY` | 中 | 无 |
-| Amp 密钥文件 | `~/.local/share/amp/secrets.json` | 低 | 5 分钟 |
-
-**建议**：日常使用时采用 Amp 密钥文件（最低优先级）。该文件由 `amp login` 自动管理。
-
-### 安全设置
-
-**`amp-restrict-management-to-localhost`**（默认：`true`）
-
-启用后，管理路由（`/api/auth`、`/api/user`、`/api/threads` 等）只接受来自 localhost（127.0.0.1、::1）的连接，可防止：
-- 浏览器探测式攻击
-- 对管理端点的远程访问
-- 基于 CORS 的攻击
-- 伪造头攻击（例如 `X-Forwarded-For: 127.0.0.1`）
-
-#### 工作原理
-
-此限制使用**实际的 TCP 连接地址**（`RemoteAddr`），而非 `X-Forwarded-For` 等 HTTP 头，能防止头部伪造，但有重要影响：
-
-- ✅ **直接连接可用**：在本机或服务器直接运行 CLIProxyAPI 时适用
-- ⚠️ **可能不适用于反向代理场景**：部署在 nginx、Cloudflare 等代理后，请求源会显示为代理 IP 而非 localhost
-
-#### 反向代理部署
-
-若需要在反向代理（nginx、Caddy、Cloudflare Tunnel 等）后运行 CLIProxyAPI：
-
-1. **关闭 localhost 限制**：
-   ```yaml
-   amp-restrict-management-to-localhost: false
-   ```
-
-2. **使用替代安全措施**：
-   - 防火墙规则限制管理路由访问
-   - 代理层认证（HTTP Basic Auth、OAuth）
-   - 网络隔离（VPN、Tailscale、Cloudflare Access）
-   - 将 CLIProxyAPI 仅绑定 `127.0.0.1`，并通过 SSH 隧道访问
-
-3. **nginx 示例配置**（阻止外部访问管理路由）：
-   ```nginx
-   location /api/auth { deny all; }
-   location /api/user { deny all; }
-   location /api/threads { deny all; }
-   location /api/internal { deny all; }
-   ```
-
-**重要**：只有在理解安全影响并已采取其他防护措施时，才关闭 `amp-restrict-management-to-localhost`。
-
-## 设置
-
-### 1. 配置 CLIProxyAPI
-
-创建或编辑 `config.yaml`：
-
-```yaml
-port: 8317
-auth-dir: "~/.cli-proxy-api"
-
-# Amp 集成
-amp-upstream-url: "https://ampcode.com"
-amp-restrict-management-to-localhost: true
-
-# 其他常规设置...
-debug: false
-logging-to-file: true
-```
-
-### 2. 认证提供商
-
-为要使用的提供商执行 OAuth 登录：
-
-**Google 账号（Gemini 2.5 Pro、Gemini 2.5 Flash、Gemini 3 Pro Preview）：**
-```bash
-./cli-proxy-api --login
-```
-
-**ChatGPT Plus/Pro（GPT-5、GPT-5 Codex）：**
-```bash
-./cli-proxy-api --codex-login
-```
-
-**Claude Pro/Max（Claude Sonnet 4.5、Opus 4.1）：**
-```bash
-./cli-proxy-api --claude-login
-```
-
-令牌会保存到：
-- Gemini: `~/.cli-proxy-api/gemini-<email>.json`
-- OpenAI Codex: `~/.cli-proxy-api/codex-<email>.json`
-- Claude: `~/.cli-proxy-api/claude-<email>.json`
-
-### 3. 启动代理
-
-```bash
-./cli-proxy-api --config config.yaml
-```
-
-或使用 tmux 在后台运行（推荐用于远程服务器）：
-
-```bash
-tmux new-session -d -s proxy "./cli-proxy-api --config config.yaml"
-```
-
-### 4. 配置 Amp CLI
-
-#### 方案 A：配置文件
-
-编辑 `~/.config/amp/settings.json`：
-
-```json
-{
-  "amp.url": "http://localhost:8317"
-}
-```
-
-#### 方案 B：环境变量
-
-```bash
-export AMP_URL=http://localhost:8317
-```
-
-### 5. 登录并使用 Amp
-
-通过代理登录（请求会被代理到 ampcode.com）：
-
-```bash
-amp login
-```
-
-像平常一样使用 Amp：
-
-```bash
-amp "Write a hello world program in Python"
-```
-
-### 6. （可选）配置 Amp IDE 扩展
-
-该代理同样适用于 VS Code、Cursor、Windsurf 等 Amp IDE 扩展。
-
-1. 在 IDE 中打开 Amp 扩展设置
-2. 将 **Amp URL** 设置为 `http://localhost:8317`
-3. 用你的 Amp 账号登录
-4. 在 IDE 中开始使用 Amp
-
-CLI 和 IDE 可同时使用该代理。
-
-## 用法
-
-### 支持的路由
-
-#### 提供商别名（始终可用）
-
-这些路由即使未配置 `amp-upstream-url` 也可使用：
-
-- `/api/provider/openai/v1/chat/completions`
-- `/api/provider/openai/v1/responses`
-- `/api/provider/anthropic/v1/messages`
-- `/api/provider/google/v1beta/models/:action`
-
-Amp CLI 会使用你在 CLIProxyAPI 中通过 OAuth 认证的模型来调用这些路由。
-
-#### 管理路由（需要 `amp-upstream-url`）
-
-这些路由会被代理到 ampcode.com：
-
-- `/api/auth` - 认证
-- `/api/user` - 用户资料
-- `/api/meta` - 元数据
-- `/api/threads` - 会话线程
-- `/api/telemetry` - 使用遥测
-- `/api/internal` - 内部 API
-
-**安全性**：默认限制为 localhost。
-
-### 模型回退行为
-
-当 Amp 请求模型时：
-
-1. **检查本地配置**：CLIProxyAPI 是否有该模型提供商的 OAuth 令牌？
-2. **如果有**：路由到本地处理器（使用你的 OAuth 订阅）
-3. **如果没有**：转发到 ampcode.com（使用 Amp 的默认路由）
-
-这实现了无缝混用：
-- 你已配置的模型（Gemini、ChatGPT、Claude）→ 你的 OAuth 订阅
-- 未配置的模型 → Amp 的默认提供商
-
-### 示例 API 调用
-
-**使用本地 OAuth 的聊天补全：**
-```bash
-curl http://localhost:8317/api/provider/openai/v1/chat/completions \
-  -H "Content-Type: application/json" \
-  -d '{
-    "model": "gpt-5",
-    "messages": [{"role": "user", "content": "Hello"}]
-  }'
-```
-
-**管理端点（仅限 localhost）：**
-```bash
-curl http://localhost:8317/api/user
-```
-
-## 故障排查
-
-### 常见问题
-
-| 症状 | 可能原因 | 解决方案 |
-|------|----------|----------|
-| `/api/provider/...` 返回 404 | 路径错误 | 确保路径准确：`/api/provider/{provider}/v1...` |
-| `/api/user` 返回 403 | 非 localhost 请求 | 在同一机器上访问，或关闭 `amp-restrict-management-to-localhost`（不推荐） |
-| 提供商返回 401/403 | OAuth 缺失或过期 | 重新运行 `--codex-login` 或 `--claude-login` |
-| Amp gzip 错误 | 响应解压问题 | 更新到最新构建；自动解压应能处理 |
-| 模型未走代理 | Amp URL 设置错误 | 检查 `amp.url` 设置或 `AMP_URL` 环境变量 |
-| CORS 错误 | 受保护的管理端点 | 使用 CLI/终端而非浏览器 |
-
-### 诊断
-
-**查看代理日志：**
-```bash
-# 若 logging-to-file: true
-tail -f logs/requests.log
-
-# 若运行在 tmux 中
-tmux attach-session -t proxy
-```
-
-**临时开启调试模式：**
-```yaml
-debug: true
-```
-
-**测试基础连通性：**
-```bash
-# 检查代理是否运行
-curl http://localhost:8317/v1/models
-
-# 检查 Amp 特定路由
-curl http://localhost:8317/api/provider/openai/v1/models
-```
-
-**验证 Amp 配置：**
-```bash
-# 检查 Amp 是否使用代理
-amp config get amp.url
-
-# 或检查环境变量
-echo $AMP_URL
-```
-
-### 安全清单
-
-- ✅ 保持 `amp-restrict-management-to-localhost: true`（默认）
-- ✅ 不要将代理暴露到公共网络（绑定到 localhost 或使用防火墙/VPN）
-- ✅ 使用 `amp login` 管理的 Amp 密钥文件（`~/.local/share/amp/secrets.json`）
-- ✅ 定期重新登录轮换 OAuth 令牌
-- ✅ 若处理敏感数据，使用加密磁盘存储配置和 auth-dir
-- ✅ 保持代理二进制为最新版本以获取安全修复
-
-## 其他资源
-
-- [CLIProxyAPI 主文档](https://help.router-for.me/)
-- [Amp CLI 官方手册](https://ampcode.com/manual)
-- [管理 API 参考](https://help.router-for.me/management/api)
-- [SDK 文档](sdk-usage.md)
-
-## 免责声明
-
-此集成仅用于个人或教育用途。使用反向代理或替代 API 基址可能违反提供商的服务条款。你需要对自己的使用方式负责。账号可能会被限速、锁定或封禁。软件不附带任何保证，使用风险自负。

go.mod

@@ -18,8 +18,8 @@ require (
 	github.com/tidwall/gjson v1.18.0
 	github.com/tidwall/sjson v1.2.5
 	github.com/tiktoken-go/tokenizer v0.7.0
-	golang.org/x/crypto v0.43.0
-	golang.org/x/net v0.46.0
+	golang.org/x/crypto v0.45.0
+	golang.org/x/net v0.47.0
 	golang.org/x/oauth2 v0.30.0
 	gopkg.in/natefinch/lumberjack.v2 v2.2.1
 	gopkg.in/yaml.v3 v3.0.1
@@ -68,9 +68,9 @@ require (
 	github.com/twitchyliquid64/golang-asm v0.15.1 // indirect
 	github.com/ugorji/go/codec v1.2.12 // indirect
 	golang.org/x/arch v0.8.0 // indirect
-	golang.org/x/sync v0.17.0 // indirect
-	golang.org/x/sys v0.37.0 // indirect
-	golang.org/x/text v0.30.0 // indirect
+	golang.org/x/sync v0.18.0 // indirect
+	golang.org/x/sys v0.38.0 // indirect
+	golang.org/x/text v0.31.0 // indirect
 	google.golang.org/protobuf v1.34.1 // indirect
 	gopkg.in/ini.v1 v1.67.0 // indirect
 )

go.sum

@@ -160,22 +160,22 @@ github.com/ugorji/go/codec v1.2.12/go.mod h1:UNopzCgEMSXjBc6AOMqYvWC1ktqTAfzJZUZ
 golang.org/x/arch v0.0.0-20210923205945-b76863e36670/go.mod h1:5om86z9Hs0C8fWVUuoMHwpExlXzs5Tkyp9hOrfG7pp8=
 golang.org/x/arch v0.8.0 h1:3wRIsP3pM4yUptoR96otTUOXI367OS0+c9eeRi9doIc=
 golang.org/x/arch v0.8.0/go.mod h1:FEVrYAQjsQXMVJ1nsMoVVXPZg6p2JE2mx8psSWTDQys=
-golang.org/x/crypto v0.43.0 h1:dduJYIi3A3KOfdGOHX8AVZ/jGiyPa3IbBozJ5kNuE04=
-golang.org/x/crypto v0.43.0/go.mod h1:BFbav4mRNlXJL4wNeejLpWxB7wMbc79PdRGhWKncxR0=
-golang.org/x/net v0.46.0 h1:giFlY12I07fugqwPuWJi68oOnpfqFnJIJzaIIm2JVV4=
-golang.org/x/net v0.46.0/go.mod h1:Q9BGdFy1y4nkUwiLvT5qtyhAnEHgnQ/zd8PfU6nc210=
+golang.org/x/crypto v0.45.0 h1:jMBrvKuj23MTlT0bQEOBcAE0mjg8mK9RXFhRH6nyF3Q=
+golang.org/x/crypto v0.45.0/go.mod h1:XTGrrkGJve7CYK7J8PEww4aY7gM3qMCElcJQ8n8JdX4=
+golang.org/x/net v0.47.0 h1:Mx+4dIFzqraBXUugkia1OOvlD6LemFo1ALMHjrXDOhY=
+golang.org/x/net v0.47.0/go.mod h1:/jNxtkgq5yWUGYkaZGqo27cfGZ1c5Nen03aYrrKpVRU=
 golang.org/x/oauth2 v0.30.0 h1:dnDm7JmhM45NNpd8FDDeLhK6FwqbOf4MLCM9zb1BOHI=
 golang.org/x/oauth2 v0.30.0/go.mod h1:B++QgG3ZKulg6sRPGD/mqlHQs5rB3Ml9erfeDY7xKlU=
-golang.org/x/sync v0.17.0 h1:l60nONMj9l5drqw6jlhIELNv9I0A4OFgRsG9k2oT9Ug=
-golang.org/x/sync v0.17.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
+golang.org/x/sync v0.18.0 h1:kr88TuHDroi+UVf+0hZnirlk8o8T+4MrK6mr60WkH/I=
+golang.org/x/sync v0.18.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
 golang.org/x/sys v0.0.0-20220715151400-c0bba94af5f8/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.37.0 h1:fdNQudmxPjkdUTPnLn5mdQv7Zwvbvpaxqs831goi9kQ=
-golang.org/x/sys v0.37.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
-golang.org/x/term v0.36.0 h1:zMPR+aF8gfksFprF/Nc/rd1wRS1EI6nDBGyWAvDzx2Q=
-golang.org/x/term v0.36.0/go.mod h1:Qu394IJq6V6dCBRgwqshf3mPF85AqzYEzofzRdZkWss=
-golang.org/x/text v0.30.0 h1:yznKA/E9zq54KzlzBEAWn1NXSQ8DIp/NYMy88xJjl4k=
-golang.org/x/text v0.30.0/go.mod h1:yDdHFIX9t+tORqspjENWgzaCVXgk0yYnYuSZ8UzzBVM=
+golang.org/x/sys v0.38.0 h1:3yZWxaJjBmCWXqhN1qh02AkOnCQ1poK6oF+a7xWL6Gc=
+golang.org/x/sys v0.38.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
+golang.org/x/term v0.37.0 h1:8EGAD0qCmHYZg6J17DvsMy9/wJ7/D/4pV/wfnld5lTU=
+golang.org/x/term v0.37.0/go.mod h1:5pB4lxRNYYVZuTLmy8oR2BH8dflOR+IbTYFD8fi3254=
+golang.org/x/text v0.31.0 h1:aC8ghyu4JhP8VojJ2lEHBnochRno1sgL6nEi9WGFGMM=
+golang.org/x/text v0.31.0/go.mod h1:tKRAlv61yKIjGGHX/4tP1LTbc13YSec1pxVEWXzfoeM=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543 h1:E7g+9GITq07hpfrRu66IVDexMakfv52eLZ2CXBWiKr4=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 google.golang.org/protobuf v1.34.1 h1:9ddQBjfCyZPOHPUiPxpYESBLc+T8P3E+Vo4IbKZgFWg=

internal/api/middleware/response_writer.go

@@ -71,22 +71,64 @@ func (w *ResponseWriterWrapper) Write(data []byte) (int, error) {
 	n, err := w.ResponseWriter.Write(data)
 
 	// THEN: Handle logging based on response type
-	if w.isStreaming {
+	if w.isStreaming && w.chunkChannel != nil {
 		// For streaming responses: Send to async logging channel (non-blocking)
-		if w.chunkChannel != nil {
-			select {
-			case w.chunkChannel <- append([]byte(nil), data...): // Non-blocking send with copy
-			default: // Channel full, skip logging to avoid blocking
-			}
+		select {
+		case w.chunkChannel <- append([]byte(nil), data...): // Non-blocking send with copy
+		default: // Channel full, skip logging to avoid blocking
 		}
-	} else {
-		// For non-streaming responses: Buffer complete response
+		return n, err
+	}
+
+	if w.shouldBufferResponseBody() {
 		w.body.Write(data)
 	}
 
 	return n, err
 }
 
+func (w *ResponseWriterWrapper) shouldBufferResponseBody() bool {
+	if w.logger != nil && w.logger.IsEnabled() {
+		return true
+	}
+	if !w.logOnErrorOnly {
+		return false
+	}
+	status := w.statusCode
+	if status == 0 {
+		if statusWriter, ok := w.ResponseWriter.(interface{ Status() int }); ok && statusWriter != nil {
+			status = statusWriter.Status()
+		} else {
+			status = http.StatusOK
+		}
+	}
+	return status >= http.StatusBadRequest
+}
+
+// WriteString wraps the underlying ResponseWriter's WriteString method to capture response data.
+// Some handlers (and fmt/io helpers) write via io.StringWriter; without this override, those writes
+// bypass Write() and would be missing from request logs.
+func (w *ResponseWriterWrapper) WriteString(data string) (int, error) {
+	w.ensureHeadersCaptured()
+
+	// CRITICAL: Write to client first (zero latency)
+	n, err := w.ResponseWriter.WriteString(data)
+
+	// THEN: Capture for logging
+	if w.isStreaming && w.chunkChannel != nil {
+		select {
+		case w.chunkChannel <- []byte(data):
+		default:
+		}
+		return n, err
+	}
+
+	if w.shouldBufferResponseBody() {
+		w.body.WriteString(data)
+	}
+	return n, err
+}
+
 // WriteHeader wraps the underlying ResponseWriter's WriteHeader method.
 // It captures the status code, detects if the response is streaming based on the Content-Type header,
 // and initializes the appropriate logging mechanism (standard or streaming).
@@ -160,12 +202,16 @@ func (w *ResponseWriterWrapper) detectStreaming(contentType string) bool {
 		return true
 	}
 
-	// Check request body for streaming indicators
-	if w.requestInfo.Body != nil {
+	// If a concrete Content-Type is already set (e.g., application/json for error responses),
+	// treat it as non-streaming instead of inferring from the request payload.
+	if strings.TrimSpace(contentType) != "" {
+		return false
+	}
+
+	// Only fall back to request payload hints when Content-Type is not set yet.
+	if w.requestInfo != nil && len(w.requestInfo.Body) > 0 {
 		bodyStr := string(w.requestInfo.Body)
-		if strings.Contains(bodyStr, `"stream": true`) || strings.Contains(bodyStr, `"stream":true`) {
-			return true
-		}
+		return strings.Contains(bodyStr, `"stream": true`) || strings.Contains(bodyStr, `"stream":true`)
 	}
 
 	return false
@@ -221,7 +267,7 @@ func (w *ResponseWriterWrapper) Finalize(c *gin.Context) error {
 		return nil
 	}
 
-	if w.isStreaming {
+	if w.isStreaming && w.streamWriter != nil {
 		if w.chunkChannel != nil {
 			close(w.chunkChannel)
 			w.chunkChannel = nil
@@ -233,24 +279,19 @@ func (w *ResponseWriterWrapper) Finalize(c *gin.Context) error {
 		}
 
 		// Write API Request and Response to the streaming log before closing
-		if w.streamWriter != nil {
-			apiRequest := w.extractAPIRequest(c)
-			if len(apiRequest) > 0 {
-				_ = w.streamWriter.WriteAPIRequest(apiRequest)
-			}
-			apiResponse := w.extractAPIResponse(c)
-			if len(apiResponse) > 0 {
-				_ = w.streamWriter.WriteAPIResponse(apiResponse)
-			}
-			if err := w.streamWriter.Close(); err != nil {
-				w.streamWriter = nil
-				return err
-			}
+		apiRequest := w.extractAPIRequest(c)
+		if len(apiRequest) > 0 {
+			_ = w.streamWriter.WriteAPIRequest(apiRequest)
+		}
+		apiResponse := w.extractAPIResponse(c)
+		if len(apiResponse) > 0 {
+			_ = w.streamWriter.WriteAPIResponse(apiResponse)
+		}
+		if err := w.streamWriter.Close(); err != nil {
 			w.streamWriter = nil
+			return err
 		}
-		if forceLog {
-			return w.logRequest(finalStatusCode, w.cloneHeaders(), w.body.Bytes(), w.extractAPIRequest(c), w.extractAPIResponse(c), slicesAPIResponseError, forceLog)
-		}
+		w.streamWriter = nil
 		return nil
 	}
 
@@ -335,26 +376,3 @@ func (w *ResponseWriterWrapper) logRequest(statusCode int, headers map[string][]
 		apiResponseErrors,
 	)
 }
-
-// Status returns the HTTP response status code captured by the wrapper.
-// It defaults to 200 if WriteHeader has not been called.
-func (w *ResponseWriterWrapper) Status() int {
-	if w.statusCode == 0 {
-		return 200 // Default status code
-	}
-	return w.statusCode
-}
-
-// Size returns the size of the response body in bytes for non-streaming responses.
-// For streaming responses, it returns -1, as the total size is unknown.
-func (w *ResponseWriterWrapper) Size() int {
-	if w.isStreaming {
-		return -1 // Unknown size for streaming responses
-	}
-	return w.body.Len()
-}
-
-// Written returns true if the response header has been written (i.e., a status code has been set).
-func (w *ResponseWriterWrapper) Written() bool {
-	return w.statusCode != 0
-}

internal/api/modules/amp/amp_test.go

@@ -146,6 +146,9 @@ func TestAmpModule_OnConfigUpdated_CacheInvalidation(t *testing.T) {
 	m := &AmpModule{enabled: true}
 	ms := NewMultiSourceSecretWithPath("", p, time.Minute)
 	m.secretSource = ms
+	m.lastConfig = &config.AmpCode{
+		UpstreamAPIKey: "old-key",
+	}
 
 	// Warm the cache
 	if _, err := ms.Get(context.Background()); err != nil {
@@ -157,7 +160,7 @@ func TestAmpModule_OnConfigUpdated_CacheInvalidation(t *testing.T) {
 	}
 
 	// Update config - should invalidate cache
-	if err := m.OnConfigUpdated(&config.Config{AmpCode: config.AmpCode{UpstreamURL: "http://x"}}); err != nil {
+	if err := m.OnConfigUpdated(&config.Config{AmpCode: config.AmpCode{UpstreamURL: "http://x", UpstreamAPIKey: "new-key"}}); err != nil {
 		t.Fatal(err)
 	}
 

internal/api/modules/amp/routes.go

@@ -267,7 +267,7 @@ func (m *AmpModule) registerProviderAliases(engine *gin.Engine, baseHandler *han
 	v1betaAmp := provider.Group("/v1beta")
 	{
 		v1betaAmp.GET("/models", geminiHandlers.GeminiModels)
-		v1betaAmp.POST("/models/:action", fallbackHandler.WrapHandler(geminiHandlers.GeminiHandler))
-		v1betaAmp.GET("/models/:action", geminiHandlers.GeminiGetHandler)
+		v1betaAmp.POST("/models/*action", fallbackHandler.WrapHandler(geminiHandlers.GeminiHandler))
+		v1betaAmp.GET("/models/*action", geminiHandlers.GeminiGetHandler)
 	}
 }

internal/api/modules/amp/routes_test.go

@@ -32,7 +32,9 @@ func TestRegisterManagementRoutes(t *testing.T) {
 	m.setProxy(proxy)
 
 	base := &handlers.BaseAPIHandler{}
-	m.registerManagementRoutes(r, base)
+	m.registerManagementRoutes(r, base, nil)
+	srv := httptest.NewServer(r)
+	defer srv.Close()
 
 	managementPaths := []struct {
 		path   string
@@ -63,11 +65,17 @@ func TestRegisterManagementRoutes(t *testing.T) {
 	for _, path := range managementPaths {
 		t.Run(path.path, func(t *testing.T) {
 			proxyCalled = false
-			req := httptest.NewRequest(path.method, path.path, nil)
-			w := httptest.NewRecorder()
-			r.ServeHTTP(w, req)
+			req, err := http.NewRequest(path.method, srv.URL+path.path, nil)
+			if err != nil {
+				t.Fatalf("failed to build request: %v", err)
+			}
+			resp, err := http.DefaultClient.Do(req)
+			if err != nil {
+				t.Fatalf("request failed: %v", err)
+			}
+			defer resp.Body.Close()
 
-			if w.Code == http.StatusNotFound {
+			if resp.StatusCode == http.StatusNotFound {
 				t.Fatalf("route %s not registered", path.path)
 			}
 			if !proxyCalled {

internal/api/server.go

@@ -230,13 +230,9 @@ func NewServer(cfg *config.Config, authManager *auth.Manager, accessManager *sdk
 	envManagementSecret := envAdminPasswordSet && envAdminPassword != ""
 
 	// Create server instance
-	providerNames := make([]string, 0, len(cfg.OpenAICompatibility))
-	for _, p := range cfg.OpenAICompatibility {
-		providerNames = append(providerNames, p.Name)
-	}
 	s := &Server{
 		engine:              engine,
-		handlers:            handlers.NewBaseAPIHandlers(&cfg.SDKConfig, authManager, providerNames),
+		handlers:            handlers.NewBaseAPIHandlers(&cfg.SDKConfig, authManager),
 		cfg:                 cfg,
 		accessManager:       accessManager,
 		requestLogger:       requestLogger,
@@ -334,8 +330,8 @@ func (s *Server) setupRoutes() {
 	v1beta.Use(AuthMiddleware(s.accessManager))
 	{
 		v1beta.GET("/models", geminiHandlers.GeminiModels)
-		v1beta.POST("/models/:action", geminiHandlers.GeminiHandler)
-		v1beta.GET("/models/:action", geminiHandlers.GeminiGetHandler)
+		v1beta.POST("/models/*action", geminiHandlers.GeminiHandler)
+		v1beta.GET("/models/*action", geminiHandlers.GeminiGetHandler)
 	}
 
 	// Root endpoint
@@ -609,7 +605,7 @@ func (s *Server) serveManagementControlPanel(c *gin.Context) {
 
 	if _, err := os.Stat(filePath); err != nil {
 		if os.IsNotExist(err) {
-			go managementasset.EnsureLatestManagementHTML(context.Background(), managementasset.StaticDir(s.configFilePath), cfg.ProxyURL)
+			go managementasset.EnsureLatestManagementHTML(context.Background(), managementasset.StaticDir(s.configFilePath), cfg.ProxyURL, cfg.RemoteManagement.PanelGitHubRepository)
 			c.AbortWithStatus(http.StatusNotFound)
 			return
 		}
@@ -919,17 +915,11 @@ func (s *Server) UpdateClients(cfg *config.Config) {
 	// Save YAML snapshot for next comparison
 	s.oldConfigYaml, _ = yaml.Marshal(cfg)
 
-	providerNames := make([]string, 0, len(cfg.OpenAICompatibility))
-	for _, p := range cfg.OpenAICompatibility {
-		providerNames = append(providerNames, p.Name)
-	}
-	s.handlers.OpenAICompatProviders = providerNames
-
 	s.handlers.UpdateClients(&cfg.SDKConfig)
 
 	if !cfg.RemoteManagement.DisableControlPanel {
 		staticDir := managementasset.StaticDir(s.configFilePath)
-		go managementasset.EnsureLatestManagementHTML(context.Background(), staticDir, cfg.ProxyURL)
+		go managementasset.EnsureLatestManagementHTML(context.Background(), staticDir, cfg.ProxyURL, cfg.RemoteManagement.PanelGitHubRepository)
 	}
 	if s.mgmt != nil {
 		s.mgmt.SetConfig(cfg)

internal/config/config.go

@@ -17,6 +17,8 @@ import (
 	"gopkg.in/yaml.v3"
 )
 
+const DefaultPanelGitHubRepository = "https://github.com/router-for-me/Cli-Proxy-API-Management-Center"
+
 // Config represents the application's configuration, loaded from a YAML file.
 type Config struct {
 	config.SDKConfig `yaml:",inline"`
@@ -104,6 +106,9 @@ type RemoteManagement struct {
 	SecretKey string `yaml:"secret-key"`
 	// DisableControlPanel skips serving and syncing the bundled management UI when true.
 	DisableControlPanel bool `yaml:"disable-control-panel"`
+	// PanelGitHubRepository overrides the GitHub repository used to fetch the management panel asset.
+	// Accepts either a repository URL (https://github.com/org/repo) or an API releases endpoint.
+	PanelGitHubRepository string `yaml:"panel-github-repository"`
 }
 
 // QuotaExceeded defines the behavior when API quota limits are exceeded.
@@ -182,6 +187,9 @@ type ClaudeKey struct {
 	// APIKey is the authentication key for accessing Claude API services.
 	APIKey string `yaml:"api-key" json:"api-key"`
 
+	// Prefix optionally namespaces models for this credential (e.g., "teamA/claude-sonnet-4").
+	Prefix string `yaml:"prefix,omitempty" json:"prefix,omitempty"`
+
 	// BaseURL is the base URL for the Claude API endpoint.
 	// If empty, the default Claude API URL will be used.
 	BaseURL string `yaml:"base-url" json:"base-url"`
@@ -214,6 +222,9 @@ type CodexKey struct {
 	// APIKey is the authentication key for accessing Codex API services.
 	APIKey string `yaml:"api-key" json:"api-key"`
 
+	// Prefix optionally namespaces models for this credential (e.g., "teamA/gpt-5-codex").
+	Prefix string `yaml:"prefix,omitempty" json:"prefix,omitempty"`
+
 	// BaseURL is the base URL for the Codex API endpoint.
 	// If empty, the default Codex API URL will be used.
 	BaseURL string `yaml:"base-url" json:"base-url"`
@@ -234,6 +245,9 @@ type GeminiKey struct {
 	// APIKey is the authentication key for accessing Gemini API services.
 	APIKey string `yaml:"api-key" json:"api-key"`
 
+	// Prefix optionally namespaces models for this credential (e.g., "teamA/gemini-3-pro-preview").
+	Prefix string `yaml:"prefix,omitempty" json:"prefix,omitempty"`
+
 	// BaseURL optionally overrides the Gemini API endpoint.
 	BaseURL string `yaml:"base-url,omitempty" json:"base-url,omitempty"`
 
@@ -253,6 +267,9 @@ type OpenAICompatibility struct {
 	// Name is the identifier for this OpenAI compatibility configuration.
 	Name string `yaml:"name" json:"name"`
 
+	// Prefix optionally namespaces model aliases for this provider (e.g., "teamA/kimi-k2").
+	Prefix string `yaml:"prefix,omitempty" json:"prefix,omitempty"`
+
 	// BaseURL is the base URL for the external OpenAI-compatible API endpoint.
 	BaseURL string `yaml:"base-url" json:"base-url"`
 
@@ -328,6 +345,7 @@ func LoadConfigOptional(configFile string, optional bool) (*Config, error) {
 	cfg.UsageStatisticsEnabled = false
 	cfg.DisableCooling = false
 	cfg.AmpCode.RestrictManagementToLocalhost = false // Default to false: API key auth is sufficient
+	cfg.RemoteManagement.PanelGitHubRepository = DefaultPanelGitHubRepository
 	if err = yaml.Unmarshal(data, &cfg); err != nil {
 		if optional {
 			// In cloud deploy mode, if YAML parsing fails, return empty config instead of error.
@@ -363,6 +381,11 @@ func LoadConfigOptional(configFile string, optional bool) (*Config, error) {
 		_ = SaveConfigPreserveCommentsUpdateNestedScalar(configFile, []string{"remote-management", "secret-key"}, hashed)
 	}
 
+	cfg.RemoteManagement.PanelGitHubRepository = strings.TrimSpace(cfg.RemoteManagement.PanelGitHubRepository)
+	if cfg.RemoteManagement.PanelGitHubRepository == "" {
+		cfg.RemoteManagement.PanelGitHubRepository = DefaultPanelGitHubRepository
+	}
+
 	// Sync request authentication providers with inline API keys for backwards compatibility.
 	syncInlineAccessProvider(&cfg)
 
@@ -411,6 +434,7 @@ func (cfg *Config) SanitizeOpenAICompatibility() {
 	for i := range cfg.OpenAICompatibility {
 		e := cfg.OpenAICompatibility[i]
 		e.Name = strings.TrimSpace(e.Name)
+		e.Prefix = normalizeModelPrefix(e.Prefix)
 		e.BaseURL = strings.TrimSpace(e.BaseURL)
 		e.Headers = NormalizeHeaders(e.Headers)
 		if e.BaseURL == "" {
@@ -431,6 +455,7 @@ func (cfg *Config) SanitizeCodexKeys() {
 	out := make([]CodexKey, 0, len(cfg.CodexKey))
 	for i := range cfg.CodexKey {
 		e := cfg.CodexKey[i]
+		e.Prefix = normalizeModelPrefix(e.Prefix)
 		e.BaseURL = strings.TrimSpace(e.BaseURL)
 		e.Headers = NormalizeHeaders(e.Headers)
 		e.ExcludedModels = NormalizeExcludedModels(e.ExcludedModels)
@@ -449,6 +474,7 @@ func (cfg *Config) SanitizeClaudeKeys() {
 	}
 	for i := range cfg.ClaudeKey {
 		entry := &cfg.ClaudeKey[i]
+		entry.Prefix = normalizeModelPrefix(entry.Prefix)
 		entry.Headers = NormalizeHeaders(entry.Headers)
 		entry.ExcludedModels = NormalizeExcludedModels(entry.ExcludedModels)
 	}
@@ -468,6 +494,7 @@ func (cfg *Config) SanitizeGeminiKeys() {
 		if entry.APIKey == "" {
 			continue
 		}
+		entry.Prefix = normalizeModelPrefix(entry.Prefix)
 		entry.BaseURL = strings.TrimSpace(entry.BaseURL)
 		entry.ProxyURL = strings.TrimSpace(entry.ProxyURL)
 		entry.Headers = NormalizeHeaders(entry.Headers)
@@ -481,6 +508,18 @@ func (cfg *Config) SanitizeGeminiKeys() {
 	cfg.GeminiKey = out
 }
 
+func normalizeModelPrefix(prefix string) string {
+	trimmed := strings.TrimSpace(prefix)
+	trimmed = strings.Trim(trimmed, "/")
+	if trimmed == "" {
+		return ""
+	}
+	if strings.Contains(trimmed, "/") {
+		return ""
+	}
+	return trimmed
+}
+
 func syncInlineAccessProvider(cfg *Config) {
 	if cfg == nil {
 		return

internal/config/vertex_compat.go

@@ -13,6 +13,9 @@ type VertexCompatKey struct {
 	// Maps to the x-goog-api-key header.
 	APIKey string `yaml:"api-key" json:"api-key"`
 
+	// Prefix optionally namespaces model aliases for this credential (e.g., "teamA/vertex-pro").
+	Prefix string `yaml:"prefix,omitempty" json:"prefix,omitempty"`
+
 	// BaseURL is the base URL for the Vertex-compatible API endpoint.
 	// The executor will append "/v1/publishers/google/models/{model}:action" to this.
 	// Example: "https://zenmux.ai/api" becomes "https://zenmux.ai/api/v1/publishers/google/models/..."
@@ -53,6 +56,7 @@ func (cfg *Config) SanitizeVertexCompatKeys() {
 		if entry.APIKey == "" {
 			continue
 		}
+		entry.Prefix = normalizeModelPrefix(entry.Prefix)
 		entry.BaseURL = strings.TrimSpace(entry.BaseURL)
 		if entry.BaseURL == "" {
 			// BaseURL is required for Vertex API key entries

internal/managementasset/updater.go

@@ -9,6 +9,7 @@ import (
 	"fmt"
 	"io"
 	"net/http"
+	"net/url"
 	"os"
 	"path/filepath"
 	"strings"
@@ -23,10 +24,10 @@ import (
 )
 
 const (
-	managementReleaseURL = "https://api.github.com/repos/router-for-me/Cli-Proxy-API-Management-Center/releases/latest"
-	managementAssetName  = "management.html"
-	httpUserAgent        = "CLIProxyAPI-management-updater"
-	updateCheckInterval  = 3 * time.Hour
+	defaultManagementReleaseURL = "https://api.github.com/repos/router-for-me/Cli-Proxy-API-Management-Center/releases/latest"
+	managementAssetName         = "management.html"
+	httpUserAgent               = "CLIProxyAPI-management-updater"
+	updateCheckInterval         = 3 * time.Hour
 )
 
 // ManagementFileName exposes the control panel asset filename.
@@ -97,7 +98,7 @@ func runAutoUpdater(ctx context.Context) {
 
 		configPath, _ := schedulerConfigPath.Load().(string)
 		staticDir := StaticDir(configPath)
-		EnsureLatestManagementHTML(ctx, staticDir, cfg.ProxyURL)
+		EnsureLatestManagementHTML(ctx, staticDir, cfg.ProxyURL, cfg.RemoteManagement.PanelGitHubRepository)
 	}
 
 	runOnce()
@@ -181,7 +182,7 @@ func FilePath(configFilePath string) string {
 // EnsureLatestManagementHTML checks the latest management.html asset and updates the local copy when needed.
 // The function is designed to run in a background goroutine and will never panic.
 // It enforces a 3-hour rate limit to avoid frequent checks on config/auth file changes.
-func EnsureLatestManagementHTML(ctx context.Context, staticDir string, proxyURL string) {
+func EnsureLatestManagementHTML(ctx context.Context, staticDir string, proxyURL string, panelRepository string) {
 	if ctx == nil {
 		ctx = context.Background()
 	}
@@ -214,6 +215,7 @@ func EnsureLatestManagementHTML(ctx context.Context, staticDir string, proxyURL
 		return
 	}
 
+	releaseURL := resolveReleaseURL(panelRepository)
 	client := newHTTPClient(proxyURL)
 
 	localPath := filepath.Join(staticDir, managementAssetName)
@@ -225,7 +227,7 @@ func EnsureLatestManagementHTML(ctx context.Context, staticDir string, proxyURL
 		localHash = ""
 	}
 
-	asset, remoteHash, err := fetchLatestAsset(ctx, client)
+	asset, remoteHash, err := fetchLatestAsset(ctx, client, releaseURL)
 	if err != nil {
 		log.WithError(err).Warn("failed to fetch latest management release information")
 		return
@@ -254,8 +256,44 @@ func EnsureLatestManagementHTML(ctx context.Context, staticDir string, proxyURL
 	log.Infof("management asset updated successfully (hash=%s)", downloadedHash)
 }
 
-func fetchLatestAsset(ctx context.Context, client *http.Client) (*releaseAsset, string, error) {
-	req, err := http.NewRequestWithContext(ctx, http.MethodGet, managementReleaseURL, nil)
+func resolveReleaseURL(repo string) string {
+	repo = strings.TrimSpace(repo)
+	if repo == "" {
+		return defaultManagementReleaseURL
+	}
+
+	parsed, err := url.Parse(repo)
+	if err != nil || parsed.Host == "" {
+		return defaultManagementReleaseURL
+	}
+
+	host := strings.ToLower(parsed.Host)
+	parsed.Path = strings.TrimSuffix(parsed.Path, "/")
+
+	if host == "api.github.com" {
+		if !strings.HasSuffix(strings.ToLower(parsed.Path), "/releases/latest") {
+			parsed.Path = parsed.Path + "/releases/latest"
+		}
+		return parsed.String()
+	}
+
+	if host == "github.com" {
+		parts := strings.Split(strings.Trim(parsed.Path, "/"), "/")
+		if len(parts) >= 2 && parts[0] != "" && parts[1] != "" {
+			repoName := strings.TrimSuffix(parts[1], ".git")
+			return fmt.Sprintf("https://api.github.com/repos/%s/%s/releases/latest", parts[0], repoName)
+		}
+	}
+
+	return defaultManagementReleaseURL
+}
+
+func fetchLatestAsset(ctx context.Context, client *http.Client, releaseURL string) (*releaseAsset, string, error) {
+	if strings.TrimSpace(releaseURL) == "" {
+		releaseURL = defaultManagementReleaseURL
+	}
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, releaseURL, nil)
 	if err != nil {
 		return nil, "", fmt.Errorf("create release request: %w", err)
 	}

internal/registry/model_definitions.go

@@ -630,6 +630,13 @@ func GetQwenModels() []*ModelInfo {
 	}
 }
 
+// iFlowThinkingSupport is a shared ThinkingSupport configuration for iFlow models
+// that support thinking mode via chat_template_kwargs.enable_thinking (boolean toggle).
+// Uses level-based configuration so standard normalization flows apply before conversion.
+var iFlowThinkingSupport = &ThinkingSupport{
+	Levels: []string{"none", "auto", "minimal", "low", "medium", "high", "xhigh"},
+}
+
 // GetIFlowModels returns supported models for iFlow OAuth accounts.
 func GetIFlowModels() []*ModelInfo {
 	entries := []struct {
@@ -645,9 +652,9 @@ func GetIFlowModels() []*ModelInfo {
 		{ID: "qwen3-vl-plus", DisplayName: "Qwen3-VL-Plus", Description: "Qwen3 multimodal vision-language", Created: 1758672000},
 		{ID: "qwen3-max-preview", DisplayName: "Qwen3-Max-Preview", Description: "Qwen3 Max preview build", Created: 1757030400},
 		{ID: "kimi-k2-0905", DisplayName: "Kimi-K2-Instruct-0905", Description: "Moonshot Kimi K2 instruct 0905", Created: 1757030400},
-		{ID: "glm-4.6", DisplayName: "GLM-4.6", Description: "Zhipu GLM 4.6 general model", Created: 1759190400},
+		{ID: "glm-4.6", DisplayName: "GLM-4.6", Description: "Zhipu GLM 4.6 general model", Created: 1759190400, Thinking: iFlowThinkingSupport},
 		{ID: "kimi-k2", DisplayName: "Kimi-K2", Description: "Moonshot Kimi K2 general model", Created: 1752192000},
-		{ID: "kimi-k2-thinking", DisplayName: "Kimi-K2-Thinking", Description: "Moonshot Kimi K2 thinking model", Created: 1762387200, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}}},
+		{ID: "kimi-k2-thinking", DisplayName: "Kimi-K2-Thinking", Description: "Moonshot Kimi K2 thinking model", Created: 1762387200},
 		{ID: "deepseek-v3.2-chat", DisplayName: "DeepSeek-V3.2", Description: "DeepSeek V3.2 Chat", Created: 1764576000},
 		{ID: "deepseek-v3.2-reasoner", DisplayName: "DeepSeek-V3.2", Description: "DeepSeek V3.2 Reasoner", Created: 1764576000},
 		{ID: "deepseek-v3.2", DisplayName: "DeepSeek-V3.2-Exp", Description: "DeepSeek V3.2 experimental", Created: 1759104000},
@@ -655,10 +662,10 @@ func GetIFlowModels() []*ModelInfo {
 		{ID: "deepseek-r1", DisplayName: "DeepSeek-R1", Description: "DeepSeek reasoning model R1", Created: 1737331200},
 		{ID: "deepseek-v3", DisplayName: "DeepSeek-V3-671B", Description: "DeepSeek V3 671B", Created: 1734307200},
 		{ID: "qwen3-32b", DisplayName: "Qwen3-32B", Description: "Qwen3 32B", Created: 1747094400},
-		{ID: "qwen3-235b-a22b-thinking-2507", DisplayName: "Qwen3-235B-A22B-Thinking", Description: "Qwen3 235B A22B Thinking (2507)", Created: 1753401600, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}}},
+		{ID: "qwen3-235b-a22b-thinking-2507", DisplayName: "Qwen3-235B-A22B-Thinking", Description: "Qwen3 235B A22B Thinking (2507)", Created: 1753401600},
 		{ID: "qwen3-235b-a22b-instruct", DisplayName: "Qwen3-235B-A22B-Instruct", Description: "Qwen3 235B A22B Instruct", Created: 1753401600},
 		{ID: "qwen3-235b", DisplayName: "Qwen3-235B-A22B", Description: "Qwen3 235B A22B", Created: 1753401600},
-		{ID: "minimax-m2", DisplayName: "MiniMax-M2", Description: "MiniMax M2", Created: 1758672000, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}}},
+		{ID: "minimax-m2", DisplayName: "MiniMax-M2", Description: "MiniMax M2", Created: 1758672000},
 	}
 	models := make([]*ModelInfo, 0, len(entries))
 	for _, entry := range entries {

internal/runtime/executor/iflow_executor.go

@@ -66,6 +66,7 @@ func (e *IFlowExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	if errValidate := ValidateThinkingConfig(body, upstreamModel); errValidate != nil {
 		return resp, errValidate
 	}
+	body = applyIFlowThinkingConfig(body)
 	body = applyPayloadConfig(e.cfg, req.Model, body)
 
 	endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint
@@ -157,6 +158,7 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	if errValidate := ValidateThinkingConfig(body, upstreamModel); errValidate != nil {
 		return nil, errValidate
 	}
+	body = applyIFlowThinkingConfig(body)
 	// Ensure tools array exists to avoid provider quirks similar to Qwen's behaviour.
 	toolsResult := gjson.GetBytes(body, "tools")
 	if toolsResult.Exists() && toolsResult.IsArray() && len(toolsResult.Array()) == 0 {
@@ -442,3 +444,21 @@ func ensureToolsArray(body []byte) []byte {
 	}
 	return updated
 }
+
+// applyIFlowThinkingConfig converts normalized reasoning_effort to iFlow chat_template_kwargs.enable_thinking.
+// This should be called after NormalizeThinkingConfig has processed the payload.
+// iFlow only supports boolean enable_thinking, so any non-"none" effort enables thinking.
+func applyIFlowThinkingConfig(body []byte) []byte {
+	effort := gjson.GetBytes(body, "reasoning_effort")
+	if !effort.Exists() {
+		return body
+	}
+
+	val := strings.ToLower(strings.TrimSpace(effort.String()))
+	enableThinking := val != "none" && val != ""
+
+	body, _ = sjson.DeleteBytes(body, "reasoning_effort")
+	body, _ = sjson.SetBytes(body, "chat_template_kwargs.enable_thinking", enableThinking)
+
+	return body
+}

internal/runtime/executor/payload_helpers.go

@@ -72,13 +72,7 @@ func ApplyReasoningEffortMetadata(payload []byte, metadata map[string]any, model
 	// Fallback: numeric thinking_budget suffix for level-based (OpenAI-style) models.
 	if util.ModelUsesThinkingLevels(baseModel) || allowCompat {
 		if budget, _, _, matched := util.ThinkingFromMetadata(metadata); matched && budget != nil {
-			if effort, ok := util.OpenAIThinkingBudgetToEffort(baseModel, *budget); ok && effort != "" {
-				if *budget == 0 && effort == "none" && util.ModelUsesThinkingLevels(baseModel) {
-					if _, supported := util.NormalizeReasoningEffortLevel(baseModel, effort); !supported {
-						return StripThinkingFields(payload, false)
-					}
-				}
-
+			if effort, ok := util.ThinkingBudgetToEffort(baseModel, *budget); ok && effort != "" {
 				if updated, err := sjson.SetBytes(payload, field, effort); err == nil {
 					return updated
 				}
@@ -273,7 +267,7 @@ func StripThinkingFields(payload []byte, effortOnly bool) []byte {
 		"reasoning.effort",
 	}
 	if !effortOnly {
-		fieldsToRemove = append([]string{"reasoning"}, fieldsToRemove...)
+		fieldsToRemove = append([]string{"reasoning", "thinking"}, fieldsToRemove...)
 	}
 	out := payload
 	for _, field := range fieldsToRemove {

internal/translator/antigravity/claude/antigravity_claude_request.go

@@ -84,13 +84,18 @@ func ConvertClaudeRequestToAntigravity(modelName string, inputRawJSON []byte, _
 					contentResult := contentResults[j]
 					contentTypeResult := contentResult.Get("type")
 					if contentTypeResult.Type == gjson.String && contentTypeResult.String() == "thinking" {
-						prompt := contentResult.Get("thinking").String()
+						// Claude "thinking" blocks are internal-only. They also require a valid provider signature
+						// when replayed as conversation history. Since we cannot mint signatures, only forward
+						// thinking blocks when the client provides a non-empty signature; otherwise, drop them.
 						signatureResult := contentResult.Get("signature")
-						signature := geminiCLIClaudeThoughtSignature
-						if signatureResult.Exists() {
-							signature = signatureResult.String()
+						if signatureResult.Type == gjson.String && signatureResult.String() != "" {
+							prompt := contentResult.Get("thinking").String()
+							clientContent.Parts = append(clientContent.Parts, client.Part{
+								Text:             prompt,
+								Thought:          true,
+								ThoughtSignature: signatureResult.String(),
+							})
 						}
-						clientContent.Parts = append(clientContent.Parts, client.Part{Text: prompt, Thought: true, ThoughtSignature: signature})
 					} else if contentTypeResult.Type == gjson.String && contentTypeResult.String() == "text" {
 						prompt := contentResult.Get("text").String()
 						clientContent.Parts = append(clientContent.Parts, client.Part{Text: prompt})
@@ -117,9 +122,17 @@ func ConvertClaudeRequestToAntigravity(modelName string, inputRawJSON []byte, _
 							funcName := toolCallID
 							toolCallIDs := strings.Split(toolCallID, "-")
 							if len(toolCallIDs) > 1 {
-								funcName = strings.Join(toolCallIDs[0:len(toolCallIDs)-1], "-")
+								funcName = strings.Join(toolCallIDs[0:len(toolCallIDs)-2], "-")
 							}
-							responseData := contentResult.Get("content").Raw
+							functionResponseResult := contentResult.Get("content")
+
+							responseData := ""
+							if functionResponseResult.Type == gjson.String {
+								responseData = functionResponseResult.String()
+							} else {
+								responseData = contentResult.Get("content").Raw
+							}
+
 							functionResponse := client.FunctionResponse{ID: toolCallID, Name: funcName, Response: map[string]interface{}{"result": responseData}}
 							clientContent.Parts = append(clientContent.Parts, client.Part{FunctionResponse: &functionResponse})
 						}
@@ -134,7 +147,9 @@ func ConvertClaudeRequestToAntigravity(modelName string, inputRawJSON []byte, _
 						}
 					}
 				}
-				contents = append(contents, clientContent)
+				if len(clientContent.Parts) > 0 {
+					contents = append(contents, clientContent)
+				}
 			} else if contentsResult.Type == gjson.String {
 				prompt := contentsResult.String()
 				contents = append(contents, client.Content{Role: role, Parts: []client.Part{{Text: prompt}}})

internal/translator/antigravity/claude/antigravity_claude_response.go

@@ -114,44 +114,54 @@ func ConvertAntigravityResponseToClaude(_ context.Context, _ string, originalReq
 			// Extract the different types of content from each part
 			partTextResult := partResult.Get("text")
 			functionCallResult := partResult.Get("functionCall")
+			thoughtSignatureResult := partResult.Get("thoughtSignature")
+			if !thoughtSignatureResult.Exists() {
+				thoughtSignatureResult = partResult.Get("thought_signature")
+			}
+			hasThoughtSignature := thoughtSignatureResult.Exists() && thoughtSignatureResult.String() != ""
+			isThought := partResult.Get("thought").Bool()
+
+			// Some Antigravity/Vertex Claude streams emit the thought signature as a standalone part
+			// (no text payload). Claude requires this signature to be replayed verbatim on subsequent turns.
+			if isThought && hasThoughtSignature && !partTextResult.Exists() && !functionCallResult.Exists() {
+				if params.ResponseType == 2 {
+					output = output + "event: content_block_delta\n"
+					data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":""}}`, params.ResponseIndex), "delta.signature", thoughtSignatureResult.String())
+					output = output + fmt.Sprintf("data: %s\n\n\n", data)
+					params.HasContent = true
+				}
+				continue
+			}
 
 			// Handle text content (both regular content and thinking)
 			if partTextResult.Exists() {
 				// Process thinking content (internal reasoning)
-				if partResult.Get("thought").Bool() {
-					if thoughtSignature := partResult.Get("thoughtSignature"); thoughtSignature.Exists() && thoughtSignature.String() != "" {
-						output = output + "event: content_block_delta\n"
-						data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":""}}`, params.ResponseIndex), "delta.signature", thoughtSignature.String())
-						output = output + fmt.Sprintf("data: %s\n\n\n", data)
-						params.HasContent = true
-					} else if params.ResponseType == 2 { // Continue existing thinking block if already in thinking state
-						output = output + "event: content_block_delta\n"
-						data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, params.ResponseIndex), "delta.thinking", partTextResult.String())
-						output = output + fmt.Sprintf("data: %s\n\n\n", data)
-						params.HasContent = true
-					} else {
-						// Transition from another state to thinking
-						// First, close any existing content block
+				if isThought {
+					// Ensure we have an open thinking block to attach thinking/signature deltas to.
+					if params.ResponseType != 2 {
 						if params.ResponseType != 0 {
-							if params.ResponseType == 2 {
-								// output = output + "event: content_block_delta\n"
-								// output = output + fmt.Sprintf(`data: {"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":null}}`, params.ResponseIndex)
-								// output = output + "\n\n\n"
-							}
 							output = output + "event: content_block_stop\n"
 							output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, params.ResponseIndex)
 							output = output + "\n\n\n"
 							params.ResponseIndex++
 						}
-
-						// Start a new thinking content block
 						output = output + "event: content_block_start\n"
 						output = output + fmt.Sprintf(`data: {"type":"content_block_start","index":%d,"content_block":{"type":"thinking","thinking":""}}`, params.ResponseIndex)
 						output = output + "\n\n\n"
+						params.ResponseType = 2
+					}
+
+					if partTextResult.String() != "" {
 						output = output + "event: content_block_delta\n"
 						data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, params.ResponseIndex), "delta.thinking", partTextResult.String())
 						output = output + fmt.Sprintf("data: %s\n\n\n", data)
-						params.ResponseType = 2 // Set state to thinking
+						params.HasContent = true
+					}
+
+					if hasThoughtSignature {
+						output = output + "event: content_block_delta\n"
+						data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":""}}`, params.ResponseIndex), "delta.signature", thoughtSignatureResult.String())
+						output = output + fmt.Sprintf("data: %s\n\n\n", data)
 						params.HasContent = true
 					}
 				} else {
@@ -368,6 +378,7 @@ func ConvertAntigravityResponseToClaudeNonStream(_ context.Context, _ string, or
 	var contentBlocks []interface{}
 	textBuilder := strings.Builder{}
 	thinkingBuilder := strings.Builder{}
+	thinkingSignature := ""
 	toolIDCounter := 0
 	hasToolCall := false
 
@@ -386,19 +397,37 @@ func ConvertAntigravityResponseToClaudeNonStream(_ context.Context, _ string, or
 		if thinkingBuilder.Len() == 0 {
 			return
 		}
-		contentBlocks = append(contentBlocks, map[string]interface{}{
+		block := map[string]interface{}{
 			"type":     "thinking",
 			"thinking": thinkingBuilder.String(),
-		})
+		}
+		if thinkingSignature != "" {
+			block["signature"] = thinkingSignature
+		}
+		contentBlocks = append(contentBlocks, block)
 		thinkingBuilder.Reset()
+		thinkingSignature = ""
 	}
 
 	if parts.IsArray() {
 		for _, part := range parts.Array() {
+			thoughtSignatureResult := part.Get("thoughtSignature")
+			if !thoughtSignatureResult.Exists() {
+				thoughtSignatureResult = part.Get("thought_signature")
+			}
+			if part.Get("thought").Bool() && thoughtSignatureResult.Exists() && thoughtSignatureResult.String() != "" && (!part.Get("text").Exists() || part.Get("text").String() == "") {
+				// Signature-only thought part (no text payload).
+				thinkingSignature = thoughtSignatureResult.String()
+				continue
+			}
+
 			if text := part.Get("text"); text.Exists() && text.String() != "" {
 				if part.Get("thought").Bool() {
 					flushText()
 					thinkingBuilder.WriteString(text.String())
+					if thoughtSignatureResult.Exists() && thoughtSignatureResult.String() != "" {
+						thinkingSignature = thoughtSignatureResult.String()
+					}
 					continue
 				}
 				flushThinking()

internal/translator/codex/claude/codex_claude_request.go

@@ -219,15 +219,20 @@ func ConvertClaudeRequestToCodex(modelName string, inputRawJSON []byte, _ bool)
 	// Convert thinking.budget_tokens to reasoning.effort for level-based models
 	reasoningEffort := "medium" // default
 	if thinking := rootResult.Get("thinking"); thinking.Exists() && thinking.IsObject() {
-		if thinking.Get("type").String() == "enabled" {
+		switch thinking.Get("type").String() {
+		case "enabled":
 			if util.ModelUsesThinkingLevels(modelName) {
 				if budgetTokens := thinking.Get("budget_tokens"); budgetTokens.Exists() {
 					budget := int(budgetTokens.Int())
-					if effort, ok := util.OpenAIThinkingBudgetToEffort(modelName, budget); ok && effort != "" {
+					if effort, ok := util.ThinkingBudgetToEffort(modelName, budget); ok && effort != "" {
 						reasoningEffort = effort
 					}
 				}
 			}
+		case "disabled":
+			if effort, ok := util.ThinkingBudgetToEffort(modelName, 0); ok && effort != "" {
+				reasoningEffort = effort
+			}
 		}
 	}
 	template, _ = sjson.Set(template, "reasoning.effort", reasoningEffort)

internal/translator/codex/gemini/codex_gemini_request.go

@@ -253,7 +253,7 @@ func ConvertGeminiRequestToCodex(modelName string, inputRawJSON []byte, _ bool)
 			if util.ModelUsesThinkingLevels(modelName) {
 				if thinkingBudget := thinkingConfig.Get("thinkingBudget"); thinkingBudget.Exists() {
 					budget := int(thinkingBudget.Int())
-					if effort, ok := util.OpenAIThinkingBudgetToEffort(modelName, budget); ok && effort != "" {
+					if effort, ok := util.ThinkingBudgetToEffort(modelName, budget); ok && effort != "" {
 						reasoningEffort = effort
 					}
 				}

internal/translator/openai/claude/openai_claude_request.go

@@ -63,10 +63,22 @@ func ConvertClaudeRequestToOpenAI(modelName string, inputRawJSON []byte, stream
 
 	// Thinking: Convert Claude thinking.budget_tokens to OpenAI reasoning_effort
 	if thinking := root.Get("thinking"); thinking.Exists() && thinking.IsObject() {
-		if thinkingType := thinking.Get("type"); thinkingType.Exists() && thinkingType.String() == "enabled" {
-			if budgetTokens := thinking.Get("budget_tokens"); budgetTokens.Exists() {
-				budget := int(budgetTokens.Int())
-				if effort, ok := util.OpenAIThinkingBudgetToEffort(modelName, budget); ok && effort != "" {
+		if thinkingType := thinking.Get("type"); thinkingType.Exists() {
+			switch thinkingType.String() {
+			case "enabled":
+				if budgetTokens := thinking.Get("budget_tokens"); budgetTokens.Exists() {
+					budget := int(budgetTokens.Int())
+					if effort, ok := util.ThinkingBudgetToEffort(modelName, budget); ok && effort != "" {
+						out, _ = sjson.Set(out, "reasoning_effort", effort)
+					}
+				} else {
+					// No budget_tokens specified, default to "auto" for enabled thinking
+					if effort, ok := util.ThinkingBudgetToEffort(modelName, -1); ok && effort != "" {
+						out, _ = sjson.Set(out, "reasoning_effort", effort)
+					}
+				}
+			case "disabled":
+				if effort, ok := util.ThinkingBudgetToEffort(modelName, 0); ok && effort != "" {
 					out, _ = sjson.Set(out, "reasoning_effort", effort)
 				}
 			}

internal/translator/openai/claude/openai_claude_response.go

@@ -128,9 +128,10 @@ func convertOpenAIStreamingChunkToAnthropic(rawJSON []byte, param *ConvertOpenAI
 		param.CreatedAt = root.Get("created").Int()
 	}
 
-	// Check if this is the first chunk (has role)
+	// Emit message_start on the very first chunk, regardless of whether it has a role field.
+	// Some providers (like Copilot) may send tool_calls in the first chunk without a role field.
 	if delta := root.Get("choices.0.delta"); delta.Exists() {
-		if role := delta.Get("role"); role.Exists() && role.String() == "assistant" && !param.MessageStarted {
+		if !param.MessageStarted {
 			// Send message_start event
 			messageStart := map[string]interface{}{
 				"type": "message_start",

internal/translator/openai/gemini/openai_gemini_request.go

@@ -83,7 +83,7 @@ func ConvertGeminiRequestToOpenAI(modelName string, inputRawJSON []byte, stream
 		if thinkingConfig := genConfig.Get("thinkingConfig"); thinkingConfig.Exists() && thinkingConfig.IsObject() {
 			if thinkingBudget := thinkingConfig.Get("thinkingBudget"); thinkingBudget.Exists() {
 				budget := int(thinkingBudget.Int())
-				if effort, ok := util.OpenAIThinkingBudgetToEffort(modelName, budget); ok && effort != "" {
+				if effort, ok := util.ThinkingBudgetToEffort(modelName, budget); ok && effort != "" {
 					out, _ = sjson.Set(out, "reasoning_effort", effort)
 				}
 			}

internal/util/openai_thinking.go

@@ -1,37 +0,0 @@
-package util
-
-// OpenAIThinkingBudgetToEffort maps a numeric thinking budget (tokens)
-// into an OpenAI-style reasoning effort level for level-based models.
-//
-// Ranges:
-//   - 0            -> "none"
-//   - -1           -> "auto"
-//   - 1..1024      -> "low"
-//   - 1025..8192   -> "medium"
-//   - 8193..24576  -> "high"
-//   - 24577..      -> highest supported level for the model (defaults to "xhigh")
-//
-// Negative values other than -1 are treated as unsupported.
-func OpenAIThinkingBudgetToEffort(model string, budget int) (string, bool) {
-	switch {
-	case budget == -1:
-		return "auto", true
-	case budget < -1:
-		return "", false
-	case budget == 0:
-		return "none", true
-	case budget > 0 && budget <= 1024:
-		return "low", true
-	case budget <= 8192:
-		return "medium", true
-	case budget <= 24576:
-		return "high", true
-	case budget > 24576:
-		if levels := GetModelThinkingLevels(model); len(levels) > 0 {
-			return levels[len(levels)-1], true
-		}
-		return "xhigh", true
-	default:
-		return "", false
-	}
-}

internal/util/thinking.go

@@ -118,3 +118,83 @@ func IsOpenAICompatibilityModel(model string) bool {
 	}
 	return strings.EqualFold(strings.TrimSpace(info.Type), "openai-compatibility")
 }
+
+// ThinkingEffortToBudget maps a reasoning effort level to a numeric thinking budget (tokens),
+// clamping the result to the model's supported range.
+//
+// Mappings (values are normalized to model's supported range):
+//   - "none"    -> 0
+//   - "auto"    -> -1
+//   - "minimal" -> 512
+//   - "low"     -> 1024
+//   - "medium"  -> 8192
+//   - "high"    -> 24576
+//   - "xhigh"   -> 32768
+//
+// Returns false when the effort level is empty or unsupported.
+func ThinkingEffortToBudget(model, effort string) (int, bool) {
+	if effort == "" {
+		return 0, false
+	}
+	normalized, ok := NormalizeReasoningEffortLevel(model, effort)
+	if !ok {
+		normalized = strings.ToLower(strings.TrimSpace(effort))
+	}
+	switch normalized {
+	case "none":
+		return 0, true
+	case "auto":
+		return NormalizeThinkingBudget(model, -1), true
+	case "minimal":
+		return NormalizeThinkingBudget(model, 512), true
+	case "low":
+		return NormalizeThinkingBudget(model, 1024), true
+	case "medium":
+		return NormalizeThinkingBudget(model, 8192), true
+	case "high":
+		return NormalizeThinkingBudget(model, 24576), true
+	case "xhigh":
+		return NormalizeThinkingBudget(model, 32768), true
+	default:
+		return 0, false
+	}
+}
+
+// ThinkingBudgetToEffort maps a numeric thinking budget (tokens)
+// to a reasoning effort level for level-based models.
+//
+// Mappings:
+//   - 0            -> "none" (or lowest supported level if model doesn't support "none")
+//   - -1           -> "auto"
+//   - 1..1024      -> "low"
+//   - 1025..8192   -> "medium"
+//   - 8193..24576  -> "high"
+//   - 24577..      -> highest supported level for the model (defaults to "xhigh")
+//
+// Returns false when the budget is unsupported (negative values other than -1).
+func ThinkingBudgetToEffort(model string, budget int) (string, bool) {
+	switch {
+	case budget == -1:
+		return "auto", true
+	case budget < -1:
+		return "", false
+	case budget == 0:
+		if levels := GetModelThinkingLevels(model); len(levels) > 0 {
+			return levels[0], true
+		}
+		return "none", true
+	case budget > 0 && budget <= 1024:
+		return "low", true
+	case budget <= 8192:
+		return "medium", true
+	case budget <= 24576:
+		return "high", true
+	case budget > 24576:
+		if levels := GetModelThinkingLevels(model); len(levels) > 0 {
+			return levels[len(levels)-1], true
+		}
+		return "xhigh", true
+	default:
+		return "", false
+	}
+}

internal/util/thinking_suffix.go

@@ -201,36 +201,6 @@ func ReasoningEffortFromMetadata(metadata map[string]any) (string, bool) {
 	return "", true
 }
 
-// ThinkingEffortToBudget maps reasoning effort levels to approximate budgets,
-// clamping the result to the model's supported range.
-func ThinkingEffortToBudget(model, effort string) (int, bool) {
-	if effort == "" {
-		return 0, false
-	}
-	normalized, ok := NormalizeReasoningEffortLevel(model, effort)
-	if !ok {
-		normalized = strings.ToLower(strings.TrimSpace(effort))
-	}
-	switch normalized {
-	case "none":
-		return 0, true
-	case "auto":
-		return NormalizeThinkingBudget(model, -1), true
-	case "minimal":
-		return NormalizeThinkingBudget(model, 512), true
-	case "low":
-		return NormalizeThinkingBudget(model, 1024), true
-	case "medium":
-		return NormalizeThinkingBudget(model, 8192), true
-	case "high":
-		return NormalizeThinkingBudget(model, 24576), true
-	case "xhigh":
-		return NormalizeThinkingBudget(model, 32768), true
-	default:
-		return 0, false
-	}
-}
-
 // ResolveOriginalModel returns the original model name stored in metadata (if present),
 // otherwise falls back to the provided model.
 func ResolveOriginalModel(model string, metadata map[string]any) string {

internal/watcher/diff/config_diff.go

@@ -0,0 +1,303 @@
+package diff
+
+import (
+	"fmt"
+	"net/url"
+	"reflect"
+	"strings"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+)
+
+// BuildConfigChangeDetails computes a redacted, human-readable list of config changes.
+// Secrets are never printed; only structural or non-sensitive fields are surfaced.
+func BuildConfigChangeDetails(oldCfg, newCfg *config.Config) []string {
+	changes := make([]string, 0, 16)
+	if oldCfg == nil || newCfg == nil {
+		return changes
+	}
+
+	// Simple scalars
+	if oldCfg.Port != newCfg.Port {
+		changes = append(changes, fmt.Sprintf("port: %d -> %d", oldCfg.Port, newCfg.Port))
+	}
+	if oldCfg.AuthDir != newCfg.AuthDir {
+		changes = append(changes, fmt.Sprintf("auth-dir: %s -> %s", oldCfg.AuthDir, newCfg.AuthDir))
+	}
+	if oldCfg.Debug != newCfg.Debug {
+		changes = append(changes, fmt.Sprintf("debug: %t -> %t", oldCfg.Debug, newCfg.Debug))
+	}
+	if oldCfg.LoggingToFile != newCfg.LoggingToFile {
+		changes = append(changes, fmt.Sprintf("logging-to-file: %t -> %t", oldCfg.LoggingToFile, newCfg.LoggingToFile))
+	}
+	if oldCfg.UsageStatisticsEnabled != newCfg.UsageStatisticsEnabled {
+		changes = append(changes, fmt.Sprintf("usage-statistics-enabled: %t -> %t", oldCfg.UsageStatisticsEnabled, newCfg.UsageStatisticsEnabled))
+	}
+	if oldCfg.DisableCooling != newCfg.DisableCooling {
+		changes = append(changes, fmt.Sprintf("disable-cooling: %t -> %t", oldCfg.DisableCooling, newCfg.DisableCooling))
+	}
+	if oldCfg.RequestLog != newCfg.RequestLog {
+		changes = append(changes, fmt.Sprintf("request-log: %t -> %t", oldCfg.RequestLog, newCfg.RequestLog))
+	}
+	if oldCfg.RequestRetry != newCfg.RequestRetry {
+		changes = append(changes, fmt.Sprintf("request-retry: %d -> %d", oldCfg.RequestRetry, newCfg.RequestRetry))
+	}
+	if oldCfg.MaxRetryInterval != newCfg.MaxRetryInterval {
+		changes = append(changes, fmt.Sprintf("max-retry-interval: %d -> %d", oldCfg.MaxRetryInterval, newCfg.MaxRetryInterval))
+	}
+	if oldCfg.ProxyURL != newCfg.ProxyURL {
+		changes = append(changes, fmt.Sprintf("proxy-url: %s -> %s", formatProxyURL(oldCfg.ProxyURL), formatProxyURL(newCfg.ProxyURL)))
+	}
+	if oldCfg.WebsocketAuth != newCfg.WebsocketAuth {
+		changes = append(changes, fmt.Sprintf("ws-auth: %t -> %t", oldCfg.WebsocketAuth, newCfg.WebsocketAuth))
+	}
+	if oldCfg.ForceModelPrefix != newCfg.ForceModelPrefix {
+		changes = append(changes, fmt.Sprintf("force-model-prefix: %t -> %t", oldCfg.ForceModelPrefix, newCfg.ForceModelPrefix))
+	}
+
+	// Quota-exceeded behavior
+	if oldCfg.QuotaExceeded.SwitchProject != newCfg.QuotaExceeded.SwitchProject {
+		changes = append(changes, fmt.Sprintf("quota-exceeded.switch-project: %t -> %t", oldCfg.QuotaExceeded.SwitchProject, newCfg.QuotaExceeded.SwitchProject))
+	}
+	if oldCfg.QuotaExceeded.SwitchPreviewModel != newCfg.QuotaExceeded.SwitchPreviewModel {
+		changes = append(changes, fmt.Sprintf("quota-exceeded.switch-preview-model: %t -> %t", oldCfg.QuotaExceeded.SwitchPreviewModel, newCfg.QuotaExceeded.SwitchPreviewModel))
+	}
+
+	// API keys (redacted) and counts
+	if len(oldCfg.APIKeys) != len(newCfg.APIKeys) {
+		changes = append(changes, fmt.Sprintf("api-keys count: %d -> %d", len(oldCfg.APIKeys), len(newCfg.APIKeys)))
+	} else if !reflect.DeepEqual(trimStrings(oldCfg.APIKeys), trimStrings(newCfg.APIKeys)) {
+		changes = append(changes, "api-keys: values updated (count unchanged, redacted)")
+	}
+	if len(oldCfg.GeminiKey) != len(newCfg.GeminiKey) {
+		changes = append(changes, fmt.Sprintf("gemini-api-key count: %d -> %d", len(oldCfg.GeminiKey), len(newCfg.GeminiKey)))
+	} else {
+		for i := range oldCfg.GeminiKey {
+			o := oldCfg.GeminiKey[i]
+			n := newCfg.GeminiKey[i]
+			if strings.TrimSpace(o.BaseURL) != strings.TrimSpace(n.BaseURL) {
+				changes = append(changes, fmt.Sprintf("gemini[%d].base-url: %s -> %s", i, strings.TrimSpace(o.BaseURL), strings.TrimSpace(n.BaseURL)))
+			}
+			if strings.TrimSpace(o.ProxyURL) != strings.TrimSpace(n.ProxyURL) {
+				changes = append(changes, fmt.Sprintf("gemini[%d].proxy-url: %s -> %s", i, formatProxyURL(o.ProxyURL), formatProxyURL(n.ProxyURL)))
+			}
+			if strings.TrimSpace(o.Prefix) != strings.TrimSpace(n.Prefix) {
+				changes = append(changes, fmt.Sprintf("gemini[%d].prefix: %s -> %s", i, formatProxyURL(o.Prefix), formatProxyURL(n.Prefix)))
+			}
+			if strings.TrimSpace(o.APIKey) != strings.TrimSpace(n.APIKey) {
+				changes = append(changes, fmt.Sprintf("gemini[%d].api-key: updated", i))
+			}
+			if !equalStringMap(o.Headers, n.Headers) {
+				changes = append(changes, fmt.Sprintf("gemini[%d].headers: updated", i))
+			}
+			oldExcluded := SummarizeExcludedModels(o.ExcludedModels)
+			newExcluded := SummarizeExcludedModels(n.ExcludedModels)
+			if oldExcluded.hash != newExcluded.hash {
+				changes = append(changes, fmt.Sprintf("gemini[%d].excluded-models: updated (%d -> %d entries)", i, oldExcluded.count, newExcluded.count))
+			}
+		}
+	}
+
+	// Claude keys (do not print key material)
+	if len(oldCfg.ClaudeKey) != len(newCfg.ClaudeKey) {
+		changes = append(changes, fmt.Sprintf("claude-api-key count: %d -> %d", len(oldCfg.ClaudeKey), len(newCfg.ClaudeKey)))
+	} else {
+		for i := range oldCfg.ClaudeKey {
+			o := oldCfg.ClaudeKey[i]
+			n := newCfg.ClaudeKey[i]
+			if strings.TrimSpace(o.BaseURL) != strings.TrimSpace(n.BaseURL) {
+				changes = append(changes, fmt.Sprintf("claude[%d].base-url: %s -> %s", i, strings.TrimSpace(o.BaseURL), strings.TrimSpace(n.BaseURL)))
+			}
+			if strings.TrimSpace(o.ProxyURL) != strings.TrimSpace(n.ProxyURL) {
+				changes = append(changes, fmt.Sprintf("claude[%d].proxy-url: %s -> %s", i, formatProxyURL(o.ProxyURL), formatProxyURL(n.ProxyURL)))
+			}
+			if strings.TrimSpace(o.Prefix) != strings.TrimSpace(n.Prefix) {
+				changes = append(changes, fmt.Sprintf("claude[%d].prefix: %s -> %s", i, formatProxyURL(o.Prefix), formatProxyURL(n.Prefix)))
+			}
+			if strings.TrimSpace(o.APIKey) != strings.TrimSpace(n.APIKey) {
+				changes = append(changes, fmt.Sprintf("claude[%d].api-key: updated", i))
+			}
+			if !equalStringMap(o.Headers, n.Headers) {
+				changes = append(changes, fmt.Sprintf("claude[%d].headers: updated", i))
+			}
+			oldExcluded := SummarizeExcludedModels(o.ExcludedModels)
+			newExcluded := SummarizeExcludedModels(n.ExcludedModels)
+			if oldExcluded.hash != newExcluded.hash {
+				changes = append(changes, fmt.Sprintf("claude[%d].excluded-models: updated (%d -> %d entries)", i, oldExcluded.count, newExcluded.count))
+			}
+		}
+	}
+
+	// Codex keys (do not print key material)
+	if len(oldCfg.CodexKey) != len(newCfg.CodexKey) {
+		changes = append(changes, fmt.Sprintf("codex-api-key count: %d -> %d", len(oldCfg.CodexKey), len(newCfg.CodexKey)))
+	} else {
+		for i := range oldCfg.CodexKey {
+			o := oldCfg.CodexKey[i]
+			n := newCfg.CodexKey[i]
+			if strings.TrimSpace(o.BaseURL) != strings.TrimSpace(n.BaseURL) {
+				changes = append(changes, fmt.Sprintf("codex[%d].base-url: %s -> %s", i, strings.TrimSpace(o.BaseURL), strings.TrimSpace(n.BaseURL)))
+			}
+			if strings.TrimSpace(o.ProxyURL) != strings.TrimSpace(n.ProxyURL) {
+				changes = append(changes, fmt.Sprintf("codex[%d].proxy-url: %s -> %s", i, formatProxyURL(o.ProxyURL), formatProxyURL(n.ProxyURL)))
+			}
+			if strings.TrimSpace(o.Prefix) != strings.TrimSpace(n.Prefix) {
+				changes = append(changes, fmt.Sprintf("codex[%d].prefix: %s -> %s", i, formatProxyURL(o.Prefix), formatProxyURL(n.Prefix)))
+			}
+			if strings.TrimSpace(o.APIKey) != strings.TrimSpace(n.APIKey) {
+				changes = append(changes, fmt.Sprintf("codex[%d].api-key: updated", i))
+			}
+			if !equalStringMap(o.Headers, n.Headers) {
+				changes = append(changes, fmt.Sprintf("codex[%d].headers: updated", i))
+			}
+			oldExcluded := SummarizeExcludedModels(o.ExcludedModels)
+			newExcluded := SummarizeExcludedModels(n.ExcludedModels)
+			if oldExcluded.hash != newExcluded.hash {
+				changes = append(changes, fmt.Sprintf("codex[%d].excluded-models: updated (%d -> %d entries)", i, oldExcluded.count, newExcluded.count))
+			}
+		}
+	}
+
+	// AmpCode settings (redacted where needed)
+	oldAmpURL := strings.TrimSpace(oldCfg.AmpCode.UpstreamURL)
+	newAmpURL := strings.TrimSpace(newCfg.AmpCode.UpstreamURL)
+	if oldAmpURL != newAmpURL {
+		changes = append(changes, fmt.Sprintf("ampcode.upstream-url: %s -> %s", oldAmpURL, newAmpURL))
+	}
+	oldAmpKey := strings.TrimSpace(oldCfg.AmpCode.UpstreamAPIKey)
+	newAmpKey := strings.TrimSpace(newCfg.AmpCode.UpstreamAPIKey)
+	switch {
+	case oldAmpKey == "" && newAmpKey != "":
+		changes = append(changes, "ampcode.upstream-api-key: added")
+	case oldAmpKey != "" && newAmpKey == "":
+		changes = append(changes, "ampcode.upstream-api-key: removed")
+	case oldAmpKey != newAmpKey:
+		changes = append(changes, "ampcode.upstream-api-key: updated")
+	}
+	if oldCfg.AmpCode.RestrictManagementToLocalhost != newCfg.AmpCode.RestrictManagementToLocalhost {
+		changes = append(changes, fmt.Sprintf("ampcode.restrict-management-to-localhost: %t -> %t", oldCfg.AmpCode.RestrictManagementToLocalhost, newCfg.AmpCode.RestrictManagementToLocalhost))
+	}
+	oldMappings := SummarizeAmpModelMappings(oldCfg.AmpCode.ModelMappings)
+	newMappings := SummarizeAmpModelMappings(newCfg.AmpCode.ModelMappings)
+	if oldMappings.hash != newMappings.hash {
+		changes = append(changes, fmt.Sprintf("ampcode.model-mappings: updated (%d -> %d entries)", oldMappings.count, newMappings.count))
+	}
+	if oldCfg.AmpCode.ForceModelMappings != newCfg.AmpCode.ForceModelMappings {
+		changes = append(changes, fmt.Sprintf("ampcode.force-model-mappings: %t -> %t", oldCfg.AmpCode.ForceModelMappings, newCfg.AmpCode.ForceModelMappings))
+	}
+
+	if entries, _ := DiffOAuthExcludedModelChanges(oldCfg.OAuthExcludedModels, newCfg.OAuthExcludedModels); len(entries) > 0 {
+		changes = append(changes, entries...)
+	}
+
+	// Remote management (never print the key)
+	if oldCfg.RemoteManagement.AllowRemote != newCfg.RemoteManagement.AllowRemote {
+		changes = append(changes, fmt.Sprintf("remote-management.allow-remote: %t -> %t", oldCfg.RemoteManagement.AllowRemote, newCfg.RemoteManagement.AllowRemote))
+	}
+	if oldCfg.RemoteManagement.DisableControlPanel != newCfg.RemoteManagement.DisableControlPanel {
+		changes = append(changes, fmt.Sprintf("remote-management.disable-control-panel: %t -> %t", oldCfg.RemoteManagement.DisableControlPanel, newCfg.RemoteManagement.DisableControlPanel))
+	}
+	oldPanelRepo := strings.TrimSpace(oldCfg.RemoteManagement.PanelGitHubRepository)
+	newPanelRepo := strings.TrimSpace(newCfg.RemoteManagement.PanelGitHubRepository)
+	if oldPanelRepo != newPanelRepo {
+		changes = append(changes, fmt.Sprintf("remote-management.panel-github-repository: %s -> %s", oldPanelRepo, newPanelRepo))
+	}
+	if oldCfg.RemoteManagement.SecretKey != newCfg.RemoteManagement.SecretKey {
+		switch {
+		case oldCfg.RemoteManagement.SecretKey == "" && newCfg.RemoteManagement.SecretKey != "":
+			changes = append(changes, "remote-management.secret-key: created")
+		case oldCfg.RemoteManagement.SecretKey != "" && newCfg.RemoteManagement.SecretKey == "":
+			changes = append(changes, "remote-management.secret-key: deleted")
+		default:
+			changes = append(changes, "remote-management.secret-key: updated")
+		}
+	}
+
+	// OpenAI compatibility providers (summarized)
+	if compat := DiffOpenAICompatibility(oldCfg.OpenAICompatibility, newCfg.OpenAICompatibility); len(compat) > 0 {
+		changes = append(changes, "openai-compatibility:")
+		for _, c := range compat {
+			changes = append(changes, "  "+c)
+		}
+	}
+
+	// Vertex-compatible API keys
+	if len(oldCfg.VertexCompatAPIKey) != len(newCfg.VertexCompatAPIKey) {
+		changes = append(changes, fmt.Sprintf("vertex-api-key count: %d -> %d", len(oldCfg.VertexCompatAPIKey), len(newCfg.VertexCompatAPIKey)))
+	} else {
+		for i := range oldCfg.VertexCompatAPIKey {
+			o := oldCfg.VertexCompatAPIKey[i]
+			n := newCfg.VertexCompatAPIKey[i]
+			if strings.TrimSpace(o.BaseURL) != strings.TrimSpace(n.BaseURL) {
+				changes = append(changes, fmt.Sprintf("vertex[%d].base-url: %s -> %s", i, strings.TrimSpace(o.BaseURL), strings.TrimSpace(n.BaseURL)))
+			}
+			if strings.TrimSpace(o.ProxyURL) != strings.TrimSpace(n.ProxyURL) {
+				changes = append(changes, fmt.Sprintf("vertex[%d].proxy-url: %s -> %s", i, formatProxyURL(o.ProxyURL), formatProxyURL(n.ProxyURL)))
+			}
+			if strings.TrimSpace(o.Prefix) != strings.TrimSpace(n.Prefix) {
+				changes = append(changes, fmt.Sprintf("vertex[%d].prefix: %s -> %s", i, formatProxyURL(o.Prefix), formatProxyURL(n.Prefix)))
+			}
+			if strings.TrimSpace(o.APIKey) != strings.TrimSpace(n.APIKey) {
+				changes = append(changes, fmt.Sprintf("vertex[%d].api-key: updated", i))
+			}
+			oldModels := SummarizeVertexModels(o.Models)
+			newModels := SummarizeVertexModels(n.Models)
+			if oldModels.hash != newModels.hash {
+				changes = append(changes, fmt.Sprintf("vertex[%d].models: updated (%d -> %d entries)", i, oldModels.count, newModels.count))
+			}
+			if !equalStringMap(o.Headers, n.Headers) {
+				changes = append(changes, fmt.Sprintf("vertex[%d].headers: updated", i))
+			}
+		}
+	}
+
+	return changes
+}
+
+func trimStrings(in []string) []string {
+	out := make([]string, len(in))
+	for i := range in {
+		out[i] = strings.TrimSpace(in[i])
+	}
+	return out
+}
+
+func equalStringMap(a, b map[string]string) bool {
+	if len(a) != len(b) {
+		return false
+	}
+	for k, v := range a {
+		if b[k] != v {
+			return false
+		}
+	}
+	return true
+}
+
+func formatProxyURL(raw string) string {
+	trimmed := strings.TrimSpace(raw)
+	if trimmed == "" {
+		return "<none>"
+	}
+	parsed, err := url.Parse(trimmed)
+	if err != nil {
+		return "<redacted>"
+	}
+	host := strings.TrimSpace(parsed.Host)
+	scheme := strings.TrimSpace(parsed.Scheme)
+	if host == "" {
+		// Allow host:port style without scheme.
+		parsed2, err2 := url.Parse("http://" + trimmed)
+		if err2 == nil {
+			host = strings.TrimSpace(parsed2.Host)
+		}
+		scheme = ""
+	}
+	if host == "" {
+		return "<redacted>"
+	}
+	if scheme == "" {
+		return host
+	}
+	return scheme + "://" + host
+}

internal/watcher/diff/config_diff_test.go

@@ -0,0 +1,526 @@
+package diff
+
+import (
+	"testing"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+	sdkconfig "github.com/router-for-me/CLIProxyAPI/v6/sdk/config"
+)
+
+func TestBuildConfigChangeDetails(t *testing.T) {
+	oldCfg := &config.Config{
+		Port:    8080,
+		AuthDir: "/tmp/auth-old",
+		GeminiKey: []config.GeminiKey{
+			{APIKey: "old", BaseURL: "http://old", ExcludedModels: []string{"old-model"}},
+		},
+		AmpCode: config.AmpCode{
+			UpstreamURL:                   "http://old-upstream",
+			ModelMappings:                 []config.AmpModelMapping{{From: "from-old", To: "to-old"}},
+			RestrictManagementToLocalhost: false,
+		},
+		RemoteManagement: config.RemoteManagement{
+			AllowRemote:           false,
+			SecretKey:             "old",
+			DisableControlPanel:   false,
+			PanelGitHubRepository: "repo-old",
+		},
+		OAuthExcludedModels: map[string][]string{
+			"providerA": {"m1"},
+		},
+		OpenAICompatibility: []config.OpenAICompatibility{
+			{
+				Name: "compat-a",
+				APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+					{APIKey: "k1"},
+				},
+				Models: []config.OpenAICompatibilityModel{{Name: "m1"}},
+			},
+		},
+	}
+
+	newCfg := &config.Config{
+		Port:    9090,
+		AuthDir: "/tmp/auth-new",
+		GeminiKey: []config.GeminiKey{
+			{APIKey: "old", BaseURL: "http://old", ExcludedModels: []string{"old-model", "extra"}},
+		},
+		AmpCode: config.AmpCode{
+			UpstreamURL:                   "http://new-upstream",
+			RestrictManagementToLocalhost: true,
+			ModelMappings: []config.AmpModelMapping{
+				{From: "from-old", To: "to-old"},
+				{From: "from-new", To: "to-new"},
+			},
+		},
+		RemoteManagement: config.RemoteManagement{
+			AllowRemote:           true,
+			SecretKey:             "new",
+			DisableControlPanel:   true,
+			PanelGitHubRepository: "repo-new",
+		},
+		OAuthExcludedModels: map[string][]string{
+			"providerA": {"m1", "m2"},
+			"providerB": {"x"},
+		},
+		OpenAICompatibility: []config.OpenAICompatibility{
+			{
+				Name: "compat-a",
+				APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+					{APIKey: "k1"},
+				},
+				Models: []config.OpenAICompatibilityModel{{Name: "m1"}, {Name: "m2"}},
+			},
+			{
+				Name: "compat-b",
+				APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+					{APIKey: "k2"},
+				},
+			},
+		},
+	}
+
+	details := BuildConfigChangeDetails(oldCfg, newCfg)
+
+	expectContains(t, details, "port: 8080 -> 9090")
+	expectContains(t, details, "auth-dir: /tmp/auth-old -> /tmp/auth-new")
+	expectContains(t, details, "gemini[0].excluded-models: updated (1 -> 2 entries)")
+	expectContains(t, details, "ampcode.upstream-url: http://old-upstream -> http://new-upstream")
+	expectContains(t, details, "ampcode.model-mappings: updated (1 -> 2 entries)")
+	expectContains(t, details, "remote-management.allow-remote: false -> true")
+	expectContains(t, details, "remote-management.secret-key: updated")
+	expectContains(t, details, "oauth-excluded-models[providera]: updated (1 -> 2 entries)")
+	expectContains(t, details, "oauth-excluded-models[providerb]: added (1 entries)")
+	expectContains(t, details, "openai-compatibility:")
+	expectContains(t, details, "  provider added: compat-b (api-keys=1, models=0)")
+	expectContains(t, details, "  provider updated: compat-a (models 1 -> 2)")
+}
+
+func TestBuildConfigChangeDetails_NoChanges(t *testing.T) {
+	cfg := &config.Config{
+		Port: 8080,
+	}
+	if details := BuildConfigChangeDetails(cfg, cfg); len(details) != 0 {
+		t.Fatalf("expected no change entries, got %v", details)
+	}
+}
+
+func TestBuildConfigChangeDetails_GeminiVertexHeadersAndForceMappings(t *testing.T) {
+	oldCfg := &config.Config{
+		GeminiKey: []config.GeminiKey{
+			{APIKey: "g1", Headers: map[string]string{"H": "1"}, ExcludedModels: []string{"a"}},
+		},
+		VertexCompatAPIKey: []config.VertexCompatKey{
+			{APIKey: "v1", BaseURL: "http://v-old", Models: []config.VertexCompatModel{{Name: "m1"}}},
+		},
+		AmpCode: config.AmpCode{
+			ModelMappings:      []config.AmpModelMapping{{From: "a", To: "b"}},
+			ForceModelMappings: false,
+		},
+	}
+	newCfg := &config.Config{
+		GeminiKey: []config.GeminiKey{
+			{APIKey: "g1", Headers: map[string]string{"H": "2"}, ExcludedModels: []string{"a", "b"}},
+		},
+		VertexCompatAPIKey: []config.VertexCompatKey{
+			{APIKey: "v1", BaseURL: "http://v-new", Models: []config.VertexCompatModel{{Name: "m1"}, {Name: "m2"}}},
+		},
+		AmpCode: config.AmpCode{
+			ModelMappings:      []config.AmpModelMapping{{From: "a", To: "c"}},
+			ForceModelMappings: true,
+		},
+	}
+
+	details := BuildConfigChangeDetails(oldCfg, newCfg)
+	expectContains(t, details, "gemini[0].headers: updated")
+	expectContains(t, details, "gemini[0].excluded-models: updated (1 -> 2 entries)")
+	expectContains(t, details, "ampcode.model-mappings: updated (1 -> 1 entries)")
+	expectContains(t, details, "ampcode.force-model-mappings: false -> true")
+}
+
+func TestBuildConfigChangeDetails_ModelPrefixes(t *testing.T) {
+	oldCfg := &config.Config{
+		GeminiKey: []config.GeminiKey{
+			{APIKey: "g1", Prefix: "old-g", BaseURL: "http://g", ProxyURL: "http://gp"},
+		},
+		ClaudeKey: []config.ClaudeKey{
+			{APIKey: "c1", Prefix: "old-c", BaseURL: "http://c", ProxyURL: "http://cp"},
+		},
+		CodexKey: []config.CodexKey{
+			{APIKey: "x1", Prefix: "old-x", BaseURL: "http://x", ProxyURL: "http://xp"},
+		},
+		VertexCompatAPIKey: []config.VertexCompatKey{
+			{APIKey: "v1", Prefix: "old-v", BaseURL: "http://v", ProxyURL: "http://vp"},
+		},
+	}
+	newCfg := &config.Config{
+		GeminiKey: []config.GeminiKey{
+			{APIKey: "g1", Prefix: "new-g", BaseURL: "http://g", ProxyURL: "http://gp"},
+		},
+		ClaudeKey: []config.ClaudeKey{
+			{APIKey: "c1", Prefix: "new-c", BaseURL: "http://c", ProxyURL: "http://cp"},
+		},
+		CodexKey: []config.CodexKey{
+			{APIKey: "x1", Prefix: "new-x", BaseURL: "http://x", ProxyURL: "http://xp"},
+		},
+		VertexCompatAPIKey: []config.VertexCompatKey{
+			{APIKey: "v1", Prefix: "new-v", BaseURL: "http://v", ProxyURL: "http://vp"},
+		},
+	}
+
+	changes := BuildConfigChangeDetails(oldCfg, newCfg)
+	expectContains(t, changes, "gemini[0].prefix: old-g -> new-g")
+	expectContains(t, changes, "claude[0].prefix: old-c -> new-c")
+	expectContains(t, changes, "codex[0].prefix: old-x -> new-x")
+	expectContains(t, changes, "vertex[0].prefix: old-v -> new-v")
+}
+
+func TestBuildConfigChangeDetails_NilSafe(t *testing.T) {
+	if details := BuildConfigChangeDetails(nil, &config.Config{}); len(details) != 0 {
+		t.Fatalf("expected empty change list when old nil, got %v", details)
+	}
+	if details := BuildConfigChangeDetails(&config.Config{}, nil); len(details) != 0 {
+		t.Fatalf("expected empty change list when new nil, got %v", details)
+	}
+}
+
+func TestBuildConfigChangeDetails_SecretsAndCounts(t *testing.T) {
+	oldCfg := &config.Config{
+		SDKConfig: sdkconfig.SDKConfig{
+			APIKeys: []string{"a"},
+		},
+		AmpCode: config.AmpCode{
+			UpstreamAPIKey: "",
+		},
+		RemoteManagement: config.RemoteManagement{
+			SecretKey: "",
+		},
+	}
+	newCfg := &config.Config{
+		SDKConfig: sdkconfig.SDKConfig{
+			APIKeys: []string{"a", "b", "c"},
+		},
+		AmpCode: config.AmpCode{
+			UpstreamAPIKey: "new-key",
+		},
+		RemoteManagement: config.RemoteManagement{
+			SecretKey: "new-secret",
+		},
+	}
+
+	details := BuildConfigChangeDetails(oldCfg, newCfg)
+	expectContains(t, details, "api-keys count: 1 -> 3")
+	expectContains(t, details, "ampcode.upstream-api-key: added")
+	expectContains(t, details, "remote-management.secret-key: created")
+}
+
+func TestBuildConfigChangeDetails_FlagsAndKeys(t *testing.T) {
+	oldCfg := &config.Config{
+		Port:                   1000,
+		AuthDir:                "/old",
+		Debug:                  false,
+		LoggingToFile:          false,
+		UsageStatisticsEnabled: false,
+		DisableCooling:         false,
+		RequestRetry:           1,
+		MaxRetryInterval:       1,
+		WebsocketAuth:          false,
+		QuotaExceeded:          config.QuotaExceeded{SwitchProject: false, SwitchPreviewModel: false},
+		ClaudeKey:              []config.ClaudeKey{{APIKey: "c1"}},
+		CodexKey:               []config.CodexKey{{APIKey: "x1"}},
+		AmpCode:                config.AmpCode{UpstreamAPIKey: "keep", RestrictManagementToLocalhost: false},
+		RemoteManagement:       config.RemoteManagement{DisableControlPanel: false, PanelGitHubRepository: "old/repo", SecretKey: "keep"},
+		SDKConfig: sdkconfig.SDKConfig{
+			RequestLog: false,
+			ProxyURL:   "http://old-proxy",
+			APIKeys:    []string{"key-1"},
+		},
+	}
+	newCfg := &config.Config{
+		Port:                   2000,
+		AuthDir:                "/new",
+		Debug:                  true,
+		LoggingToFile:          true,
+		UsageStatisticsEnabled: true,
+		DisableCooling:         true,
+		RequestRetry:           2,
+		MaxRetryInterval:       3,
+		WebsocketAuth:          true,
+		QuotaExceeded:          config.QuotaExceeded{SwitchProject: true, SwitchPreviewModel: true},
+		ClaudeKey: []config.ClaudeKey{
+			{APIKey: "c1", BaseURL: "http://new", ProxyURL: "http://p", Headers: map[string]string{"H": "1"}, ExcludedModels: []string{"a"}},
+			{APIKey: "c2"},
+		},
+		CodexKey: []config.CodexKey{
+			{APIKey: "x1", BaseURL: "http://x", ProxyURL: "http://px", Headers: map[string]string{"H": "2"}, ExcludedModels: []string{"b"}},
+			{APIKey: "x2"},
+		},
+		AmpCode: config.AmpCode{
+			UpstreamAPIKey:                "",
+			RestrictManagementToLocalhost: true,
+			ModelMappings:                 []config.AmpModelMapping{{From: "a", To: "b"}},
+		},
+		RemoteManagement: config.RemoteManagement{
+			DisableControlPanel:   true,
+			PanelGitHubRepository: "new/repo",
+			SecretKey:             "",
+		},
+		SDKConfig: sdkconfig.SDKConfig{
+			RequestLog: true,
+			ProxyURL:   "http://new-proxy",
+			APIKeys:    []string{" key-1 ", "key-2"},
+		},
+	}
+
+	details := BuildConfigChangeDetails(oldCfg, newCfg)
+	expectContains(t, details, "debug: false -> true")
+	expectContains(t, details, "logging-to-file: false -> true")
+	expectContains(t, details, "usage-statistics-enabled: false -> true")
+	expectContains(t, details, "disable-cooling: false -> true")
+	expectContains(t, details, "request-log: false -> true")
+	expectContains(t, details, "request-retry: 1 -> 2")
+	expectContains(t, details, "max-retry-interval: 1 -> 3")
+	expectContains(t, details, "proxy-url: http://old-proxy -> http://new-proxy")
+	expectContains(t, details, "ws-auth: false -> true")
+	expectContains(t, details, "quota-exceeded.switch-project: false -> true")
+	expectContains(t, details, "quota-exceeded.switch-preview-model: false -> true")
+	expectContains(t, details, "api-keys count: 1 -> 2")
+	expectContains(t, details, "claude-api-key count: 1 -> 2")
+	expectContains(t, details, "codex-api-key count: 1 -> 2")
+	expectContains(t, details, "ampcode.restrict-management-to-localhost: false -> true")
+	expectContains(t, details, "ampcode.upstream-api-key: removed")
+	expectContains(t, details, "remote-management.disable-control-panel: false -> true")
+	expectContains(t, details, "remote-management.panel-github-repository: old/repo -> new/repo")
+	expectContains(t, details, "remote-management.secret-key: deleted")
+}
+
+func TestBuildConfigChangeDetails_AllBranches(t *testing.T) {
+	oldCfg := &config.Config{
+		Port:                   1,
+		AuthDir:                "/a",
+		Debug:                  false,
+		LoggingToFile:          false,
+		UsageStatisticsEnabled: false,
+		DisableCooling:         false,
+		RequestRetry:           1,
+		MaxRetryInterval:       1,
+		WebsocketAuth:          false,
+		QuotaExceeded:          config.QuotaExceeded{SwitchProject: false, SwitchPreviewModel: false},
+		GeminiKey: []config.GeminiKey{
+			{APIKey: "g-old", BaseURL: "http://g-old", ProxyURL: "http://gp-old", Headers: map[string]string{"A": "1"}},
+		},
+		ClaudeKey: []config.ClaudeKey{
+			{APIKey: "c-old", BaseURL: "http://c-old", ProxyURL: "http://cp-old", Headers: map[string]string{"H": "1"}, ExcludedModels: []string{"x"}},
+		},
+		CodexKey: []config.CodexKey{
+			{APIKey: "x-old", BaseURL: "http://x-old", ProxyURL: "http://xp-old", Headers: map[string]string{"H": "1"}, ExcludedModels: []string{"x"}},
+		},
+		VertexCompatAPIKey: []config.VertexCompatKey{
+			{APIKey: "v-old", BaseURL: "http://v-old", ProxyURL: "http://vp-old", Headers: map[string]string{"H": "1"}, Models: []config.VertexCompatModel{{Name: "m1"}}},
+		},
+		AmpCode: config.AmpCode{
+			UpstreamURL:                   "http://amp-old",
+			UpstreamAPIKey:                "old-key",
+			RestrictManagementToLocalhost: false,
+			ModelMappings:                 []config.AmpModelMapping{{From: "a", To: "b"}},
+			ForceModelMappings:            false,
+		},
+		RemoteManagement: config.RemoteManagement{
+			AllowRemote:           false,
+			DisableControlPanel:   false,
+			PanelGitHubRepository: "old/repo",
+			SecretKey:             "old",
+		},
+		SDKConfig: sdkconfig.SDKConfig{
+			RequestLog: false,
+			ProxyURL:   "http://old-proxy",
+			APIKeys:    []string{" keyA "},
+		},
+		OAuthExcludedModels: map[string][]string{"p1": {"a"}},
+		OpenAICompatibility: []config.OpenAICompatibility{
+			{
+				Name: "prov-old",
+				APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+					{APIKey: "k1"},
+				},
+				Models: []config.OpenAICompatibilityModel{{Name: "m1"}},
+			},
+		},
+	}
+	newCfg := &config.Config{
+		Port:                   2,
+		AuthDir:                "/b",
+		Debug:                  true,
+		LoggingToFile:          true,
+		UsageStatisticsEnabled: true,
+		DisableCooling:         true,
+		RequestRetry:           2,
+		MaxRetryInterval:       3,
+		WebsocketAuth:          true,
+		QuotaExceeded:          config.QuotaExceeded{SwitchProject: true, SwitchPreviewModel: true},
+		GeminiKey: []config.GeminiKey{
+			{APIKey: "g-new", BaseURL: "http://g-new", ProxyURL: "http://gp-new", Headers: map[string]string{"A": "2"}, ExcludedModels: []string{"x", "y"}},
+		},
+		ClaudeKey: []config.ClaudeKey{
+			{APIKey: "c-new", BaseURL: "http://c-new", ProxyURL: "http://cp-new", Headers: map[string]string{"H": "2"}, ExcludedModels: []string{"x", "y"}},
+		},
+		CodexKey: []config.CodexKey{
+			{APIKey: "x-new", BaseURL: "http://x-new", ProxyURL: "http://xp-new", Headers: map[string]string{"H": "2"}, ExcludedModels: []string{"x", "y"}},
+		},
+		VertexCompatAPIKey: []config.VertexCompatKey{
+			{APIKey: "v-new", BaseURL: "http://v-new", ProxyURL: "http://vp-new", Headers: map[string]string{"H": "2"}, Models: []config.VertexCompatModel{{Name: "m1"}, {Name: "m2"}}},
+		},
+		AmpCode: config.AmpCode{
+			UpstreamURL:                   "http://amp-new",
+			UpstreamAPIKey:                "",
+			RestrictManagementToLocalhost: true,
+			ModelMappings:                 []config.AmpModelMapping{{From: "a", To: "c"}},
+			ForceModelMappings:            true,
+		},
+		RemoteManagement: config.RemoteManagement{
+			AllowRemote:           true,
+			DisableControlPanel:   true,
+			PanelGitHubRepository: "new/repo",
+			SecretKey:             "",
+		},
+		SDKConfig: sdkconfig.SDKConfig{
+			RequestLog: true,
+			ProxyURL:   "http://new-proxy",
+			APIKeys:    []string{"keyB"},
+		},
+		OAuthExcludedModels: map[string][]string{"p1": {"b", "c"}, "p2": {"d"}},
+		OpenAICompatibility: []config.OpenAICompatibility{
+			{
+				Name: "prov-old",
+				APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+					{APIKey: "k1"},
+					{APIKey: "k2"},
+				},
+				Models: []config.OpenAICompatibilityModel{{Name: "m1"}, {Name: "m2"}},
+			},
+			{
+				Name:          "prov-new",
+				APIKeyEntries: []config.OpenAICompatibilityAPIKey{{APIKey: "k3"}},
+			},
+		},
+	}
+
+	changes := BuildConfigChangeDetails(oldCfg, newCfg)
+	expectContains(t, changes, "port: 1 -> 2")
+	expectContains(t, changes, "auth-dir: /a -> /b")
+	expectContains(t, changes, "debug: false -> true")
+	expectContains(t, changes, "logging-to-file: false -> true")
+	expectContains(t, changes, "usage-statistics-enabled: false -> true")
+	expectContains(t, changes, "disable-cooling: false -> true")
+	expectContains(t, changes, "request-retry: 1 -> 2")
+	expectContains(t, changes, "max-retry-interval: 1 -> 3")
+	expectContains(t, changes, "proxy-url: http://old-proxy -> http://new-proxy")
+	expectContains(t, changes, "ws-auth: false -> true")
+	expectContains(t, changes, "quota-exceeded.switch-project: false -> true")
+	expectContains(t, changes, "quota-exceeded.switch-preview-model: false -> true")
+	expectContains(t, changes, "api-keys: values updated (count unchanged, redacted)")
+	expectContains(t, changes, "gemini[0].base-url: http://g-old -> http://g-new")
+	expectContains(t, changes, "gemini[0].proxy-url: http://gp-old -> http://gp-new")
+	expectContains(t, changes, "gemini[0].api-key: updated")
+	expectContains(t, changes, "gemini[0].headers: updated")
+	expectContains(t, changes, "gemini[0].excluded-models: updated (0 -> 2 entries)")
+	expectContains(t, changes, "claude[0].base-url: http://c-old -> http://c-new")
+	expectContains(t, changes, "claude[0].proxy-url: http://cp-old -> http://cp-new")
+	expectContains(t, changes, "claude[0].api-key: updated")
+	expectContains(t, changes, "claude[0].headers: updated")
+	expectContains(t, changes, "claude[0].excluded-models: updated (1 -> 2 entries)")
+	expectContains(t, changes, "codex[0].base-url: http://x-old -> http://x-new")
+	expectContains(t, changes, "codex[0].proxy-url: http://xp-old -> http://xp-new")
+	expectContains(t, changes, "codex[0].api-key: updated")
+	expectContains(t, changes, "codex[0].headers: updated")
+	expectContains(t, changes, "codex[0].excluded-models: updated (1 -> 2 entries)")
+	expectContains(t, changes, "vertex[0].base-url: http://v-old -> http://v-new")
+	expectContains(t, changes, "vertex[0].proxy-url: http://vp-old -> http://vp-new")
+	expectContains(t, changes, "vertex[0].api-key: updated")
+	expectContains(t, changes, "vertex[0].models: updated (1 -> 2 entries)")
+	expectContains(t, changes, "vertex[0].headers: updated")
+	expectContains(t, changes, "ampcode.upstream-url: http://amp-old -> http://amp-new")
+	expectContains(t, changes, "ampcode.upstream-api-key: removed")
+	expectContains(t, changes, "ampcode.restrict-management-to-localhost: false -> true")
+	expectContains(t, changes, "ampcode.model-mappings: updated (1 -> 1 entries)")
+	expectContains(t, changes, "ampcode.force-model-mappings: false -> true")
+	expectContains(t, changes, "oauth-excluded-models[p1]: updated (1 -> 2 entries)")
+	expectContains(t, changes, "oauth-excluded-models[p2]: added (1 entries)")
+	expectContains(t, changes, "remote-management.allow-remote: false -> true")
+	expectContains(t, changes, "remote-management.disable-control-panel: false -> true")
+	expectContains(t, changes, "remote-management.panel-github-repository: old/repo -> new/repo")
+	expectContains(t, changes, "remote-management.secret-key: deleted")
+	expectContains(t, changes, "openai-compatibility:")
+}
+
+func TestFormatProxyURL(t *testing.T) {
+	tests := []struct {
+		name string
+		in   string
+		want string
+	}{
+		{name: "empty", in: "", want: "<none>"},
+		{name: "invalid", in: "http://[::1", want: "<redacted>"},
+		{name: "fullURLRedactsUserinfoAndPath", in: "http://user:pass@example.com:8080/path?x=1#frag", want: "http://example.com:8080"},
+		{name: "socks5RedactsUserinfoAndPath", in: "socks5://user:pass@192.168.1.1:1080/path?x=1", want: "socks5://192.168.1.1:1080"},
+		{name: "socks5HostPort", in: "socks5://proxy.example.com:1080/", want: "socks5://proxy.example.com:1080"},
+		{name: "hostPortNoScheme", in: "example.com:1234/path?x=1", want: "example.com:1234"},
+		{name: "relativePathRedacted", in: "/just/path", want: "<redacted>"},
+		{name: "schemeAndHost", in: "https://example.com", want: "https://example.com"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if got := formatProxyURL(tt.in); got != tt.want {
+				t.Fatalf("expected %q, got %q", tt.want, got)
+			}
+		})
+	}
+}
+
+func TestBuildConfigChangeDetails_SecretAndUpstreamUpdates(t *testing.T) {
+	oldCfg := &config.Config{
+		AmpCode: config.AmpCode{
+			UpstreamAPIKey: "old",
+		},
+		RemoteManagement: config.RemoteManagement{
+			SecretKey: "old",
+		},
+	}
+	newCfg := &config.Config{
+		AmpCode: config.AmpCode{
+			UpstreamAPIKey: "new",
+		},
+		RemoteManagement: config.RemoteManagement{
+			SecretKey: "new",
+		},
+	}
+
+	changes := BuildConfigChangeDetails(oldCfg, newCfg)
+	expectContains(t, changes, "ampcode.upstream-api-key: updated")
+	expectContains(t, changes, "remote-management.secret-key: updated")
+}
+
+func TestBuildConfigChangeDetails_CountBranches(t *testing.T) {
+	oldCfg := &config.Config{}
+	newCfg := &config.Config{
+		GeminiKey: []config.GeminiKey{{APIKey: "g"}},
+		ClaudeKey: []config.ClaudeKey{{APIKey: "c"}},
+		CodexKey:  []config.CodexKey{{APIKey: "x"}},
+		VertexCompatAPIKey: []config.VertexCompatKey{
+			{APIKey: "v", BaseURL: "http://v"},
+		},
+	}
+
+	changes := BuildConfigChangeDetails(oldCfg, newCfg)
+	expectContains(t, changes, "gemini-api-key count: 0 -> 1")
+	expectContains(t, changes, "claude-api-key count: 0 -> 1")
+	expectContains(t, changes, "codex-api-key count: 0 -> 1")
+	expectContains(t, changes, "vertex-api-key count: 0 -> 1")
+}
+
+func TestTrimStrings(t *testing.T) {
+	out := trimStrings([]string{" a ", "b", "  c"})
+	if len(out) != 3 || out[0] != "a" || out[1] != "b" || out[2] != "c" {
+		t.Fatalf("unexpected trimmed strings: %v", out)
+	}
+}

internal/watcher/diff/model_hash.go

@@ -0,0 +1,102 @@
+package diff
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"encoding/json"
+	"sort"
+	"strings"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+)
+
+// ComputeOpenAICompatModelsHash returns a stable hash for OpenAI-compat models.
+// Used to detect model list changes during hot reload.
+func ComputeOpenAICompatModelsHash(models []config.OpenAICompatibilityModel) string {
+	keys := normalizeModelPairs(func(out func(key string)) {
+		for _, model := range models {
+			name := strings.TrimSpace(model.Name)
+			alias := strings.TrimSpace(model.Alias)
+			if name == "" && alias == "" {
+				continue
+			}
+			out(strings.ToLower(name) + "|" + strings.ToLower(alias))
+		}
+	})
+	return hashJoined(keys)
+}
+
+// ComputeVertexCompatModelsHash returns a stable hash for Vertex-compatible models.
+func ComputeVertexCompatModelsHash(models []config.VertexCompatModel) string {
+	keys := normalizeModelPairs(func(out func(key string)) {
+		for _, model := range models {
+			name := strings.TrimSpace(model.Name)
+			alias := strings.TrimSpace(model.Alias)
+			if name == "" && alias == "" {
+				continue
+			}
+			out(strings.ToLower(name) + "|" + strings.ToLower(alias))
+		}
+	})
+	return hashJoined(keys)
+}
+
+// ComputeClaudeModelsHash returns a stable hash for Claude model aliases.
+func ComputeClaudeModelsHash(models []config.ClaudeModel) string {
+	keys := normalizeModelPairs(func(out func(key string)) {
+		for _, model := range models {
+			name := strings.TrimSpace(model.Name)
+			alias := strings.TrimSpace(model.Alias)
+			if name == "" && alias == "" {
+				continue
+			}
+			out(strings.ToLower(name) + "|" + strings.ToLower(alias))
+		}
+	})
+	return hashJoined(keys)
+}
+
+// ComputeExcludedModelsHash returns a normalized hash for excluded model lists.
+func ComputeExcludedModelsHash(excluded []string) string {
+	if len(excluded) == 0 {
+		return ""
+	}
+	normalized := make([]string, 0, len(excluded))
+	for _, entry := range excluded {
+		if trimmed := strings.TrimSpace(entry); trimmed != "" {
+			normalized = append(normalized, strings.ToLower(trimmed))
+		}
+	}
+	if len(normalized) == 0 {
+		return ""
+	}
+	sort.Strings(normalized)
+	data, _ := json.Marshal(normalized)
+	sum := sha256.Sum256(data)
+	return hex.EncodeToString(sum[:])
+}
+
+func normalizeModelPairs(collect func(out func(key string))) []string {
+	seen := make(map[string]struct{})
+	keys := make([]string, 0)
+	collect(func(key string) {
+		if _, exists := seen[key]; exists {
+			return
+		}
+		seen[key] = struct{}{}
+		keys = append(keys, key)
+	})
+	if len(keys) == 0 {
+		return nil
+	}
+	sort.Strings(keys)
+	return keys
+}
+
+func hashJoined(keys []string) string {
+	if len(keys) == 0 {
+		return ""
+	}
+	sum := sha256.Sum256([]byte(strings.Join(keys, "\n")))
+	return hex.EncodeToString(sum[:])
+}

internal/watcher/diff/model_hash_test.go

@@ -0,0 +1,159 @@
+package diff
+
+import (
+	"testing"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+)
+
+func TestComputeOpenAICompatModelsHash_Deterministic(t *testing.T) {
+	models := []config.OpenAICompatibilityModel{
+		{Name: "gpt-4", Alias: "gpt4"},
+		{Name: "gpt-3.5-turbo"},
+	}
+	hash1 := ComputeOpenAICompatModelsHash(models)
+	hash2 := ComputeOpenAICompatModelsHash(models)
+	if hash1 == "" {
+		t.Fatal("hash should not be empty")
+	}
+	if hash1 != hash2 {
+		t.Fatalf("hash should be deterministic, got %s vs %s", hash1, hash2)
+	}
+	changed := ComputeOpenAICompatModelsHash([]config.OpenAICompatibilityModel{{Name: "gpt-4"}, {Name: "gpt-4.1"}})
+	if hash1 == changed {
+		t.Fatal("hash should change when model list changes")
+	}
+}
+
+func TestComputeOpenAICompatModelsHash_NormalizesAndDedups(t *testing.T) {
+	a := []config.OpenAICompatibilityModel{
+		{Name: "gpt-4", Alias: "gpt4"},
+		{Name: " "},
+		{Name: "GPT-4", Alias: "GPT4"},
+		{Alias: "a1"},
+	}
+	b := []config.OpenAICompatibilityModel{
+		{Alias: "A1"},
+		{Name: "gpt-4", Alias: "gpt4"},
+	}
+	h1 := ComputeOpenAICompatModelsHash(a)
+	h2 := ComputeOpenAICompatModelsHash(b)
+	if h1 == "" || h2 == "" {
+		t.Fatal("expected non-empty hashes for non-empty model sets")
+	}
+	if h1 != h2 {
+		t.Fatalf("expected normalized hashes to match, got %s / %s", h1, h2)
+	}
+}
+
+func TestComputeVertexCompatModelsHash_DifferentInputs(t *testing.T) {
+	models := []config.VertexCompatModel{{Name: "gemini-pro", Alias: "pro"}}
+	hash1 := ComputeVertexCompatModelsHash(models)
+	hash2 := ComputeVertexCompatModelsHash([]config.VertexCompatModel{{Name: "gemini-1.5-pro", Alias: "pro"}})
+	if hash1 == "" || hash2 == "" {
+		t.Fatal("hashes should not be empty for non-empty models")
+	}
+	if hash1 == hash2 {
+		t.Fatal("hash should differ when model content differs")
+	}
+}
+
+func TestComputeVertexCompatModelsHash_IgnoresBlankAndOrder(t *testing.T) {
+	a := []config.VertexCompatModel{
+		{Name: "m1", Alias: "a1"},
+		{Name: " "},
+		{Name: "M1", Alias: "A1"},
+	}
+	b := []config.VertexCompatModel{
+		{Name: "m1", Alias: "a1"},
+	}
+	if h1, h2 := ComputeVertexCompatModelsHash(a), ComputeVertexCompatModelsHash(b); h1 == "" || h1 != h2 {
+		t.Fatalf("expected same hash ignoring blanks/dupes, got %q / %q", h1, h2)
+	}
+}
+
+func TestComputeClaudeModelsHash_Empty(t *testing.T) {
+	if got := ComputeClaudeModelsHash(nil); got != "" {
+		t.Fatalf("expected empty hash for nil models, got %q", got)
+	}
+	if got := ComputeClaudeModelsHash([]config.ClaudeModel{}); got != "" {
+		t.Fatalf("expected empty hash for empty slice, got %q", got)
+	}
+}
+
+func TestComputeClaudeModelsHash_IgnoresBlankAndDedup(t *testing.T) {
+	a := []config.ClaudeModel{
+		{Name: "m1", Alias: "a1"},
+		{Name: " "},
+		{Name: "M1", Alias: "A1"},
+	}
+	b := []config.ClaudeModel{
+		{Name: "m1", Alias: "a1"},
+	}
+	if h1, h2 := ComputeClaudeModelsHash(a), ComputeClaudeModelsHash(b); h1 == "" || h1 != h2 {
+		t.Fatalf("expected same hash ignoring blanks/dupes, got %q / %q", h1, h2)
+	}
+}
+
+func TestComputeExcludedModelsHash_Normalizes(t *testing.T) {
+	hash1 := ComputeExcludedModelsHash([]string{" A ", "b", "a"})
+	hash2 := ComputeExcludedModelsHash([]string{"a", " b", "A"})
+	if hash1 == "" || hash2 == "" {
+		t.Fatal("hash should not be empty for non-empty input")
+	}
+	if hash1 != hash2 {
+		t.Fatalf("hash should be order/space insensitive for same multiset, got %s vs %s", hash1, hash2)
+	}
+	hash3 := ComputeExcludedModelsHash([]string{"c"})
+	if hash1 == hash3 {
+		t.Fatal("hash should differ for different normalized sets")
+	}
+}
+
+func TestComputeOpenAICompatModelsHash_Empty(t *testing.T) {
+	if got := ComputeOpenAICompatModelsHash(nil); got != "" {
+		t.Fatalf("expected empty hash for nil input, got %q", got)
+	}
+	if got := ComputeOpenAICompatModelsHash([]config.OpenAICompatibilityModel{}); got != "" {
+		t.Fatalf("expected empty hash for empty slice, got %q", got)
+	}
+	if got := ComputeOpenAICompatModelsHash([]config.OpenAICompatibilityModel{{Name: " "}, {Alias: ""}}); got != "" {
+		t.Fatalf("expected empty hash for blank models, got %q", got)
+	}
+}
+
+func TestComputeVertexCompatModelsHash_Empty(t *testing.T) {
+	if got := ComputeVertexCompatModelsHash(nil); got != "" {
+		t.Fatalf("expected empty hash for nil input, got %q", got)
+	}
+	if got := ComputeVertexCompatModelsHash([]config.VertexCompatModel{}); got != "" {
+		t.Fatalf("expected empty hash for empty slice, got %q", got)
+	}
+	if got := ComputeVertexCompatModelsHash([]config.VertexCompatModel{{Name: " "}}); got != "" {
+		t.Fatalf("expected empty hash for blank models, got %q", got)
+	}
+}
+
+func TestComputeExcludedModelsHash_Empty(t *testing.T) {
+	if got := ComputeExcludedModelsHash(nil); got != "" {
+		t.Fatalf("expected empty hash for nil input, got %q", got)
+	}
+	if got := ComputeExcludedModelsHash([]string{}); got != "" {
+		t.Fatalf("expected empty hash for empty slice, got %q", got)
+	}
+	if got := ComputeExcludedModelsHash([]string{"  ", ""}); got != "" {
+		t.Fatalf("expected empty hash for whitespace-only entries, got %q", got)
+	}
+}
+
+func TestComputeClaudeModelsHash_Deterministic(t *testing.T) {
+	models := []config.ClaudeModel{{Name: "a", Alias: "A"}, {Name: "b"}}
+	h1 := ComputeClaudeModelsHash(models)
+	h2 := ComputeClaudeModelsHash(models)
+	if h1 == "" || h1 != h2 {
+		t.Fatalf("expected deterministic hash, got %s / %s", h1, h2)
+	}
+	if h3 := ComputeClaudeModelsHash([]config.ClaudeModel{{Name: "a"}}); h3 == h1 {
+		t.Fatalf("expected different hash when models change, got %s", h3)
+	}
+}

internal/watcher/diff/oauth_excluded.go

@@ -0,0 +1,151 @@
+package diff
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"fmt"
+	"sort"
+	"strings"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+)
+
+type ExcludedModelsSummary struct {
+	hash  string
+	count int
+}
+
+// SummarizeExcludedModels normalizes and hashes an excluded-model list.
+func SummarizeExcludedModels(list []string) ExcludedModelsSummary {
+	if len(list) == 0 {
+		return ExcludedModelsSummary{}
+	}
+	seen := make(map[string]struct{}, len(list))
+	normalized := make([]string, 0, len(list))
+	for _, entry := range list {
+		if trimmed := strings.ToLower(strings.TrimSpace(entry)); trimmed != "" {
+			if _, exists := seen[trimmed]; exists {
+				continue
+			}
+			seen[trimmed] = struct{}{}
+			normalized = append(normalized, trimmed)
+		}
+	}
+	sort.Strings(normalized)
+	return ExcludedModelsSummary{
+		hash:  ComputeExcludedModelsHash(normalized),
+		count: len(normalized),
+	}
+}
+
+// SummarizeOAuthExcludedModels summarizes OAuth excluded models per provider.
+func SummarizeOAuthExcludedModels(entries map[string][]string) map[string]ExcludedModelsSummary {
+	if len(entries) == 0 {
+		return nil
+	}
+	out := make(map[string]ExcludedModelsSummary, len(entries))
+	for k, v := range entries {
+		key := strings.ToLower(strings.TrimSpace(k))
+		if key == "" {
+			continue
+		}
+		out[key] = SummarizeExcludedModels(v)
+	}
+	return out
+}
+
+// DiffOAuthExcludedModelChanges compares OAuth excluded models maps.
+func DiffOAuthExcludedModelChanges(oldMap, newMap map[string][]string) ([]string, []string) {
+	oldSummary := SummarizeOAuthExcludedModels(oldMap)
+	newSummary := SummarizeOAuthExcludedModels(newMap)
+	keys := make(map[string]struct{}, len(oldSummary)+len(newSummary))
+	for k := range oldSummary {
+		keys[k] = struct{}{}
+	}
+	for k := range newSummary {
+		keys[k] = struct{}{}
+	}
+	changes := make([]string, 0, len(keys))
+	affected := make([]string, 0, len(keys))
+	for key := range keys {
+		oldInfo, okOld := oldSummary[key]
+		newInfo, okNew := newSummary[key]
+		switch {
+		case okOld && !okNew:
+			changes = append(changes, fmt.Sprintf("oauth-excluded-models[%s]: removed", key))
+			affected = append(affected, key)
+		case !okOld && okNew:
+			changes = append(changes, fmt.Sprintf("oauth-excluded-models[%s]: added (%d entries)", key, newInfo.count))
+			affected = append(affected, key)
+		case okOld && okNew && oldInfo.hash != newInfo.hash:
+			changes = append(changes, fmt.Sprintf("oauth-excluded-models[%s]: updated (%d -> %d entries)", key, oldInfo.count, newInfo.count))
+			affected = append(affected, key)
+		}
+	}
+	sort.Strings(changes)
+	sort.Strings(affected)
+	return changes, affected
+}
+
+type AmpModelMappingsSummary struct {
+	hash  string
+	count int
+}
+
+// SummarizeAmpModelMappings hashes Amp model mappings for change detection.
+func SummarizeAmpModelMappings(mappings []config.AmpModelMapping) AmpModelMappingsSummary {
+	if len(mappings) == 0 {
+		return AmpModelMappingsSummary{}
+	}
+	entries := make([]string, 0, len(mappings))
+	for _, mapping := range mappings {
+		from := strings.TrimSpace(mapping.From)
+		to := strings.TrimSpace(mapping.To)
+		if from == "" && to == "" {
+			continue
+		}
+		entries = append(entries, from+"->"+to)
+	}
+	if len(entries) == 0 {
+		return AmpModelMappingsSummary{}
+	}
+	sort.Strings(entries)
+	sum := sha256.Sum256([]byte(strings.Join(entries, "|")))
+	return AmpModelMappingsSummary{
+		hash:  hex.EncodeToString(sum[:]),
+		count: len(entries),
+	}
+}
+
+type VertexModelsSummary struct {
+	hash  string
+	count int
+}
+
+// SummarizeVertexModels hashes vertex-compatible models for change detection.
+func SummarizeVertexModels(models []config.VertexCompatModel) VertexModelsSummary {
+	if len(models) == 0 {
+		return VertexModelsSummary{}
+	}
+	names := make([]string, 0, len(models))
+	for _, m := range models {
+		name := strings.TrimSpace(m.Name)
+		alias := strings.TrimSpace(m.Alias)
+		if name == "" && alias == "" {
+			continue
+		}
+		if alias != "" {
+			name = alias
+		}
+		names = append(names, name)
+	}
+	if len(names) == 0 {
+		return VertexModelsSummary{}
+	}
+	sort.Strings(names)
+	sum := sha256.Sum256([]byte(strings.Join(names, "|")))
+	return VertexModelsSummary{
+		hash:  hex.EncodeToString(sum[:]),
+		count: len(names),
+	}
+}

internal/watcher/diff/oauth_excluded_test.go

@@ -0,0 +1,109 @@
+package diff
+
+import (
+	"testing"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+)
+
+func TestSummarizeExcludedModels_NormalizesAndDedupes(t *testing.T) {
+	summary := SummarizeExcludedModels([]string{"A", " a ", "B", "b"})
+	if summary.count != 2 {
+		t.Fatalf("expected 2 unique entries, got %d", summary.count)
+	}
+	if summary.hash == "" {
+		t.Fatal("expected non-empty hash")
+	}
+	if empty := SummarizeExcludedModels(nil); empty.count != 0 || empty.hash != "" {
+		t.Fatalf("expected empty summary for nil input, got %+v", empty)
+	}
+}
+
+func TestDiffOAuthExcludedModelChanges(t *testing.T) {
+	oldMap := map[string][]string{
+		"ProviderA": {"model-1", "model-2"},
+		"providerB": {"x"},
+	}
+	newMap := map[string][]string{
+		"providerA": {"model-1", "model-3"},
+		"providerC": {"y"},
+	}
+
+	changes, affected := DiffOAuthExcludedModelChanges(oldMap, newMap)
+	expectContains(t, changes, "oauth-excluded-models[providera]: updated (2 -> 2 entries)")
+	expectContains(t, changes, "oauth-excluded-models[providerb]: removed")
+	expectContains(t, changes, "oauth-excluded-models[providerc]: added (1 entries)")
+
+	if len(affected) != 3 {
+		t.Fatalf("expected 3 affected providers, got %d", len(affected))
+	}
+}
+
+func TestSummarizeAmpModelMappings(t *testing.T) {
+	summary := SummarizeAmpModelMappings([]config.AmpModelMapping{
+		{From: "a", To: "A"},
+		{From: "b", To: "B"},
+		{From: " ", To: " "}, // ignored
+	})
+	if summary.count != 2 {
+		t.Fatalf("expected 2 entries, got %d", summary.count)
+	}
+	if summary.hash == "" {
+		t.Fatal("expected non-empty hash")
+	}
+	if empty := SummarizeAmpModelMappings(nil); empty.count != 0 || empty.hash != "" {
+		t.Fatalf("expected empty summary for nil input, got %+v", empty)
+	}
+	if blank := SummarizeAmpModelMappings([]config.AmpModelMapping{{From: " ", To: " "}}); blank.count != 0 || blank.hash != "" {
+		t.Fatalf("expected blank mappings ignored, got %+v", blank)
+	}
+}
+
+func TestSummarizeOAuthExcludedModels_NormalizesKeys(t *testing.T) {
+	out := SummarizeOAuthExcludedModels(map[string][]string{
+		"ProvA": {"X"},
+		"":      {"ignored"},
+	})
+	if len(out) != 1 {
+		t.Fatalf("expected only non-empty key summary, got %d", len(out))
+	}
+	if _, ok := out["prova"]; !ok {
+		t.Fatalf("expected normalized key 'prova', got keys %v", out)
+	}
+	if out["prova"].count != 1 || out["prova"].hash == "" {
+		t.Fatalf("unexpected summary %+v", out["prova"])
+	}
+	if outEmpty := SummarizeOAuthExcludedModels(nil); outEmpty != nil {
+		t.Fatalf("expected nil map for nil input, got %v", outEmpty)
+	}
+}
+
+func TestSummarizeVertexModels(t *testing.T) {
+	summary := SummarizeVertexModels([]config.VertexCompatModel{
+		{Name: "m1"},
+		{Name: " ", Alias: "alias"},
+		{}, // ignored
+	})
+	if summary.count != 2 {
+		t.Fatalf("expected 2 vertex models, got %d", summary.count)
+	}
+	if summary.hash == "" {
+		t.Fatal("expected non-empty hash")
+	}
+	if empty := SummarizeVertexModels(nil); empty.count != 0 || empty.hash != "" {
+		t.Fatalf("expected empty summary for nil input, got %+v", empty)
+	}
+	if blank := SummarizeVertexModels([]config.VertexCompatModel{{Name: " "}}); blank.count != 0 || blank.hash != "" {
+		t.Fatalf("expected blank model ignored, got %+v", blank)
+	}
+}
+
+func expectContains(t *testing.T, list []string, target string) {
+	t.Helper()
+	for _, entry := range list {
+		if entry == target {
+			return
+		}
+	}
+	t.Fatalf("expected list to contain %q, got %#v", target, list)
+}

internal/watcher/diff/openai_compat.go

@@ -0,0 +1,183 @@
+package diff
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"fmt"
+	"sort"
+	"strings"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+)
+
+// DiffOpenAICompatibility produces human-readable change descriptions.
+func DiffOpenAICompatibility(oldList, newList []config.OpenAICompatibility) []string {
+	changes := make([]string, 0)
+	oldMap := make(map[string]config.OpenAICompatibility, len(oldList))
+	oldLabels := make(map[string]string, len(oldList))
+	for idx, entry := range oldList {
+		key, label := openAICompatKey(entry, idx)
+		oldMap[key] = entry
+		oldLabels[key] = label
+	}
+	newMap := make(map[string]config.OpenAICompatibility, len(newList))
+	newLabels := make(map[string]string, len(newList))
+	for idx, entry := range newList {
+		key, label := openAICompatKey(entry, idx)
+		newMap[key] = entry
+		newLabels[key] = label
+	}
+	keySet := make(map[string]struct{}, len(oldMap)+len(newMap))
+	for key := range oldMap {
+		keySet[key] = struct{}{}
+	}
+	for key := range newMap {
+		keySet[key] = struct{}{}
+	}
+	orderedKeys := make([]string, 0, len(keySet))
+	for key := range keySet {
+		orderedKeys = append(orderedKeys, key)
+	}
+	sort.Strings(orderedKeys)
+	for _, key := range orderedKeys {
+		oldEntry, oldOk := oldMap[key]
+		newEntry, newOk := newMap[key]
+		label := oldLabels[key]
+		if label == "" {
+			label = newLabels[key]
+		}
+		switch {
+		case !oldOk:
+			changes = append(changes, fmt.Sprintf("provider added: %s (api-keys=%d, models=%d)", label, countAPIKeys(newEntry), countOpenAIModels(newEntry.Models)))
+		case !newOk:
+			changes = append(changes, fmt.Sprintf("provider removed: %s (api-keys=%d, models=%d)", label, countAPIKeys(oldEntry), countOpenAIModels(oldEntry.Models)))
+		default:
+			if detail := describeOpenAICompatibilityUpdate(oldEntry, newEntry); detail != "" {
+				changes = append(changes, fmt.Sprintf("provider updated: %s %s", label, detail))
+			}
+		}
+	}
+	return changes
+}
+
+func describeOpenAICompatibilityUpdate(oldEntry, newEntry config.OpenAICompatibility) string {
+	oldKeyCount := countAPIKeys(oldEntry)
+	newKeyCount := countAPIKeys(newEntry)
+	oldModelCount := countOpenAIModels(oldEntry.Models)
+	newModelCount := countOpenAIModels(newEntry.Models)
+	details := make([]string, 0, 3)
+	if oldKeyCount != newKeyCount {
+		details = append(details, fmt.Sprintf("api-keys %d -> %d", oldKeyCount, newKeyCount))
+	}
+	if oldModelCount != newModelCount {
+		details = append(details, fmt.Sprintf("models %d -> %d", oldModelCount, newModelCount))
+	}
+	if !equalStringMap(oldEntry.Headers, newEntry.Headers) {
+		details = append(details, "headers updated")
+	}
+	if len(details) == 0 {
+		return ""
+	}
+	return "(" + strings.Join(details, ", ") + ")"
+}
+
+func countAPIKeys(entry config.OpenAICompatibility) int {
+	count := 0
+	for _, keyEntry := range entry.APIKeyEntries {
+		if strings.TrimSpace(keyEntry.APIKey) != "" {
+			count++
+		}
+	}
+	return count
+}
+
+func countOpenAIModels(models []config.OpenAICompatibilityModel) int {
+	count := 0
+	for _, model := range models {
+		name := strings.TrimSpace(model.Name)
+		alias := strings.TrimSpace(model.Alias)
+		if name == "" && alias == "" {
+			continue
+		}
+		count++
+	}
+	return count
+}
+
+func openAICompatKey(entry config.OpenAICompatibility, index int) (string, string) {
+	name := strings.TrimSpace(entry.Name)
+	if name != "" {
+		return "name:" + name, name
+	}
+	base := strings.TrimSpace(entry.BaseURL)
+	if base != "" {
+		return "base:" + base, base
+	}
+	for _, model := range entry.Models {
+		alias := strings.TrimSpace(model.Alias)
+		if alias == "" {
+			alias = strings.TrimSpace(model.Name)
+		}
+		if alias != "" {
+			return "alias:" + alias, alias
+		}
+	}
+	sig := openAICompatSignature(entry)
+	if sig == "" {
+		return fmt.Sprintf("index:%d", index), fmt.Sprintf("entry-%d", index+1)
+	}
+	short := sig
+	if len(short) > 8 {
+		short = short[:8]
+	}
+	return "sig:" + sig, "compat-" + short
+}
+
+func openAICompatSignature(entry config.OpenAICompatibility) string {
+	var parts []string
+
+	if v := strings.TrimSpace(entry.Name); v != "" {
+		parts = append(parts, "name="+strings.ToLower(v))
+	}
+	if v := strings.TrimSpace(entry.BaseURL); v != "" {
+		parts = append(parts, "base="+v)
+	}
+
+	models := make([]string, 0, len(entry.Models))
+	for _, model := range entry.Models {
+		name := strings.TrimSpace(model.Name)
+		alias := strings.TrimSpace(model.Alias)
+		if name == "" && alias == "" {
+			continue
+		}
+		models = append(models, strings.ToLower(name)+"|"+strings.ToLower(alias))
+	}
+	if len(models) > 0 {
+		sort.Strings(models)
+		parts = append(parts, "models="+strings.Join(models, ","))
+	}
+
+	if len(entry.Headers) > 0 {
+		keys := make([]string, 0, len(entry.Headers))
+		for k := range entry.Headers {
+			if trimmed := strings.TrimSpace(k); trimmed != "" {
+				keys = append(keys, strings.ToLower(trimmed))
+			}
+		}
+		if len(keys) > 0 {
+			sort.Strings(keys)
+			parts = append(parts, "headers="+strings.Join(keys, ","))
+		}
+	}
+
+	// Intentionally exclude API key material; only count non-empty entries.
+	if count := countAPIKeys(entry); count > 0 {
+		parts = append(parts, fmt.Sprintf("api_keys=%d", count))
+	}
+
+	if len(parts) == 0 {
+		return ""
+	}
+	sum := sha256.Sum256([]byte(strings.Join(parts, "|")))
+	return hex.EncodeToString(sum[:])
+}

internal/watcher/diff/openai_compat_test.go

@@ -0,0 +1,187 @@
+package diff
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+)
+
+func TestDiffOpenAICompatibility(t *testing.T) {
+	oldList := []config.OpenAICompatibility{
+		{
+			Name: "provider-a",
+			APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+				{APIKey: "key-a"},
+			},
+			Models: []config.OpenAICompatibilityModel{
+				{Name: "m1"},
+			},
+		},
+	}
+	newList := []config.OpenAICompatibility{
+		{
+			Name: "provider-a",
+			APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+				{APIKey: "key-a"},
+				{APIKey: "key-b"},
+			},
+			Models: []config.OpenAICompatibilityModel{
+				{Name: "m1"},
+				{Name: "m2"},
+			},
+			Headers: map[string]string{"X-Test": "1"},
+		},
+		{
+			Name:          "provider-b",
+			APIKeyEntries: []config.OpenAICompatibilityAPIKey{{APIKey: "key-b"}},
+		},
+	}
+
+	changes := DiffOpenAICompatibility(oldList, newList)
+	expectContains(t, changes, "provider added: provider-b (api-keys=1, models=0)")
+	expectContains(t, changes, "provider updated: provider-a (api-keys 1 -> 2, models 1 -> 2, headers updated)")
+}
+
+func TestDiffOpenAICompatibility_RemovedAndUnchanged(t *testing.T) {
+	oldList := []config.OpenAICompatibility{
+		{
+			Name:          "provider-a",
+			APIKeyEntries: []config.OpenAICompatibilityAPIKey{{APIKey: "key-a"}},
+			Models:        []config.OpenAICompatibilityModel{{Name: "m1"}},
+		},
+	}
+	newList := []config.OpenAICompatibility{
+		{
+			Name:          "provider-a",
+			APIKeyEntries: []config.OpenAICompatibilityAPIKey{{APIKey: "key-a"}},
+			Models:        []config.OpenAICompatibilityModel{{Name: "m1"}},
+		},
+	}
+	if changes := DiffOpenAICompatibility(oldList, newList); len(changes) != 0 {
+		t.Fatalf("expected no changes, got %v", changes)
+	}
+
+	newList = nil
+	changes := DiffOpenAICompatibility(oldList, newList)
+	expectContains(t, changes, "provider removed: provider-a (api-keys=1, models=1)")
+}
+
+func TestOpenAICompatKeyFallbacks(t *testing.T) {
+	entry := config.OpenAICompatibility{
+		BaseURL: "http://base",
+		Models:  []config.OpenAICompatibilityModel{{Alias: "alias-only"}},
+	}
+	key, label := openAICompatKey(entry, 0)
+	if key != "base:http://base" || label != "http://base" {
+		t.Fatalf("expected base key, got %s/%s", key, label)
+	}
+
+	entry.BaseURL = ""
+	key, label = openAICompatKey(entry, 1)
+	if key != "alias:alias-only" || label != "alias-only" {
+		t.Fatalf("expected alias fallback, got %s/%s", key, label)
+	}
+
+	entry.Models = nil
+	key, label = openAICompatKey(entry, 2)
+	if key != "index:2" || label != "entry-3" {
+		t.Fatalf("expected index fallback, got %s/%s", key, label)
+	}
+}
+
+func TestOpenAICompatKey_UsesName(t *testing.T) {
+	entry := config.OpenAICompatibility{Name: "My-Provider"}
+	key, label := openAICompatKey(entry, 0)
+	if key != "name:My-Provider" || label != "My-Provider" {
+		t.Fatalf("expected name key, got %s/%s", key, label)
+	}
+}
+
+func TestOpenAICompatKey_SignatureFallbackWhenOnlyAPIKeys(t *testing.T) {
+	entry := config.OpenAICompatibility{
+		APIKeyEntries: []config.OpenAICompatibilityAPIKey{{APIKey: "k1"}, {APIKey: "k2"}},
+	}
+	key, label := openAICompatKey(entry, 0)
+	if !strings.HasPrefix(key, "sig:") || !strings.HasPrefix(label, "compat-") {
+		t.Fatalf("expected signature key, got %s/%s", key, label)
+	}
+}
+
+func TestOpenAICompatSignature_EmptyReturnsEmpty(t *testing.T) {
+	if got := openAICompatSignature(config.OpenAICompatibility{}); got != "" {
+		t.Fatalf("expected empty signature, got %q", got)
+	}
+}
+
+func TestOpenAICompatSignature_StableAndNormalized(t *testing.T) {
+	a := config.OpenAICompatibility{
+		Name:    "  Provider  ",
+		BaseURL: "http://base",
+		Models: []config.OpenAICompatibilityModel{
+			{Name: "m1"},
+			{Name: "  "},
+			{Alias: "A1"},
+		},
+		Headers: map[string]string{
+			"X-Test": "1",
+			"  ":     "ignored",
+		},
+		APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+			{APIKey: "k1"},
+			{APIKey: " "},
+		},
+	}
+	b := config.OpenAICompatibility{
+		Name:    "provider",
+		BaseURL: "http://base",
+		Models: []config.OpenAICompatibilityModel{
+			{Alias: "a1"},
+			{Name: "m1"},
+		},
+		Headers: map[string]string{
+			"x-test": "2",
+		},
+		APIKeyEntries: []config.OpenAICompatibilityAPIKey{
+			{APIKey: "k2"},
+		},
+	}
+
+	sigA := openAICompatSignature(a)
+	sigB := openAICompatSignature(b)
+	if sigA == "" || sigB == "" {
+		t.Fatalf("expected non-empty signatures, got %q / %q", sigA, sigB)
+	}
+	if sigA != sigB {
+		t.Fatalf("expected normalized signatures to match, got %s / %s", sigA, sigB)
+	}
+
+	c := b
+	c.Models = append(c.Models, config.OpenAICompatibilityModel{Name: "m2"})
+	if sigC := openAICompatSignature(c); sigC == sigB {
+		t.Fatalf("expected signature to change when models change, got %s", sigC)
+	}
+}
+
+func TestCountOpenAIModelsSkipsBlanks(t *testing.T) {
+	models := []config.OpenAICompatibilityModel{
+		{Name: "m1"},
+		{Name: ""},
+		{Alias: ""},
+		{Name: " "},
+		{Alias: "a1"},
+	}
+	if got := countOpenAIModels(models); got != 2 {
+		t.Fatalf("expected 2 counted models, got %d", got)
+	}
+}
+
+func TestOpenAICompatKeyUsesModelNameWhenAliasEmpty(t *testing.T) {
+	entry := config.OpenAICompatibility{
+		Models: []config.OpenAICompatibilityModel{{Name: "model-name"}},
+	}
+	key, label := openAICompatKey(entry, 5)
+	if key != "alias:model-name" || label != "model-name" {
+		t.Fatalf("expected model-name fallback, got %s/%s", key, label)
+	}
+}

internal/watcher/watcher.go

@@ -23,6 +23,7 @@ import (
 	"github.com/fsnotify/fsnotify"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/runtime/geminicli"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/watcher/diff"
 	"gopkg.in/yaml.v3"
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
@@ -183,7 +184,7 @@ func (w *Watcher) Start(ctx context.Context) error {
 	go w.processEvents(ctx)
 
 	// Perform an initial full reload based on current config and auth dir
-	w.reloadClients(true, nil)
+	w.reloadClients(true, nil, false)
 	return nil
 }
 
@@ -276,7 +277,7 @@ func (w *Watcher) DispatchRuntimeAuthUpdate(update AuthUpdate) bool {
 	return true
 }
 
-func (w *Watcher) refreshAuthState() {
+func (w *Watcher) refreshAuthState(force bool) {
 	auths := w.SnapshotCoreAuths()
 	w.clientsMutex.Lock()
 	if len(w.runtimeAuths) > 0 {
@@ -286,12 +287,12 @@ func (w *Watcher) refreshAuthState() {
 			}
 		}
 	}
-	updates := w.prepareAuthUpdatesLocked(auths)
+	updates := w.prepareAuthUpdatesLocked(auths, force)
 	w.clientsMutex.Unlock()
 	w.dispatchAuthUpdates(updates)
 }
 
-func (w *Watcher) prepareAuthUpdatesLocked(auths []*coreauth.Auth) []AuthUpdate {
+func (w *Watcher) prepareAuthUpdatesLocked(auths []*coreauth.Auth, force bool) []AuthUpdate {
 	newState := make(map[string]*coreauth.Auth, len(auths))
 	for _, auth := range auths {
 		if auth == nil || auth.ID == "" {
@@ -318,7 +319,7 @@ func (w *Watcher) prepareAuthUpdatesLocked(auths []*coreauth.Auth) []AuthUpdate
 	for id, auth := range newState {
 		if existing, ok := w.currentAuths[id]; !ok {
 			updates = append(updates, AuthUpdate{Action: AuthUpdateActionAdd, ID: id, Auth: auth.Clone()})
-		} else if !authEqual(existing, auth) {
+		} else if force || !authEqual(existing, auth) {
 			updates = append(updates, AuthUpdate{Action: AuthUpdateActionModify, ID: id, Auth: auth.Clone()})
 		}
 	}
@@ -485,170 +486,6 @@ func normalizeAuth(a *coreauth.Auth) *coreauth.Auth {
 	return clone
 }
 
-// computeOpenAICompatModelsHash returns a stable hash for the compatibility models so that
-// changes to the model list trigger auth updates during hot reload.
-func computeOpenAICompatModelsHash(models []config.OpenAICompatibilityModel) string {
-	if len(models) == 0 {
-		return ""
-	}
-	data, err := json.Marshal(models)
-	if err != nil || len(data) == 0 {
-		return ""
-	}
-	sum := sha256.Sum256(data)
-	return hex.EncodeToString(sum[:])
-}
-
-func computeVertexCompatModelsHash(models []config.VertexCompatModel) string {
-	if len(models) == 0 {
-		return ""
-	}
-	data, err := json.Marshal(models)
-	if err != nil || len(data) == 0 {
-		return ""
-	}
-	sum := sha256.Sum256(data)
-	return hex.EncodeToString(sum[:])
-}
-
-// computeClaudeModelsHash returns a stable hash for Claude model aliases.
-func computeClaudeModelsHash(models []config.ClaudeModel) string {
-	if len(models) == 0 {
-		return ""
-	}
-	data, err := json.Marshal(models)
-	if err != nil || len(data) == 0 {
-		return ""
-	}
-	sum := sha256.Sum256(data)
-	return hex.EncodeToString(sum[:])
-}
-
-func computeExcludedModelsHash(excluded []string) string {
-	if len(excluded) == 0 {
-		return ""
-	}
-	normalized := make([]string, 0, len(excluded))
-	for _, entry := range excluded {
-		if trimmed := strings.TrimSpace(entry); trimmed != "" {
-			normalized = append(normalized, strings.ToLower(trimmed))
-		}
-	}
-	if len(normalized) == 0 {
-		return ""
-	}
-	sort.Strings(normalized)
-	data, err := json.Marshal(normalized)
-	if err != nil || len(data) == 0 {
-		return ""
-	}
-	sum := sha256.Sum256(data)
-	return hex.EncodeToString(sum[:])
-}
-
-type excludedModelsSummary struct {
-	hash  string
-	count int
-}
-
-func summarizeExcludedModels(list []string) excludedModelsSummary {
-	if len(list) == 0 {
-		return excludedModelsSummary{}
-	}
-	seen := make(map[string]struct{}, len(list))
-	normalized := make([]string, 0, len(list))
-	for _, entry := range list {
-		if trimmed := strings.ToLower(strings.TrimSpace(entry)); trimmed != "" {
-			if _, exists := seen[trimmed]; exists {
-				continue
-			}
-			seen[trimmed] = struct{}{}
-			normalized = append(normalized, trimmed)
-		}
-	}
-	sort.Strings(normalized)
-	return excludedModelsSummary{
-		hash:  computeExcludedModelsHash(normalized),
-		count: len(normalized),
-	}
-}
-
-type ampModelMappingsSummary struct {
-	hash  string
-	count int
-}
-
-func summarizeAmpModelMappings(mappings []config.AmpModelMapping) ampModelMappingsSummary {
-	if len(mappings) == 0 {
-		return ampModelMappingsSummary{}
-	}
-	entries := make([]string, 0, len(mappings))
-	for _, mapping := range mappings {
-		from := strings.TrimSpace(mapping.From)
-		to := strings.TrimSpace(mapping.To)
-		if from == "" && to == "" {
-			continue
-		}
-		entries = append(entries, from+"->"+to)
-	}
-	if len(entries) == 0 {
-		return ampModelMappingsSummary{}
-	}
-	sort.Strings(entries)
-	sum := sha256.Sum256([]byte(strings.Join(entries, "|")))
-	return ampModelMappingsSummary{
-		hash:  hex.EncodeToString(sum[:]),
-		count: len(entries),
-	}
-}
-
-func summarizeOAuthExcludedModels(entries map[string][]string) map[string]excludedModelsSummary {
-	if len(entries) == 0 {
-		return nil
-	}
-	out := make(map[string]excludedModelsSummary, len(entries))
-	for k, v := range entries {
-		key := strings.ToLower(strings.TrimSpace(k))
-		if key == "" {
-			continue
-		}
-		out[key] = summarizeExcludedModels(v)
-	}
-	return out
-}
-
-func diffOAuthExcludedModelChanges(oldMap, newMap map[string][]string) ([]string, []string) {
-	oldSummary := summarizeOAuthExcludedModels(oldMap)
-	newSummary := summarizeOAuthExcludedModels(newMap)
-	keys := make(map[string]struct{}, len(oldSummary)+len(newSummary))
-	for k := range oldSummary {
-		keys[k] = struct{}{}
-	}
-	for k := range newSummary {
-		keys[k] = struct{}{}
-	}
-	changes := make([]string, 0, len(keys))
-	affected := make([]string, 0, len(keys))
-	for key := range keys {
-		oldInfo, okOld := oldSummary[key]
-		newInfo, okNew := newSummary[key]
-		switch {
-		case okOld && !okNew:
-			changes = append(changes, fmt.Sprintf("oauth-excluded-models[%s]: removed", key))
-			affected = append(affected, key)
-		case !okOld && okNew:
-			changes = append(changes, fmt.Sprintf("oauth-excluded-models[%s]: added (%d entries)", key, newInfo.count))
-			affected = append(affected, key)
-		case okOld && okNew && oldInfo.hash != newInfo.hash:
-			changes = append(changes, fmt.Sprintf("oauth-excluded-models[%s]: updated (%d -> %d entries)", key, oldInfo.count, newInfo.count))
-			affected = append(affected, key)
-		}
-	}
-	sort.Strings(changes)
-	sort.Strings(affected)
-	return changes, affected
-}
-
 func applyAuthExcludedModelsMeta(auth *coreauth.Auth, cfg *config.Config, perKey []string, authKind string) {
 	if auth == nil || cfg == nil {
 		return
@@ -677,7 +514,7 @@ func applyAuthExcludedModelsMeta(auth *coreauth.Auth, cfg *config.Config, perKey
 		combined = append(combined, k)
 	}
 	sort.Strings(combined)
-	hash := computeExcludedModelsHash(combined)
+	hash := diff.ComputeExcludedModelsHash(combined)
 	if auth.Attributes == nil {
 		auth.Attributes = make(map[string]string)
 	}
@@ -821,7 +658,7 @@ func (w *Watcher) handleEvent(event fsnotify.Event) {
 				log.Debugf("auth file unchanged (hash match), skipping reload: %s", filepath.Base(event.Name))
 				return
 			}
-			fmt.Printf("auth file changed (%s): %s, processing incrementally\n", event.Op.String(), filepath.Base(event.Name))
+			log.Infof("auth file changed (%s): %s, processing incrementally", event.Op.String(), filepath.Base(event.Name))
 			w.addOrUpdateClient(event.Name)
 			return
 		}
@@ -829,7 +666,7 @@ func (w *Watcher) handleEvent(event fsnotify.Event) {
 			log.Debugf("ignoring remove for unknown auth file: %s", filepath.Base(event.Name))
 			return
 		}
-		fmt.Printf("auth file changed (%s): %s, processing incrementally\n", event.Op.String(), filepath.Base(event.Name))
+		log.Infof("auth file changed (%s): %s, processing incrementally", event.Op.String(), filepath.Base(event.Name))
 		w.removeClient(event.Name)
 		return
 	}
@@ -838,7 +675,7 @@ func (w *Watcher) handleEvent(event fsnotify.Event) {
 			log.Debugf("auth file unchanged (hash match), skipping reload: %s", filepath.Base(event.Name))
 			return
 		}
-		fmt.Printf("auth file changed (%s): %s, processing incrementally\n", event.Op.String(), filepath.Base(event.Name))
+		log.Infof("auth file changed (%s): %s, processing incrementally", event.Op.String(), filepath.Base(event.Name))
 		w.addOrUpdateClient(event.Name)
 	}
 }
@@ -878,7 +715,7 @@ func (w *Watcher) reloadConfigIfChanged() {
 		log.Debugf("config file content unchanged (hash match), skipping reload")
 		return
 	}
-	fmt.Printf("config file changed, reloading: %s\n", w.configPath)
+	log.Infof("config file changed, reloading: %s", w.configPath)
 	if w.reloadConfig() {
 		finalHash := newHash
 		if updatedData, errRead := os.ReadFile(w.configPath); errRead == nil && len(updatedData) > 0 {
@@ -924,7 +761,7 @@ func (w *Watcher) reloadConfig() bool {
 
 	var affectedOAuthProviders []string
 	if oldConfig != nil {
-		_, affectedOAuthProviders = diffOAuthExcludedModelChanges(oldConfig.OAuthExcludedModels, newConfig.OAuthExcludedModels)
+		_, affectedOAuthProviders = diff.DiffOAuthExcludedModelChanges(oldConfig.OAuthExcludedModels, newConfig.OAuthExcludedModels)
 	}
 
 	// Always apply the current log level based on the latest config.
@@ -937,7 +774,7 @@ func (w *Watcher) reloadConfig() bool {
 
 	// Log configuration changes in debug mode, only when there are material diffs
 	if oldConfig != nil {
-		details := buildConfigChangeDetails(oldConfig, newConfig)
+		details := diff.BuildConfigChangeDetails(oldConfig, newConfig)
 		if len(details) > 0 {
 			log.Debugf("config changes detected:")
 			for _, d := range details {
@@ -949,15 +786,16 @@ func (w *Watcher) reloadConfig() bool {
 	}
 
 	authDirChanged := oldConfig == nil || oldConfig.AuthDir != newConfig.AuthDir
+	forceAuthRefresh := oldConfig != nil && oldConfig.ForceModelPrefix != newConfig.ForceModelPrefix
 
 	log.Infof("config successfully reloaded, triggering client reload")
 	// Reload clients with new config
-	w.reloadClients(authDirChanged, affectedOAuthProviders)
+	w.reloadClients(authDirChanged, affectedOAuthProviders, forceAuthRefresh)
 	return true
 }
 
 // reloadClients performs a full scan and reload of all clients.
-func (w *Watcher) reloadClients(rescanAuth bool, affectedOAuthProviders []string) {
+func (w *Watcher) reloadClients(rescanAuth bool, affectedOAuthProviders []string, forceAuthRefresh bool) {
 	log.Debugf("starting full client load process")
 
 	w.clientsMutex.RLock()
@@ -1048,7 +886,7 @@ func (w *Watcher) reloadClients(rescanAuth bool, affectedOAuthProviders []string
 		w.reloadCallback(cfg)
 	}
 
-	w.refreshAuthState()
+	w.refreshAuthState(forceAuthRefresh)
 
 	log.Infof("full client load complete - %d clients (%d auth files + %d Gemini API keys + %d Vertex API keys + %d Claude API keys + %d Codex keys + %d OpenAI-compat)",
 		totalNewClients,
@@ -1099,7 +937,7 @@ func (w *Watcher) addOrUpdateClient(path string) {
 
 	w.clientsMutex.Unlock() // Unlock before the callback
 
-	w.refreshAuthState()
+	w.refreshAuthState(false)
 
 	if w.reloadCallback != nil {
 		log.Debugf("triggering server update callback after add/update")
@@ -1118,7 +956,7 @@ func (w *Watcher) removeClient(path string) {
 
 	w.clientsMutex.Unlock() // Release the lock before the callback
 
-	w.refreshAuthState()
+	w.refreshAuthState(false)
 
 	if w.reloadCallback != nil {
 		log.Debugf("triggering server update callback after removal")
@@ -1147,6 +985,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 			if key == "" {
 				continue
 			}
+			prefix := strings.TrimSpace(entry.Prefix)
 			base := strings.TrimSpace(entry.BaseURL)
 			proxyURL := strings.TrimSpace(entry.ProxyURL)
 			id, token := idGen.next("gemini:apikey", key, base)
@@ -1162,6 +1001,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 				ID:         id,
 				Provider:   "gemini",
 				Label:      "gemini-apikey",
+				Prefix:     prefix,
 				Status:     coreauth.StatusActive,
 				ProxyURL:   proxyURL,
 				Attributes: attrs,
@@ -1179,6 +1019,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 			if key == "" {
 				continue
 			}
+			prefix := strings.TrimSpace(ck.Prefix)
 			base := strings.TrimSpace(ck.BaseURL)
 			id, token := idGen.next("claude:apikey", key, base)
 			attrs := map[string]string{
@@ -1188,7 +1029,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 			if base != "" {
 				attrs["base_url"] = base
 			}
-			if hash := computeClaudeModelsHash(ck.Models); hash != "" {
+			if hash := diff.ComputeClaudeModelsHash(ck.Models); hash != "" {
 				attrs["models_hash"] = hash
 			}
 			addConfigHeadersToAttrs(ck.Headers, attrs)
@@ -1197,6 +1038,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 				ID:         id,
 				Provider:   "claude",
 				Label:      "claude-apikey",
+				Prefix:     prefix,
 				Status:     coreauth.StatusActive,
 				ProxyURL:   proxyURL,
 				Attributes: attrs,
@@ -1213,6 +1055,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 			if key == "" {
 				continue
 			}
+			prefix := strings.TrimSpace(ck.Prefix)
 			id, token := idGen.next("codex:apikey", key, ck.BaseURL)
 			attrs := map[string]string{
 				"source":  fmt.Sprintf("config:codex[%s]", token),
@@ -1227,6 +1070,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 				ID:         id,
 				Provider:   "codex",
 				Label:      "codex-apikey",
+				Prefix:     prefix,
 				Status:     coreauth.StatusActive,
 				ProxyURL:   proxyURL,
 				Attributes: attrs,
@@ -1238,6 +1082,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 		}
 		for i := range cfg.OpenAICompatibility {
 			compat := &cfg.OpenAICompatibility[i]
+			prefix := strings.TrimSpace(compat.Prefix)
 			providerName := strings.ToLower(strings.TrimSpace(compat.Name))
 			if providerName == "" {
 				providerName = "openai-compatibility"
@@ -1261,7 +1106,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 				if key != "" {
 					attrs["api_key"] = key
 				}
-				if hash := computeOpenAICompatModelsHash(compat.Models); hash != "" {
+				if hash := diff.ComputeOpenAICompatModelsHash(compat.Models); hash != "" {
 					attrs["models_hash"] = hash
 				}
 				addConfigHeadersToAttrs(compat.Headers, attrs)
@@ -1269,6 +1114,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 					ID:         id,
 					Provider:   providerName,
 					Label:      compat.Name,
+					Prefix:     prefix,
 					Status:     coreauth.StatusActive,
 					ProxyURL:   proxyURL,
 					Attributes: attrs,
@@ -1287,7 +1133,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 					"compat_name":  compat.Name,
 					"provider_key": providerName,
 				}
-				if hash := computeOpenAICompatModelsHash(compat.Models); hash != "" {
+				if hash := diff.ComputeOpenAICompatModelsHash(compat.Models); hash != "" {
 					attrs["models_hash"] = hash
 				}
 				addConfigHeadersToAttrs(compat.Headers, attrs)
@@ -1295,6 +1141,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 					ID:         id,
 					Provider:   providerName,
 					Label:      compat.Name,
+					Prefix:     prefix,
 					Status:     coreauth.StatusActive,
 					Attributes: attrs,
 					CreatedAt:  now,
@@ -1312,8 +1159,9 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 		base := strings.TrimSpace(compat.BaseURL)
 
 		key := strings.TrimSpace(compat.APIKey)
+		prefix := strings.TrimSpace(compat.Prefix)
 		proxyURL := strings.TrimSpace(compat.ProxyURL)
-		idKind := fmt.Sprintf("vertex:apikey:%s", base)
+		idKind := "vertex:apikey"
 		id, token := idGen.next(idKind, key, base, proxyURL)
 		attrs := map[string]string{
 			"source":       fmt.Sprintf("config:vertex-apikey[%s]", token),
@@ -1323,7 +1171,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 		if key != "" {
 			attrs["api_key"] = key
 		}
-		if hash := computeVertexCompatModelsHash(compat.Models); hash != "" {
+		if hash := diff.ComputeVertexCompatModelsHash(compat.Models); hash != "" {
 			attrs["models_hash"] = hash
 		}
 		addConfigHeadersToAttrs(compat.Headers, attrs)
@@ -1331,6 +1179,7 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 			ID:         id,
 			Provider:   providerName,
 			Label:      "vertex-apikey",
+			Prefix:     prefix,
 			Status:     coreauth.StatusActive,
 			ProxyURL:   proxyURL,
 			Attributes: attrs,
@@ -1383,10 +1232,20 @@ func (w *Watcher) SnapshotCoreAuths() []*coreauth.Auth {
 			proxyURL = p
 		}
 
+		prefix := ""
+		if rawPrefix, ok := metadata["prefix"].(string); ok {
+			trimmed := strings.TrimSpace(rawPrefix)
+			trimmed = strings.Trim(trimmed, "/")
+			if trimmed != "" && !strings.Contains(trimmed, "/") {
+				prefix = trimmed
+			}
+		}
+
 		a := &coreauth.Auth{
 			ID:       id,
 			Provider: provider,
 			Label:    label,
+			Prefix:   prefix,
 			Status:   coreauth.StatusActive,
 			Attributes: map[string]string{
 				"source": full,
@@ -1473,6 +1332,7 @@ func synthesizeGeminiVirtualAuths(primary *coreauth.Auth, metadata map[string]an
 			Attributes: attrs,
 			Metadata:   metadataCopy,
 			ProxyURL:   primary.ProxyURL,
+			Prefix:     primary.Prefix,
 			CreatedAt:  now,
 			UpdatedAt:  now,
 			Runtime:    geminicli.NewVirtualCredential(projectID, shared),
@@ -1586,324 +1446,6 @@ func BuildAPIKeyClients(cfg *config.Config) (int, int, int, int, int) {
 	return geminiAPIKeyCount, vertexCompatAPIKeyCount, claudeAPIKeyCount, codexAPIKeyCount, openAICompatCount
 }
 
-func diffOpenAICompatibility(oldList, newList []config.OpenAICompatibility) []string {
-	changes := make([]string, 0)
-	oldMap := make(map[string]config.OpenAICompatibility, len(oldList))
-	oldLabels := make(map[string]string, len(oldList))
-	for idx, entry := range oldList {
-		key, label := openAICompatKey(entry, idx)
-		oldMap[key] = entry
-		oldLabels[key] = label
-	}
-	newMap := make(map[string]config.OpenAICompatibility, len(newList))
-	newLabels := make(map[string]string, len(newList))
-	for idx, entry := range newList {
-		key, label := openAICompatKey(entry, idx)
-		newMap[key] = entry
-		newLabels[key] = label
-	}
-	keySet := make(map[string]struct{}, len(oldMap)+len(newMap))
-	for key := range oldMap {
-		keySet[key] = struct{}{}
-	}
-	for key := range newMap {
-		keySet[key] = struct{}{}
-	}
-	orderedKeys := make([]string, 0, len(keySet))
-	for key := range keySet {
-		orderedKeys = append(orderedKeys, key)
-	}
-	sort.Strings(orderedKeys)
-	for _, key := range orderedKeys {
-		oldEntry, oldOk := oldMap[key]
-		newEntry, newOk := newMap[key]
-		label := oldLabels[key]
-		if label == "" {
-			label = newLabels[key]
-		}
-		switch {
-		case !oldOk:
-			changes = append(changes, fmt.Sprintf("provider added: %s (api-keys=%d, models=%d)", label, countAPIKeys(newEntry), countOpenAIModels(newEntry.Models)))
-		case !newOk:
-			changes = append(changes, fmt.Sprintf("provider removed: %s (api-keys=%d, models=%d)", label, countAPIKeys(oldEntry), countOpenAIModels(oldEntry.Models)))
-		default:
-			if detail := describeOpenAICompatibilityUpdate(oldEntry, newEntry); detail != "" {
-				changes = append(changes, fmt.Sprintf("provider updated: %s %s", label, detail))
-			}
-		}
-	}
-	return changes
-}
-
-func describeOpenAICompatibilityUpdate(oldEntry, newEntry config.OpenAICompatibility) string {
-	oldKeyCount := countAPIKeys(oldEntry)
-	newKeyCount := countAPIKeys(newEntry)
-	oldModelCount := countOpenAIModels(oldEntry.Models)
-	newModelCount := countOpenAIModels(newEntry.Models)
-	details := make([]string, 0, 3)
-	if oldKeyCount != newKeyCount {
-		details = append(details, fmt.Sprintf("api-keys %d -> %d", oldKeyCount, newKeyCount))
-	}
-	if oldModelCount != newModelCount {
-		details = append(details, fmt.Sprintf("models %d -> %d", oldModelCount, newModelCount))
-	}
-	if !equalStringMap(oldEntry.Headers, newEntry.Headers) {
-		details = append(details, "headers updated")
-	}
-	if len(details) == 0 {
-		return ""
-	}
-	return "(" + strings.Join(details, ", ") + ")"
-}
-
-func countAPIKeys(entry config.OpenAICompatibility) int {
-	count := 0
-	for _, keyEntry := range entry.APIKeyEntries {
-		if strings.TrimSpace(keyEntry.APIKey) != "" {
-			count++
-		}
-	}
-	return count
-}
-
-func countOpenAIModels(models []config.OpenAICompatibilityModel) int {
-	count := 0
-	for _, model := range models {
-		name := strings.TrimSpace(model.Name)
-		alias := strings.TrimSpace(model.Alias)
-		if name == "" && alias == "" {
-			continue
-		}
-		count++
-	}
-	return count
-}
-
-func openAICompatKey(entry config.OpenAICompatibility, index int) (string, string) {
-	name := strings.TrimSpace(entry.Name)
-	if name != "" {
-		return "name:" + name, name
-	}
-	base := strings.TrimSpace(entry.BaseURL)
-	if base != "" {
-		return "base:" + base, base
-	}
-	for _, model := range entry.Models {
-		alias := strings.TrimSpace(model.Alias)
-		if alias == "" {
-			alias = strings.TrimSpace(model.Name)
-		}
-		if alias != "" {
-			return "alias:" + alias, alias
-		}
-	}
-	return fmt.Sprintf("index:%d", index), fmt.Sprintf("entry-%d", index+1)
-}
-
-// buildConfigChangeDetails computes a redacted, human-readable list of config changes.
-// It avoids printing secrets (like API keys) and focuses on structural or non-sensitive fields.
-func buildConfigChangeDetails(oldCfg, newCfg *config.Config) []string {
-	changes := make([]string, 0, 16)
-	if oldCfg == nil || newCfg == nil {
-		return changes
-	}
-
-	// Simple scalars
-	if oldCfg.Port != newCfg.Port {
-		changes = append(changes, fmt.Sprintf("port: %d -> %d", oldCfg.Port, newCfg.Port))
-	}
-	if oldCfg.AuthDir != newCfg.AuthDir {
-		changes = append(changes, fmt.Sprintf("auth-dir: %s -> %s", oldCfg.AuthDir, newCfg.AuthDir))
-	}
-	if oldCfg.Debug != newCfg.Debug {
-		changes = append(changes, fmt.Sprintf("debug: %t -> %t", oldCfg.Debug, newCfg.Debug))
-	}
-	if oldCfg.LoggingToFile != newCfg.LoggingToFile {
-		changes = append(changes, fmt.Sprintf("logging-to-file: %t -> %t", oldCfg.LoggingToFile, newCfg.LoggingToFile))
-	}
-	if oldCfg.UsageStatisticsEnabled != newCfg.UsageStatisticsEnabled {
-		changes = append(changes, fmt.Sprintf("usage-statistics-enabled: %t -> %t", oldCfg.UsageStatisticsEnabled, newCfg.UsageStatisticsEnabled))
-	}
-	if oldCfg.DisableCooling != newCfg.DisableCooling {
-		changes = append(changes, fmt.Sprintf("disable-cooling: %t -> %t", oldCfg.DisableCooling, newCfg.DisableCooling))
-	}
-	if oldCfg.RequestLog != newCfg.RequestLog {
-		changes = append(changes, fmt.Sprintf("request-log: %t -> %t", oldCfg.RequestLog, newCfg.RequestLog))
-	}
-	if oldCfg.RequestRetry != newCfg.RequestRetry {
-		changes = append(changes, fmt.Sprintf("request-retry: %d -> %d", oldCfg.RequestRetry, newCfg.RequestRetry))
-	}
-	if oldCfg.MaxRetryInterval != newCfg.MaxRetryInterval {
-		changes = append(changes, fmt.Sprintf("max-retry-interval: %d -> %d", oldCfg.MaxRetryInterval, newCfg.MaxRetryInterval))
-	}
-	if oldCfg.ProxyURL != newCfg.ProxyURL {
-		changes = append(changes, fmt.Sprintf("proxy-url: %s -> %s", oldCfg.ProxyURL, newCfg.ProxyURL))
-	}
-	if oldCfg.WebsocketAuth != newCfg.WebsocketAuth {
-		changes = append(changes, fmt.Sprintf("ws-auth: %t -> %t", oldCfg.WebsocketAuth, newCfg.WebsocketAuth))
-	}
-
-	// Quota-exceeded behavior
-	if oldCfg.QuotaExceeded.SwitchProject != newCfg.QuotaExceeded.SwitchProject {
-		changes = append(changes, fmt.Sprintf("quota-exceeded.switch-project: %t -> %t", oldCfg.QuotaExceeded.SwitchProject, newCfg.QuotaExceeded.SwitchProject))
-	}
-	if oldCfg.QuotaExceeded.SwitchPreviewModel != newCfg.QuotaExceeded.SwitchPreviewModel {
-		changes = append(changes, fmt.Sprintf("quota-exceeded.switch-preview-model: %t -> %t", oldCfg.QuotaExceeded.SwitchPreviewModel, newCfg.QuotaExceeded.SwitchPreviewModel))
-	}
-
-	// API keys (redacted) and counts
-	if len(oldCfg.APIKeys) != len(newCfg.APIKeys) {
-		changes = append(changes, fmt.Sprintf("api-keys count: %d -> %d", len(oldCfg.APIKeys), len(newCfg.APIKeys)))
-	} else if !reflect.DeepEqual(trimStrings(oldCfg.APIKeys), trimStrings(newCfg.APIKeys)) {
-		changes = append(changes, "api-keys: values updated (count unchanged, redacted)")
-	}
-	if len(oldCfg.GeminiKey) != len(newCfg.GeminiKey) {
-		changes = append(changes, fmt.Sprintf("gemini-api-key count: %d -> %d", len(oldCfg.GeminiKey), len(newCfg.GeminiKey)))
-	} else {
-		for i := range oldCfg.GeminiKey {
-			if i >= len(newCfg.GeminiKey) {
-				break
-			}
-			o := oldCfg.GeminiKey[i]
-			n := newCfg.GeminiKey[i]
-			if strings.TrimSpace(o.BaseURL) != strings.TrimSpace(n.BaseURL) {
-				changes = append(changes, fmt.Sprintf("gemini[%d].base-url: %s -> %s", i, strings.TrimSpace(o.BaseURL), strings.TrimSpace(n.BaseURL)))
-			}
-			if strings.TrimSpace(o.ProxyURL) != strings.TrimSpace(n.ProxyURL) {
-				changes = append(changes, fmt.Sprintf("gemini[%d].proxy-url: %s -> %s", i, strings.TrimSpace(o.ProxyURL), strings.TrimSpace(n.ProxyURL)))
-			}
-			if strings.TrimSpace(o.APIKey) != strings.TrimSpace(n.APIKey) {
-				changes = append(changes, fmt.Sprintf("gemini[%d].api-key: updated", i))
-			}
-			if !equalStringMap(o.Headers, n.Headers) {
-				changes = append(changes, fmt.Sprintf("gemini[%d].headers: updated", i))
-			}
-			oldExcluded := summarizeExcludedModels(o.ExcludedModels)
-			newExcluded := summarizeExcludedModels(n.ExcludedModels)
-			if oldExcluded.hash != newExcluded.hash {
-				changes = append(changes, fmt.Sprintf("gemini[%d].excluded-models: updated (%d -> %d entries)", i, oldExcluded.count, newExcluded.count))
-			}
-		}
-	}
-
-	// Claude keys (do not print key material)
-	if len(oldCfg.ClaudeKey) != len(newCfg.ClaudeKey) {
-		changes = append(changes, fmt.Sprintf("claude-api-key count: %d -> %d", len(oldCfg.ClaudeKey), len(newCfg.ClaudeKey)))
-	} else {
-		for i := range oldCfg.ClaudeKey {
-			if i >= len(newCfg.ClaudeKey) {
-				break
-			}
-			o := oldCfg.ClaudeKey[i]
-			n := newCfg.ClaudeKey[i]
-			if strings.TrimSpace(o.BaseURL) != strings.TrimSpace(n.BaseURL) {
-				changes = append(changes, fmt.Sprintf("claude[%d].base-url: %s -> %s", i, strings.TrimSpace(o.BaseURL), strings.TrimSpace(n.BaseURL)))
-			}
-			if strings.TrimSpace(o.ProxyURL) != strings.TrimSpace(n.ProxyURL) {
-				changes = append(changes, fmt.Sprintf("claude[%d].proxy-url: %s -> %s", i, strings.TrimSpace(o.ProxyURL), strings.TrimSpace(n.ProxyURL)))
-			}
-			if strings.TrimSpace(o.APIKey) != strings.TrimSpace(n.APIKey) {
-				changes = append(changes, fmt.Sprintf("claude[%d].api-key: updated", i))
-			}
-			if !equalStringMap(o.Headers, n.Headers) {
-				changes = append(changes, fmt.Sprintf("claude[%d].headers: updated", i))
-			}
-			oldExcluded := summarizeExcludedModels(o.ExcludedModels)
-			newExcluded := summarizeExcludedModels(n.ExcludedModels)
-			if oldExcluded.hash != newExcluded.hash {
-				changes = append(changes, fmt.Sprintf("claude[%d].excluded-models: updated (%d -> %d entries)", i, oldExcluded.count, newExcluded.count))
-			}
-		}
-	}
-
-	// Codex keys (do not print key material)
-	if len(oldCfg.CodexKey) != len(newCfg.CodexKey) {
-		changes = append(changes, fmt.Sprintf("codex-api-key count: %d -> %d", len(oldCfg.CodexKey), len(newCfg.CodexKey)))
-	} else {
-		for i := range oldCfg.CodexKey {
-			if i >= len(newCfg.CodexKey) {
-				break
-			}
-			o := oldCfg.CodexKey[i]
-			n := newCfg.CodexKey[i]
-			if strings.TrimSpace(o.BaseURL) != strings.TrimSpace(n.BaseURL) {
-				changes = append(changes, fmt.Sprintf("codex[%d].base-url: %s -> %s", i, strings.TrimSpace(o.BaseURL), strings.TrimSpace(n.BaseURL)))
-			}
-			if strings.TrimSpace(o.ProxyURL) != strings.TrimSpace(n.ProxyURL) {
-				changes = append(changes, fmt.Sprintf("codex[%d].proxy-url: %s -> %s", i, strings.TrimSpace(o.ProxyURL), strings.TrimSpace(n.ProxyURL)))
-			}
-			if strings.TrimSpace(o.APIKey) != strings.TrimSpace(n.APIKey) {
-				changes = append(changes, fmt.Sprintf("codex[%d].api-key: updated", i))
-			}
-			if !equalStringMap(o.Headers, n.Headers) {
-				changes = append(changes, fmt.Sprintf("codex[%d].headers: updated", i))
-			}
-			oldExcluded := summarizeExcludedModels(o.ExcludedModels)
-			newExcluded := summarizeExcludedModels(n.ExcludedModels)
-			if oldExcluded.hash != newExcluded.hash {
-				changes = append(changes, fmt.Sprintf("codex[%d].excluded-models: updated (%d -> %d entries)", i, oldExcluded.count, newExcluded.count))
-			}
-		}
-	}
-
-	// AmpCode settings (redacted where needed)
-	oldAmpURL := strings.TrimSpace(oldCfg.AmpCode.UpstreamURL)
-	newAmpURL := strings.TrimSpace(newCfg.AmpCode.UpstreamURL)
-	if oldAmpURL != newAmpURL {
-		changes = append(changes, fmt.Sprintf("ampcode.upstream-url: %s -> %s", oldAmpURL, newAmpURL))
-	}
-	oldAmpKey := strings.TrimSpace(oldCfg.AmpCode.UpstreamAPIKey)
-	newAmpKey := strings.TrimSpace(newCfg.AmpCode.UpstreamAPIKey)
-	switch {
-	case oldAmpKey == "" && newAmpKey != "":
-		changes = append(changes, "ampcode.upstream-api-key: added")
-	case oldAmpKey != "" && newAmpKey == "":
-		changes = append(changes, "ampcode.upstream-api-key: removed")
-	case oldAmpKey != newAmpKey:
-		changes = append(changes, "ampcode.upstream-api-key: updated")
-	}
-	if oldCfg.AmpCode.RestrictManagementToLocalhost != newCfg.AmpCode.RestrictManagementToLocalhost {
-		changes = append(changes, fmt.Sprintf("ampcode.restrict-management-to-localhost: %t -> %t", oldCfg.AmpCode.RestrictManagementToLocalhost, newCfg.AmpCode.RestrictManagementToLocalhost))
-	}
-	oldMappings := summarizeAmpModelMappings(oldCfg.AmpCode.ModelMappings)
-	newMappings := summarizeAmpModelMappings(newCfg.AmpCode.ModelMappings)
-	if oldMappings.hash != newMappings.hash {
-		changes = append(changes, fmt.Sprintf("ampcode.model-mappings: updated (%d -> %d entries)", oldMappings.count, newMappings.count))
-	}
-
-	if entries, _ := diffOAuthExcludedModelChanges(oldCfg.OAuthExcludedModels, newCfg.OAuthExcludedModels); len(entries) > 0 {
-		changes = append(changes, entries...)
-	}
-
-	// Remote management (never print the key)
-	if oldCfg.RemoteManagement.AllowRemote != newCfg.RemoteManagement.AllowRemote {
-		changes = append(changes, fmt.Sprintf("remote-management.allow-remote: %t -> %t", oldCfg.RemoteManagement.AllowRemote, newCfg.RemoteManagement.AllowRemote))
-	}
-	if oldCfg.RemoteManagement.DisableControlPanel != newCfg.RemoteManagement.DisableControlPanel {
-		changes = append(changes, fmt.Sprintf("remote-management.disable-control-panel: %t -> %t", oldCfg.RemoteManagement.DisableControlPanel, newCfg.RemoteManagement.DisableControlPanel))
-	}
-	if oldCfg.RemoteManagement.SecretKey != newCfg.RemoteManagement.SecretKey {
-		switch {
-		case oldCfg.RemoteManagement.SecretKey == "" && newCfg.RemoteManagement.SecretKey != "":
-			changes = append(changes, "remote-management.secret-key: created")
-		case oldCfg.RemoteManagement.SecretKey != "" && newCfg.RemoteManagement.SecretKey == "":
-			changes = append(changes, "remote-management.secret-key: deleted")
-		default:
-			changes = append(changes, "remote-management.secret-key: updated")
-		}
-	}
-
-	// OpenAI compatibility providers (summarized)
-	if compat := diffOpenAICompatibility(oldCfg.OpenAICompatibility, newCfg.OpenAICompatibility); len(compat) > 0 {
-		changes = append(changes, "openai-compatibility:")
-		for _, c := range compat {
-			changes = append(changes, "  "+c)
-		}
-	}
-
-	return changes
-}
-
 func addConfigHeadersToAttrs(headers map[string]string, attrs map[string]string) {
 	if len(headers) == 0 || attrs == nil {
 		return
@@ -1917,23 +1459,3 @@ func addConfigHeadersToAttrs(headers map[string]string, attrs map[string]string)
 		attrs["header:"+key] = val
 	}
 }
-
-func trimStrings(in []string) []string {
-	out := make([]string, len(in))
-	for i := range in {
-		out[i] = strings.TrimSpace(in[i])
-	}
-	return out
-}
-
-func equalStringMap(a, b map[string]string) bool {
-	if len(a) != len(b) {
-		return false
-	}
-	for k, v := range a {
-		if b[k] != v {
-			return false
-		}
-	}
-	return true
-}

internal/watcher/watcher_test.go

@@ -0,0 +1,609 @@
+package watcher
+
+import (
+	"context"
+	"crypto/sha256"
+	"encoding/json"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"github.com/fsnotify/fsnotify"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/watcher/diff"
+	coreauth "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/auth"
+	"gopkg.in/yaml.v3"
+)
+
+func TestApplyAuthExcludedModelsMeta_APIKey(t *testing.T) {
+	auth := &coreauth.Auth{Attributes: map[string]string{}}
+	cfg := &config.Config{}
+	perKey := []string{" Model-1 ", "model-2"}
+
+	applyAuthExcludedModelsMeta(auth, cfg, perKey, "apikey")
+
+	expected := diff.ComputeExcludedModelsHash([]string{"model-1", "model-2"})
+	if got := auth.Attributes["excluded_models_hash"]; got != expected {
+		t.Fatalf("expected hash %s, got %s", expected, got)
+	}
+	if got := auth.Attributes["auth_kind"]; got != "apikey" {
+		t.Fatalf("expected auth_kind=apikey, got %s", got)
+	}
+}
+
+func TestApplyAuthExcludedModelsMeta_OAuthProvider(t *testing.T) {
+	auth := &coreauth.Auth{
+		Provider:   "TestProv",
+		Attributes: map[string]string{},
+	}
+	cfg := &config.Config{
+		OAuthExcludedModels: map[string][]string{
+			"testprov": {"A", "b"},
+		},
+	}
+
+	applyAuthExcludedModelsMeta(auth, cfg, nil, "oauth")
+
+	expected := diff.ComputeExcludedModelsHash([]string{"a", "b"})
+	if got := auth.Attributes["excluded_models_hash"]; got != expected {
+		t.Fatalf("expected hash %s, got %s", expected, got)
+	}
+	if got := auth.Attributes["auth_kind"]; got != "oauth" {
+		t.Fatalf("expected auth_kind=oauth, got %s", got)
+	}
+}
+
+func TestBuildAPIKeyClientsCounts(t *testing.T) {
+	cfg := &config.Config{
+		GeminiKey: []config.GeminiKey{{APIKey: "g1"}, {APIKey: "g2"}},
+		VertexCompatAPIKey: []config.VertexCompatKey{
+			{APIKey: "v1"},
+		},
+		ClaudeKey: []config.ClaudeKey{{APIKey: "c1"}},
+		CodexKey:  []config.CodexKey{{APIKey: "x1"}, {APIKey: "x2"}},
+		OpenAICompatibility: []config.OpenAICompatibility{
+			{APIKeyEntries: []config.OpenAICompatibilityAPIKey{{APIKey: "o1"}, {APIKey: "o2"}}},
+		},
+	}
+
+	gemini, vertex, claude, codex, compat := BuildAPIKeyClients(cfg)
+	if gemini != 2 || vertex != 1 || claude != 1 || codex != 2 || compat != 2 {
+		t.Fatalf("unexpected counts: %d %d %d %d %d", gemini, vertex, claude, codex, compat)
+	}
+}
+
+func TestNormalizeAuthStripsTemporalFields(t *testing.T) {
+	now := time.Now()
+	auth := &coreauth.Auth{
+		CreatedAt:        now,
+		UpdatedAt:        now,
+		LastRefreshedAt:  now,
+		NextRefreshAfter: now,
+		Quota: coreauth.QuotaState{
+			NextRecoverAt: now,
+		},
+		Runtime: map[string]any{"k": "v"},
+	}
+
+	normalized := normalizeAuth(auth)
+	if !normalized.CreatedAt.IsZero() || !normalized.UpdatedAt.IsZero() || !normalized.LastRefreshedAt.IsZero() || !normalized.NextRefreshAfter.IsZero() {
+		t.Fatal("expected time fields to be zeroed")
+	}
+	if normalized.Runtime != nil {
+		t.Fatal("expected runtime to be nil")
+	}
+	if !normalized.Quota.NextRecoverAt.IsZero() {
+		t.Fatal("expected quota.NextRecoverAt to be zeroed")
+	}
+}
+
+func TestMatchProvider(t *testing.T) {
+	if _, ok := matchProvider("OpenAI", []string{"openai", "claude"}); !ok {
+		t.Fatal("expected match to succeed ignoring case")
+	}
+	if _, ok := matchProvider("missing", []string{"openai"}); ok {
+		t.Fatal("expected match to fail for unknown provider")
+	}
+}
+
+func TestSnapshotCoreAuths_ConfigAndAuthFiles(t *testing.T) {
+	authDir := t.TempDir()
+	metadata := map[string]any{
+		"type":       "gemini",
+		"email":      "user@example.com",
+		"project_id": "proj-a, proj-b",
+		"proxy_url":  "https://proxy",
+	}
+	authFile := filepath.Join(authDir, "gemini.json")
+	data, err := json.Marshal(metadata)
+	if err != nil {
+		t.Fatalf("failed to marshal metadata: %v", err)
+	}
+	if err = os.WriteFile(authFile, data, 0o644); err != nil {
+		t.Fatalf("failed to write auth file: %v", err)
+	}
+
+	cfg := &config.Config{
+		AuthDir: authDir,
+		GeminiKey: []config.GeminiKey{
+			{
+				APIKey:         "g-key",
+				BaseURL:        "https://gemini",
+				ExcludedModels: []string{"Model-A", "model-b"},
+				Headers:        map[string]string{"X-Req": "1"},
+			},
+		},
+		OAuthExcludedModels: map[string][]string{
+			"gemini-cli": {"Foo", "bar"},
+		},
+	}
+
+	w := &Watcher{authDir: authDir}
+	w.SetConfig(cfg)
+
+	auths := w.SnapshotCoreAuths()
+	if len(auths) != 4 {
+		t.Fatalf("expected 4 auth entries (1 config + 1 primary + 2 virtual), got %d", len(auths))
+	}
+
+	var geminiAPIKeyAuth *coreauth.Auth
+	var geminiPrimary *coreauth.Auth
+	virtuals := make([]*coreauth.Auth, 0)
+	for _, a := range auths {
+		switch {
+		case a.Provider == "gemini" && a.Attributes["api_key"] == "g-key":
+			geminiAPIKeyAuth = a
+		case a.Attributes["gemini_virtual_primary"] == "true":
+			geminiPrimary = a
+		case strings.TrimSpace(a.Attributes["gemini_virtual_parent"]) != "":
+			virtuals = append(virtuals, a)
+		}
+	}
+	if geminiAPIKeyAuth == nil {
+		t.Fatal("expected synthesized Gemini API key auth")
+	}
+	expectedAPIKeyHash := diff.ComputeExcludedModelsHash([]string{"Model-A", "model-b"})
+	if geminiAPIKeyAuth.Attributes["excluded_models_hash"] != expectedAPIKeyHash {
+		t.Fatalf("expected API key excluded hash %s, got %s", expectedAPIKeyHash, geminiAPIKeyAuth.Attributes["excluded_models_hash"])
+	}
+	if geminiAPIKeyAuth.Attributes["auth_kind"] != "apikey" {
+		t.Fatalf("expected auth_kind=apikey, got %s", geminiAPIKeyAuth.Attributes["auth_kind"])
+	}
+
+	if geminiPrimary == nil {
+		t.Fatal("expected primary gemini-cli auth from file")
+	}
+	if !geminiPrimary.Disabled || geminiPrimary.Status != coreauth.StatusDisabled {
+		t.Fatal("expected primary gemini-cli auth to be disabled when virtual auths are synthesized")
+	}
+	expectedOAuthHash := diff.ComputeExcludedModelsHash([]string{"Foo", "bar"})
+	if geminiPrimary.Attributes["excluded_models_hash"] != expectedOAuthHash {
+		t.Fatalf("expected OAuth excluded hash %s, got %s", expectedOAuthHash, geminiPrimary.Attributes["excluded_models_hash"])
+	}
+	if geminiPrimary.Attributes["auth_kind"] != "oauth" {
+		t.Fatalf("expected auth_kind=oauth, got %s", geminiPrimary.Attributes["auth_kind"])
+	}
+
+	if len(virtuals) != 2 {
+		t.Fatalf("expected 2 virtual auths, got %d", len(virtuals))
+	}
+	for _, v := range virtuals {
+		if v.Attributes["gemini_virtual_parent"] != geminiPrimary.ID {
+			t.Fatalf("virtual auth missing parent link to %s", geminiPrimary.ID)
+		}
+		if v.Attributes["excluded_models_hash"] != expectedOAuthHash {
+			t.Fatalf("expected virtual excluded hash %s, got %s", expectedOAuthHash, v.Attributes["excluded_models_hash"])
+		}
+		if v.Status != coreauth.StatusActive {
+			t.Fatalf("expected virtual auth to be active, got %s", v.Status)
+		}
+	}
+}
+
+func TestReloadConfigIfChanged_TriggersOnChangeAndSkipsUnchanged(t *testing.T) {
+	tmpDir := t.TempDir()
+	authDir := filepath.Join(tmpDir, "auth")
+	if err := os.MkdirAll(authDir, 0o755); err != nil {
+		t.Fatalf("failed to create auth dir: %v", err)
+	}
+
+	configPath := filepath.Join(tmpDir, "config.yaml")
+	writeConfig := func(port int, allowRemote bool) {
+		cfg := &config.Config{
+			Port:    port,
+			AuthDir: authDir,
+			RemoteManagement: config.RemoteManagement{
+				AllowRemote: allowRemote,
+			},
+		}
+		data, err := yaml.Marshal(cfg)
+		if err != nil {
+			t.Fatalf("failed to marshal config: %v", err)
+		}
+		if err = os.WriteFile(configPath, data, 0o644); err != nil {
+			t.Fatalf("failed to write config: %v", err)
+		}
+	}
+
+	writeConfig(8080, false)
+
+	reloads := 0
+	w := &Watcher{
+		configPath:     configPath,
+		authDir:        authDir,
+		reloadCallback: func(*config.Config) { reloads++ },
+	}
+
+	w.reloadConfigIfChanged()
+	if reloads != 1 {
+		t.Fatalf("expected first reload to trigger callback once, got %d", reloads)
+	}
+
+	// Same content should be skipped by hash check.
+	w.reloadConfigIfChanged()
+	if reloads != 1 {
+		t.Fatalf("expected unchanged config to be skipped, callback count %d", reloads)
+	}
+
+	writeConfig(9090, true)
+	w.reloadConfigIfChanged()
+	if reloads != 2 {
+		t.Fatalf("expected changed config to trigger reload, callback count %d", reloads)
+	}
+	w.clientsMutex.RLock()
+	defer w.clientsMutex.RUnlock()
+	if w.config == nil || w.config.Port != 9090 || !w.config.RemoteManagement.AllowRemote {
+		t.Fatalf("expected config to be updated after reload, got %+v", w.config)
+	}
+}
+
+func TestStartAndStopSuccess(t *testing.T) {
+	tmpDir := t.TempDir()
+	authDir := filepath.Join(tmpDir, "auth")
+	if err := os.MkdirAll(authDir, 0o755); err != nil {
+		t.Fatalf("failed to create auth dir: %v", err)
+	}
+	configPath := filepath.Join(tmpDir, "config.yaml")
+	if err := os.WriteFile(configPath, []byte("auth_dir: "+authDir), 0o644); err != nil {
+		t.Fatalf("failed to create config file: %v", err)
+	}
+
+	var reloads int32
+	w, err := NewWatcher(configPath, authDir, func(*config.Config) {
+		atomic.AddInt32(&reloads, 1)
+	})
+	if err != nil {
+		t.Fatalf("failed to create watcher: %v", err)
+	}
+	w.SetConfig(&config.Config{AuthDir: authDir})
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	if err := w.Start(ctx); err != nil {
+		t.Fatalf("expected Start to succeed: %v", err)
+	}
+	cancel()
+	if err := w.Stop(); err != nil {
+		t.Fatalf("expected Stop to succeed: %v", err)
+	}
+	if got := atomic.LoadInt32(&reloads); got != 1 {
+		t.Fatalf("expected one reload callback, got %d", got)
+	}
+}
+
+func TestStartFailsWhenConfigMissing(t *testing.T) {
+	tmpDir := t.TempDir()
+	authDir := filepath.Join(tmpDir, "auth")
+	if err := os.MkdirAll(authDir, 0o755); err != nil {
+		t.Fatalf("failed to create auth dir: %v", err)
+	}
+	configPath := filepath.Join(tmpDir, "missing-config.yaml")
+
+	w, err := NewWatcher(configPath, authDir, nil)
+	if err != nil {
+		t.Fatalf("failed to create watcher: %v", err)
+	}
+	defer w.Stop()
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	if err := w.Start(ctx); err == nil {
+		t.Fatal("expected Start to fail for missing config file")
+	}
+}
+
+func TestDispatchRuntimeAuthUpdateEnqueuesAndUpdatesState(t *testing.T) {
+	queue := make(chan AuthUpdate, 4)
+	w := &Watcher{}
+	w.SetAuthUpdateQueue(queue)
+	defer w.stopDispatch()
+
+	auth := &coreauth.Auth{ID: "auth-1", Provider: "test"}
+	if ok := w.DispatchRuntimeAuthUpdate(AuthUpdate{Action: AuthUpdateActionAdd, Auth: auth}); !ok {
+		t.Fatal("expected DispatchRuntimeAuthUpdate to enqueue")
+	}
+
+	select {
+	case update := <-queue:
+		if update.Action != AuthUpdateActionAdd || update.Auth.ID != "auth-1" {
+			t.Fatalf("unexpected update: %+v", update)
+		}
+	case <-time.After(2 * time.Second):
+		t.Fatal("timed out waiting for auth update")
+	}
+
+	if ok := w.DispatchRuntimeAuthUpdate(AuthUpdate{Action: AuthUpdateActionDelete, ID: "auth-1"}); !ok {
+		t.Fatal("expected delete update to enqueue")
+	}
+	select {
+	case update := <-queue:
+		if update.Action != AuthUpdateActionDelete || update.ID != "auth-1" {
+			t.Fatalf("unexpected delete update: %+v", update)
+		}
+	case <-time.After(2 * time.Second):
+		t.Fatal("timed out waiting for delete update")
+	}
+	w.clientsMutex.RLock()
+	if _, exists := w.runtimeAuths["auth-1"]; exists {
+		w.clientsMutex.RUnlock()
+		t.Fatal("expected runtime auth to be cleared after delete")
+	}
+	w.clientsMutex.RUnlock()
+}
+
+func TestAddOrUpdateClientSkipsUnchanged(t *testing.T) {
+	tmpDir := t.TempDir()
+	authFile := filepath.Join(tmpDir, "sample.json")
+	if err := os.WriteFile(authFile, []byte(`{"type":"demo"}`), 0o644); err != nil {
+		t.Fatalf("failed to create auth file: %v", err)
+	}
+	data, _ := os.ReadFile(authFile)
+	sum := sha256.Sum256(data)
+
+	var reloads int32
+	w := &Watcher{
+		authDir: tmpDir,
+		lastAuthHashes: map[string]string{
+			filepath.Clean(authFile): hexString(sum[:]),
+		},
+		reloadCallback: func(*config.Config) {
+			atomic.AddInt32(&reloads, 1)
+		},
+	}
+	w.SetConfig(&config.Config{AuthDir: tmpDir})
+
+	w.addOrUpdateClient(authFile)
+	if got := atomic.LoadInt32(&reloads); got != 0 {
+		t.Fatalf("expected no reload for unchanged file, got %d", got)
+	}
+}
+
+func TestAddOrUpdateClientTriggersReloadAndHash(t *testing.T) {
+	tmpDir := t.TempDir()
+	authFile := filepath.Join(tmpDir, "sample.json")
+	if err := os.WriteFile(authFile, []byte(`{"type":"demo","api_key":"k"}`), 0o644); err != nil {
+		t.Fatalf("failed to create auth file: %v", err)
+	}
+
+	var reloads int32
+	w := &Watcher{
+		authDir:        tmpDir,
+		lastAuthHashes: make(map[string]string),
+		reloadCallback: func(*config.Config) {
+			atomic.AddInt32(&reloads, 1)
+		},
+	}
+	w.SetConfig(&config.Config{AuthDir: tmpDir})
+
+	w.addOrUpdateClient(authFile)
+
+	if got := atomic.LoadInt32(&reloads); got != 1 {
+		t.Fatalf("expected reload callback once, got %d", got)
+	}
+	normalized := filepath.Clean(authFile)
+	if _, ok := w.lastAuthHashes[normalized]; !ok {
+		t.Fatalf("expected hash to be stored for %s", normalized)
+	}
+}
+
+func TestRemoveClientRemovesHash(t *testing.T) {
+	tmpDir := t.TempDir()
+	authFile := filepath.Join(tmpDir, "sample.json")
+	var reloads int32
+
+	w := &Watcher{
+		authDir: tmpDir,
+		lastAuthHashes: map[string]string{
+			filepath.Clean(authFile): "hash",
+		},
+		reloadCallback: func(*config.Config) {
+			atomic.AddInt32(&reloads, 1)
+		},
+	}
+	w.SetConfig(&config.Config{AuthDir: tmpDir})
+
+	w.removeClient(authFile)
+	if _, ok := w.lastAuthHashes[filepath.Clean(authFile)]; ok {
+		t.Fatal("expected hash to be removed after deletion")
+	}
+	if got := atomic.LoadInt32(&reloads); got != 1 {
+		t.Fatalf("expected reload callback once, got %d", got)
+	}
+}
+
+func TestShouldDebounceRemove(t *testing.T) {
+	w := &Watcher{}
+	path := filepath.Clean("test.json")
+
+	if w.shouldDebounceRemove(path, time.Now()) {
+		t.Fatal("first call should not debounce")
+	}
+	if !w.shouldDebounceRemove(path, time.Now()) {
+		t.Fatal("second call within window should debounce")
+	}
+
+	w.clientsMutex.Lock()
+	w.lastRemoveTimes = map[string]time.Time{path: time.Now().Add(-2 * authRemoveDebounceWindow)}
+	w.clientsMutex.Unlock()
+
+	if w.shouldDebounceRemove(path, time.Now()) {
+		t.Fatal("call after window should not debounce")
+	}
+}
+
+func TestAuthFileUnchangedUsesHash(t *testing.T) {
+	tmpDir := t.TempDir()
+	authFile := filepath.Join(tmpDir, "sample.json")
+	content := []byte(`{"type":"demo"}`)
+	if err := os.WriteFile(authFile, content, 0o644); err != nil {
+		t.Fatalf("failed to write auth file: %v", err)
+	}
+
+	w := &Watcher{lastAuthHashes: make(map[string]string)}
+	unchanged, err := w.authFileUnchanged(authFile)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if unchanged {
+		t.Fatal("expected first check to report changed")
+	}
+
+	sum := sha256.Sum256(content)
+	w.lastAuthHashes[filepath.Clean(authFile)] = hexString(sum[:])
+
+	unchanged, err = w.authFileUnchanged(authFile)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if !unchanged {
+		t.Fatal("expected hash match to report unchanged")
+	}
+}
+
+func TestReloadClientsCachesAuthHashes(t *testing.T) {
+	tmpDir := t.TempDir()
+	authFile := filepath.Join(tmpDir, "one.json")
+	if err := os.WriteFile(authFile, []byte(`{"type":"demo"}`), 0o644); err != nil {
+		t.Fatalf("failed to write auth file: %v", err)
+	}
+	w := &Watcher{
+		authDir: tmpDir,
+		config:  &config.Config{AuthDir: tmpDir},
+	}
+
+	w.reloadClients(true, nil, false)
+
+	w.clientsMutex.RLock()
+	defer w.clientsMutex.RUnlock()
+	if len(w.lastAuthHashes) != 1 {
+		t.Fatalf("expected hash cache for one auth file, got %d", len(w.lastAuthHashes))
+	}
+}
+
+func TestReloadClientsLogsConfigDiffs(t *testing.T) {
+	tmpDir := t.TempDir()
+	oldCfg := &config.Config{AuthDir: tmpDir, Port: 1, Debug: false}
+	newCfg := &config.Config{AuthDir: tmpDir, Port: 2, Debug: true}
+
+	w := &Watcher{
+		authDir: tmpDir,
+		config:  oldCfg,
+	}
+	w.SetConfig(oldCfg)
+	w.oldConfigYaml, _ = yaml.Marshal(oldCfg)
+
+	w.clientsMutex.Lock()
+	w.config = newCfg
+	w.clientsMutex.Unlock()
+
+	w.reloadClients(false, nil, false)
+}
+
+func TestSetAuthUpdateQueueNilResetsDispatch(t *testing.T) {
+	w := &Watcher{}
+	queue := make(chan AuthUpdate, 1)
+	w.SetAuthUpdateQueue(queue)
+	if w.dispatchCond == nil || w.dispatchCancel == nil {
+		t.Fatal("expected dispatch to be initialized")
+	}
+	w.SetAuthUpdateQueue(nil)
+	if w.dispatchCancel != nil {
+		t.Fatal("expected dispatch cancel to be cleared when queue nil")
+	}
+}
+
+func TestStopConfigReloadTimerSafeWhenNil(t *testing.T) {
+	w := &Watcher{}
+	w.stopConfigReloadTimer()
+	w.configReloadMu.Lock()
+	w.configReloadTimer = time.AfterFunc(10*time.Millisecond, func() {})
+	w.configReloadMu.Unlock()
+	time.Sleep(1 * time.Millisecond)
+	w.stopConfigReloadTimer()
+}
+
+func TestHandleEventRemovesAuthFile(t *testing.T) {
+	tmpDir := t.TempDir()
+	authFile := filepath.Join(tmpDir, "remove.json")
+	if err := os.WriteFile(authFile, []byte(`{"type":"demo"}`), 0o644); err != nil {
+		t.Fatalf("failed to write auth file: %v", err)
+	}
+	if err := os.Remove(authFile); err != nil {
+		t.Fatalf("failed to remove auth file pre-check: %v", err)
+	}
+
+	var reloads int32
+	w := &Watcher{
+		authDir: tmpDir,
+		config:  &config.Config{AuthDir: tmpDir},
+		lastAuthHashes: map[string]string{
+			filepath.Clean(authFile): "hash",
+		},
+		reloadCallback: func(*config.Config) {
+			atomic.AddInt32(&reloads, 1)
+		},
+	}
+	w.handleEvent(fsnotify.Event{Name: authFile, Op: fsnotify.Remove})
+
+	if atomic.LoadInt32(&reloads) != 1 {
+		t.Fatalf("expected reload callback once, got %d", reloads)
+	}
+	if _, ok := w.lastAuthHashes[filepath.Clean(authFile)]; ok {
+		t.Fatal("expected hash entry to be removed")
+	}
+}
+
+func TestDispatchAuthUpdatesFlushesQueue(t *testing.T) {
+	queue := make(chan AuthUpdate, 4)
+	w := &Watcher{}
+	w.SetAuthUpdateQueue(queue)
+	defer w.stopDispatch()
+
+	w.dispatchAuthUpdates([]AuthUpdate{
+		{Action: AuthUpdateActionAdd, ID: "a"},
+		{Action: AuthUpdateActionModify, ID: "b"},
+	})
+
+	got := make([]AuthUpdate, 0, 2)
+	for i := 0; i < 2; i++ {
+		select {
+		case u := <-queue:
+			got = append(got, u)
+		case <-time.After(2 * time.Second):
+			t.Fatalf("timed out waiting for update %d", i)
+		}
+	}
+	if len(got) != 2 || got[0].ID != "a" || got[1].ID != "b" {
+		t.Fatalf("unexpected updates order/content: %+v", got)
+	}
+}
+
+func hexString(data []byte) string {
+	return strings.ToLower(fmt.Sprintf("%x", data))
+}

sdk/api/handlers/gemini/gemini_handlers.go

@@ -84,7 +84,8 @@ func (h *GeminiAPIHandler) GeminiGetHandler(c *gin.Context) {
 		})
 		return
 	}
-	switch request.Action {
+	action := strings.TrimPrefix(request.Action, "/")
+	switch action {
 	case "gemini-3-pro-preview":
 		c.JSON(http.StatusOK, gin.H{
 			"name":             "models/gemini-3-pro-preview",
@@ -189,7 +190,7 @@ func (h *GeminiAPIHandler) GeminiHandler(c *gin.Context) {
 		})
 		return
 	}
-	action := strings.Split(request.Action, ":")
+	action := strings.Split(strings.TrimPrefix(request.Action, "/"), ":")
 	if len(action) != 2 {
 		c.JSON(http.StatusNotFound, handlers.ErrorResponse{
 			Error: handlers.ErrorDetail{

sdk/api/handlers/handlers.go

@@ -5,6 +5,7 @@ package handlers
 
 import (
 	"bytes"
+	"encoding/json"
 	"fmt"
 	"net/http"
 	"strings"
@@ -48,9 +49,6 @@ type BaseAPIHandler struct {
 
 	// Cfg holds the current application configuration.
 	Cfg *config.SDKConfig
-
-	// OpenAICompatProviders is a list of provider names for OpenAI compatibility.
-	OpenAICompatProviders []string
 }
 
 // NewBaseAPIHandlers creates a new API handlers instance.
@@ -62,11 +60,10 @@ type BaseAPIHandler struct {
 //
 // Returns:
 //   - *BaseAPIHandler: A new API handlers instance
-func NewBaseAPIHandlers(cfg *config.SDKConfig, authManager *coreauth.Manager, openAICompatProviders []string) *BaseAPIHandler {
+func NewBaseAPIHandlers(cfg *config.SDKConfig, authManager *coreauth.Manager) *BaseAPIHandler {
 	return &BaseAPIHandler{
-		Cfg:                   cfg,
-		AuthManager:           authManager,
-		OpenAICompatProviders: openAICompatProviders,
+		Cfg:         cfg,
+		AuthManager: authManager,
 	}
 }
 
@@ -117,6 +114,16 @@ func (h *BaseAPIHandler) GetContextWithCancel(handler interfaces.APIHandler, c *
 	newCtx = context.WithValue(newCtx, "handler", handler)
 	return newCtx, func(params ...interface{}) {
 		if h.Cfg.RequestLog && len(params) == 1 {
+			if existing, exists := c.Get("API_RESPONSE"); exists {
+				if existingBytes, ok := existing.([]byte); ok && len(bytes.TrimSpace(existingBytes)) > 0 {
+					switch params[0].(type) {
+					case error, string:
+						cancel()
+						return
+					}
+				}
+			}
+
 			var payload []byte
 			switch data := params[0].(type) {
 			case []byte:
@@ -331,30 +338,19 @@ func (h *BaseAPIHandler) getRequestDetails(modelName string) (providers []string
 	// Resolve "auto" model to an actual available model first
 	resolvedModelName := util.ResolveAutoModel(modelName)
 
-	providerName, extractedModelName, isDynamic := h.parseDynamicModel(resolvedModelName)
-
-	targetModelName := resolvedModelName
-	if isDynamic {
-		targetModelName = extractedModelName
-	}
-
 	// Normalize the model name to handle dynamic thinking suffixes before determining the provider.
-	normalizedModel, metadata = normalizeModelMetadata(targetModelName)
+	normalizedModel, metadata = normalizeModelMetadata(resolvedModelName)
 
-	if isDynamic {
-		providers = []string{providerName}
-	} else {
-		// For non-dynamic models, use the normalizedModel to get the provider name.
-		providers = util.GetProviderName(normalizedModel)
-		if len(providers) == 0 && metadata != nil {
-			if originalRaw, ok := metadata[util.ThinkingOriginalModelMetadataKey]; ok {
-				if originalModel, okStr := originalRaw.(string); okStr {
-					originalModel = strings.TrimSpace(originalModel)
-					if originalModel != "" && !strings.EqualFold(originalModel, normalizedModel) {
-						if altProviders := util.GetProviderName(originalModel); len(altProviders) > 0 {
-							providers = altProviders
-							normalizedModel = originalModel
-						}
+	// Use the normalizedModel to get the provider name.
+	providers = util.GetProviderName(normalizedModel)
+	if len(providers) == 0 && metadata != nil {
+		if originalRaw, ok := metadata[util.ThinkingOriginalModelMetadataKey]; ok {
+			if originalModel, okStr := originalRaw.(string); okStr {
+				originalModel = strings.TrimSpace(originalModel)
+				if originalModel != "" && !strings.EqualFold(originalModel, normalizedModel) {
+					if altProviders := util.GetProviderName(originalModel); len(altProviders) > 0 {
+						providers = altProviders
+						normalizedModel = originalModel
 					}
 				}
 			}
@@ -372,30 +368,6 @@ func (h *BaseAPIHandler) getRequestDetails(modelName string) (providers []string
 	return providers, normalizedModel, metadata, nil
 }
 
-func (h *BaseAPIHandler) parseDynamicModel(modelName string) (providerName, model string, isDynamic bool) {
-	var providerPart, modelPart string
-	for _, sep := range []string{"://"} {
-		if parts := strings.SplitN(modelName, sep, 2); len(parts) == 2 {
-			providerPart = parts[0]
-			modelPart = parts[1]
-			break
-		}
-	}
-
-	if providerPart == "" {
-		return "", modelName, false
-	}
-
-	// Check if the provider is a configured openai-compatibility provider
-	for _, pName := range h.OpenAICompatProviders {
-		if pName == providerPart {
-			return providerPart, modelPart, true
-		}
-	}
-
-	return "", modelName, false
-}
-
 func cloneBytes(src []byte) []byte {
 	if len(src) == 0 {
 		return nil
@@ -437,12 +409,53 @@ func (h *BaseAPIHandler) WriteErrorResponse(c *gin.Context, msg *interfaces.Erro
 			}
 		}
 	}
-	c.Status(status)
+
+	errText := http.StatusText(status)
 	if msg != nil && msg.Error != nil {
-		_, _ = c.Writer.Write([]byte(msg.Error.Error()))
-	} else {
-		_, _ = c.Writer.Write([]byte(http.StatusText(status)))
+		if v := strings.TrimSpace(msg.Error.Error()); v != "" {
+			errText = v
+		}
 	}
+
+	// Prefer preserving upstream JSON error bodies when possible.
+	buildJSONBody := func() []byte {
+		trimmed := strings.TrimSpace(errText)
+		if trimmed != "" && json.Valid([]byte(trimmed)) {
+			return []byte(trimmed)
+		}
+		errType := "invalid_request_error"
+		switch status {
+		case http.StatusUnauthorized:
+			errType = "authentication_error"
+		case http.StatusForbidden:
+			errType = "permission_error"
+		case http.StatusTooManyRequests:
+			errType = "rate_limit_error"
+		default:
+			if status >= http.StatusInternalServerError {
+				errType = "server_error"
+			}
+		}
+		payload, err := json.Marshal(ErrorResponse{
+			Error: ErrorDetail{
+				Message: errText,
+				Type:    errType,
+			},
+		})
+		if err != nil {
+			return []byte(fmt.Sprintf(`{"error":{"message":%q,"type":"server_error"}}`, errText))
+		}
+		return payload
+	}
+
+	body := buildJSONBody()
+	c.Set("API_RESPONSE", bytes.Clone(body))
+
+	if !c.Writer.Written() {
+		c.Writer.Header().Set("Content-Type", "application/json")
+	}
+	c.Status(status)
+	_, _ = c.Writer.Write(body)
 }
 
 func (h *BaseAPIHandler) LoggingAPIResponseError(ctx context.Context, err *interfaces.ErrorMessage) {

sdk/cliproxy/auth/manager.go

@@ -363,10 +363,11 @@ func (m *Manager) executeWithProvider(ctx context.Context, provider string, req
 	if provider == "" {
 		return cliproxyexecutor.Response{}, &Error{Code: "provider_not_found", Message: "provider identifier is empty"}
 	}
+	routeModel := req.Model
 	tried := make(map[string]struct{})
 	var lastErr error
 	for {
-		auth, executor, errPick := m.pickNext(ctx, provider, req.Model, opts, tried)
+		auth, executor, errPick := m.pickNext(ctx, provider, routeModel, opts, tried)
 		if errPick != nil {
 			if lastErr != nil {
 				return cliproxyexecutor.Response{}, lastErr
@@ -396,8 +397,10 @@ func (m *Manager) executeWithProvider(ctx context.Context, provider string, req
 			execCtx = context.WithValue(execCtx, roundTripperContextKey{}, rt)
 			execCtx = context.WithValue(execCtx, "cliproxy.roundtripper", rt)
 		}
-		resp, errExec := executor.Execute(execCtx, auth, req, opts)
-		result := Result{AuthID: auth.ID, Provider: provider, Model: req.Model, Success: errExec == nil}
+		execReq := req
+		execReq.Model, execReq.Metadata = rewriteModelForAuth(routeModel, req.Metadata, auth)
+		resp, errExec := executor.Execute(execCtx, auth, execReq, opts)
+		result := Result{AuthID: auth.ID, Provider: provider, Model: routeModel, Success: errExec == nil}
 		if errExec != nil {
 			result.Error = &Error{Message: errExec.Error()}
 			var se cliproxyexecutor.StatusError
@@ -420,10 +423,11 @@ func (m *Manager) executeCountWithProvider(ctx context.Context, provider string,
 	if provider == "" {
 		return cliproxyexecutor.Response{}, &Error{Code: "provider_not_found", Message: "provider identifier is empty"}
 	}
+	routeModel := req.Model
 	tried := make(map[string]struct{})
 	var lastErr error
 	for {
-		auth, executor, errPick := m.pickNext(ctx, provider, req.Model, opts, tried)
+		auth, executor, errPick := m.pickNext(ctx, provider, routeModel, opts, tried)
 		if errPick != nil {
 			if lastErr != nil {
 				return cliproxyexecutor.Response{}, lastErr
@@ -453,8 +457,10 @@ func (m *Manager) executeCountWithProvider(ctx context.Context, provider string,
 			execCtx = context.WithValue(execCtx, roundTripperContextKey{}, rt)
 			execCtx = context.WithValue(execCtx, "cliproxy.roundtripper", rt)
 		}
-		resp, errExec := executor.CountTokens(execCtx, auth, req, opts)
-		result := Result{AuthID: auth.ID, Provider: provider, Model: req.Model, Success: errExec == nil}
+		execReq := req
+		execReq.Model, execReq.Metadata = rewriteModelForAuth(routeModel, req.Metadata, auth)
+		resp, errExec := executor.CountTokens(execCtx, auth, execReq, opts)
+		result := Result{AuthID: auth.ID, Provider: provider, Model: routeModel, Success: errExec == nil}
 		if errExec != nil {
 			result.Error = &Error{Message: errExec.Error()}
 			var se cliproxyexecutor.StatusError
@@ -477,10 +483,11 @@ func (m *Manager) executeStreamWithProvider(ctx context.Context, provider string
 	if provider == "" {
 		return nil, &Error{Code: "provider_not_found", Message: "provider identifier is empty"}
 	}
+	routeModel := req.Model
 	tried := make(map[string]struct{})
 	var lastErr error
 	for {
-		auth, executor, errPick := m.pickNext(ctx, provider, req.Model, opts, tried)
+		auth, executor, errPick := m.pickNext(ctx, provider, routeModel, opts, tried)
 		if errPick != nil {
 			if lastErr != nil {
 				return nil, lastErr
@@ -510,14 +517,16 @@ func (m *Manager) executeStreamWithProvider(ctx context.Context, provider string
 			execCtx = context.WithValue(execCtx, roundTripperContextKey{}, rt)
 			execCtx = context.WithValue(execCtx, "cliproxy.roundtripper", rt)
 		}
-		chunks, errStream := executor.ExecuteStream(execCtx, auth, req, opts)
+		execReq := req
+		execReq.Model, execReq.Metadata = rewriteModelForAuth(routeModel, req.Metadata, auth)
+		chunks, errStream := executor.ExecuteStream(execCtx, auth, execReq, opts)
 		if errStream != nil {
 			rerr := &Error{Message: errStream.Error()}
 			var se cliproxyexecutor.StatusError
 			if errors.As(errStream, &se) && se != nil {
 				rerr.HTTPStatus = se.StatusCode()
 			}
-			result := Result{AuthID: auth.ID, Provider: provider, Model: req.Model, Success: false, Error: rerr}
+			result := Result{AuthID: auth.ID, Provider: provider, Model: routeModel, Success: false, Error: rerr}
 			result.RetryAfter = retryAfterFromError(errStream)
 			m.MarkResult(execCtx, result)
 			lastErr = errStream
@@ -535,18 +544,66 @@ func (m *Manager) executeStreamWithProvider(ctx context.Context, provider string
 					if errors.As(chunk.Err, &se) && se != nil {
 						rerr.HTTPStatus = se.StatusCode()
 					}
-					m.MarkResult(streamCtx, Result{AuthID: streamAuth.ID, Provider: streamProvider, Model: req.Model, Success: false, Error: rerr})
+					m.MarkResult(streamCtx, Result{AuthID: streamAuth.ID, Provider: streamProvider, Model: routeModel, Success: false, Error: rerr})
 				}
 				out <- chunk
 			}
 			if !failed {
-				m.MarkResult(streamCtx, Result{AuthID: streamAuth.ID, Provider: streamProvider, Model: req.Model, Success: true})
+				m.MarkResult(streamCtx, Result{AuthID: streamAuth.ID, Provider: streamProvider, Model: routeModel, Success: true})
 			}
 		}(execCtx, auth.Clone(), provider, chunks)
 		return out, nil
 	}
 }
 
+func rewriteModelForAuth(model string, metadata map[string]any, auth *Auth) (string, map[string]any) {
+	if auth == nil || model == "" {
+		return model, metadata
+	}
+	prefix := strings.TrimSpace(auth.Prefix)
+	if prefix == "" {
+		return model, metadata
+	}
+	needle := prefix + "/"
+	if !strings.HasPrefix(model, needle) {
+		return model, metadata
+	}
+	rewritten := strings.TrimPrefix(model, needle)
+	return rewritten, stripPrefixFromMetadata(metadata, needle)
+}
+
+func stripPrefixFromMetadata(metadata map[string]any, needle string) map[string]any {
+	if len(metadata) == 0 || needle == "" {
+		return metadata
+	}
+	keys := []string{
+		util.ThinkingOriginalModelMetadataKey,
+		util.GeminiOriginalModelMetadataKey,
+	}
+	var out map[string]any
+	for _, key := range keys {
+		raw, ok := metadata[key]
+		if !ok {
+			continue
+		}
+		value, okStr := raw.(string)
+		if !okStr || !strings.HasPrefix(value, needle) {
+			continue
+		}
+		if out == nil {
+			out = make(map[string]any, len(metadata))
+			for k, v := range metadata {
+				out[k] = v
+			}
+		}
+		out[key] = strings.TrimPrefix(value, needle)
+	}
+	if out == nil {
+		return metadata
+	}
+	return out
+}
+
 func (m *Manager) normalizeProviders(providers []string) []string {
 	if len(providers) == 0 {
 		return nil

sdk/cliproxy/auth/types.go

@@ -19,6 +19,8 @@ type Auth struct {
 	Index uint64 `json:"-"`
 	// Provider is the upstream provider key (e.g. "gemini", "claude").
 	Provider string `json:"provider"`
+	// Prefix optionally namespaces models for routing (e.g., "teamA/gemini-3-pro-preview").
+	Prefix string `json:"prefix,omitempty"`
 	// FileName stores the relative or absolute path of the backing auth file.
 	FileName string `json:"-"`
 	// Storage holds the token persistence implementation used during login flows.

sdk/cliproxy/service.go

@@ -787,7 +787,7 @@ func (s *Service) registerModelsForAuth(a *coreauth.Auth) {
 						if providerKey == "" {
 							providerKey = "openai-compatibility"
 						}
-						GlobalModelRegistry().RegisterClient(a.ID, providerKey, ms)
+						GlobalModelRegistry().RegisterClient(a.ID, providerKey, applyModelPrefixes(ms, a.Prefix, s.cfg.ForceModelPrefix))
 					} else {
 						// Ensure stale registrations are cleared when model list becomes empty.
 						GlobalModelRegistry().UnregisterClient(a.ID)
@@ -807,7 +807,7 @@ func (s *Service) registerModelsForAuth(a *coreauth.Auth) {
 		if key == "" {
 			key = strings.ToLower(strings.TrimSpace(a.Provider))
 		}
-		GlobalModelRegistry().RegisterClient(a.ID, key, models)
+		GlobalModelRegistry().RegisterClient(a.ID, key, applyModelPrefixes(models, a.Prefix, s.cfg != nil && s.cfg.ForceModelPrefix))
 		return
 	}
 
@@ -987,6 +987,48 @@ func applyExcludedModels(models []*ModelInfo, excluded []string) []*ModelInfo {
 	return filtered
 }
 
+func applyModelPrefixes(models []*ModelInfo, prefix string, forceModelPrefix bool) []*ModelInfo {
+	trimmedPrefix := strings.TrimSpace(prefix)
+	if trimmedPrefix == "" || len(models) == 0 {
+		return models
+	}
+
+	out := make([]*ModelInfo, 0, len(models)*2)
+	seen := make(map[string]struct{}, len(models)*2)
+
+	addModel := func(model *ModelInfo) {
+		if model == nil {
+			return
+		}
+		id := strings.TrimSpace(model.ID)
+		if id == "" {
+			return
+		}
+		if _, exists := seen[id]; exists {
+			return
+		}
+		seen[id] = struct{}{}
+		out = append(out, model)
+	}
+
+	for _, model := range models {
+		if model == nil {
+			continue
+		}
+		baseID := strings.TrimSpace(model.ID)
+		if baseID == "" {
+			continue
+		}
+		if !forceModelPrefix || trimmedPrefix == baseID {
+			addModel(model)
+		}
+		clone := *model
+		clone.ID = trimmedPrefix + "/" + baseID
+		addModel(&clone)
+	}
+	return out
+}
+
 // matchWildcard performs case-insensitive wildcard matching where '*' matches any substring.
 func matchWildcard(pattern, value string) bool {
 	if pattern == "" {

sdk/config/config.go

@@ -9,6 +9,11 @@ type SDKConfig struct {
 	// ProxyURL is the URL of an optional proxy server to use for outbound requests.
 	ProxyURL string `yaml:"proxy-url" json:"proxy-url"`
 
+	// ForceModelPrefix requires explicit model prefixes (e.g., "teamA/gemini-3-pro-preview")
+	// to target prefixed credentials. When false, unprefixed model requests may use prefixed
+	// credentials as well.
+	ForceModelPrefix bool `yaml:"force-model-prefix" json:"force-model-prefix"`
+
 	// RequestLog enables or disables detailed request logging functionality.
 	RequestLog bool `yaml:"request-log" json:"request-log"`
 

test/antigravity_claude_signature_test.go

@@ -0,0 +1,109 @@
+package test
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	agclaude "github.com/router-for-me/CLIProxyAPI/v6/internal/translator/antigravity/claude"
+	"github.com/tidwall/gjson"
+)
+
+func TestAntigravityClaudeRequest_DropsUnsignedThinkingBlocks(t *testing.T) {
+	model := "gemini-claude-sonnet-4-5-thinking"
+	input := []byte(`{
+  "model":"` + model + `",
+  "messages":[
+    {"role":"assistant","content":[{"type":"thinking","thinking":"secret without signature"}]},
+    {"role":"user","content":[{"type":"text","text":"hi"}]}
+  ]
+}`)
+
+	out := agclaude.ConvertClaudeRequestToAntigravity(model, input, false)
+	contents := gjson.GetBytes(out, "request.contents")
+	if !contents.Exists() || !contents.IsArray() {
+		t.Fatalf("expected request.contents array, got: %s", string(out))
+	}
+	if got := len(contents.Array()); got != 1 {
+		t.Fatalf("expected 1 content message after dropping unsigned thinking-only assistant message, got %d: %s", got, contents.Raw)
+	}
+	if role := contents.Array()[0].Get("role").String(); role != "user" {
+		t.Fatalf("expected remaining message role=user, got %q", role)
+	}
+}
+
+func TestAntigravityClaudeStreamResponse_EmitsSignatureDeltaForStandaloneSignaturePart(t *testing.T) {
+	raw := []byte(`{
+  "response":{
+    "responseId":"resp_1",
+    "modelVersion":"claude-sonnet-4-5-thinking",
+    "candidates":[{
+      "content":{"parts":[
+        {"text":"THOUGHT","thought":true},
+        {"thought":true,"thoughtSignature":"sig123"},
+        {"text":"ANSWER","thought":false}
+      ]},
+      "finishReason":"STOP"
+    }],
+    "usageMetadata":{"promptTokenCount":1,"candidatesTokenCount":1,"thoughtsTokenCount":1,"totalTokenCount":3}
+  }
+}`)
+
+	var param any
+	chunks := agclaude.ConvertAntigravityResponseToClaude(context.Background(), "", nil, nil, raw, &param)
+	joined := strings.Join(chunks, "")
+	if !strings.Contains(joined, `"type":"signature_delta"`) {
+		t.Fatalf("expected signature_delta in stream output, got: %s", joined)
+	}
+	if !strings.Contains(joined, `"signature":"sig123"`) {
+		t.Fatalf("expected signature sig123 in stream output, got: %s", joined)
+	}
+	// Signature delta must be attached to the thinking content block (index 0 in this minimal stream).
+	if !strings.Contains(joined, `{"type":"content_block_delta","index":0,"delta":{"type":"signature_delta","signature":"sig123"}}`) {
+		t.Fatalf("expected signature_delta to target thinking block index 0, got: %s", joined)
+	}
+}
+
+func TestAntigravityClaudeNonStreamResponse_IncludesThinkingSignature(t *testing.T) {
+	raw := []byte(`{
+  "response":{
+    "responseId":"resp_1",
+    "modelVersion":"claude-sonnet-4-5-thinking",
+    "candidates":[{
+      "content":{"parts":[
+        {"text":"THOUGHT","thought":true},
+        {"thought":true,"thoughtSignature":"sig123"},
+        {"text":"ANSWER","thought":false}
+      ]},
+      "finishReason":"STOP"
+    }],
+    "usageMetadata":{"promptTokenCount":1,"candidatesTokenCount":1,"thoughtsTokenCount":1,"totalTokenCount":3}
+  }
+}`)
+
+	out := agclaude.ConvertAntigravityResponseToClaudeNonStream(context.Background(), "", nil, nil, raw, nil)
+	if !gjson.Valid(out) {
+		t.Fatalf("expected valid JSON output, got: %s", out)
+	}
+	content := gjson.Get(out, "content")
+	if !content.Exists() || !content.IsArray() {
+		t.Fatalf("expected content array in output, got: %s", out)
+	}
+
+	found := false
+	for _, block := range content.Array() {
+		if block.Get("type").String() != "thinking" {
+			continue
+		}
+		found = true
+		if got := block.Get("signature").String(); got != "sig123" {
+			t.Fatalf("expected thinking.signature=sig123, got %q (block=%s)", got, block.Raw)
+		}
+		if got := block.Get("thinking").String(); got != "THOUGHT" {
+			t.Fatalf("expected thinking.thinking=THOUGHT, got %q (block=%s)", got, block.Raw)
+		}
+	}
+	if !found {
+		t.Fatalf("expected a thinking block in output, got: %s", out)
+	}
+}

test/thinking_conversion_test.go

@@ -295,7 +295,7 @@ func TestThinkingConversionsAcrossProtocolsAndModels(t *testing.T) {
 									}
 									// Check numeric budget fallback for allowCompat
 									if budget, _, _, matched := util.ThinkingFromMetadata(metadata); matched && budget != nil {
-										if mapped, okMap := util.OpenAIThinkingBudgetToEffort(normalizedModel, *budget); okMap && mapped != "" {
+										if mapped, okMap := util.ThinkingBudgetToEffort(normalizedModel, *budget); okMap && mapped != "" {
 											return true, mapped, false
 										}
 									}
@@ -308,7 +308,7 @@ func TestThinkingConversionsAcrossProtocolsAndModels(t *testing.T) {
 								effort, ok := util.ReasoningEffortFromMetadata(metadata)
 								if !ok || strings.TrimSpace(effort) == "" {
 									if budget, _, _, matched := util.ThinkingFromMetadata(metadata); matched && budget != nil {
-										if mapped, okMap := util.OpenAIThinkingBudgetToEffort(normalizedModel, *budget); okMap {
+										if mapped, okMap := util.ThinkingBudgetToEffort(normalizedModel, *budget); okMap {
 											effort = mapped
 											ok = true
 										}
@@ -336,7 +336,7 @@ func TestThinkingConversionsAcrossProtocolsAndModels(t *testing.T) {
 									return false, "", true
 								}
 								if budget, _, _, matched := util.ThinkingFromMetadata(metadata); matched && budget != nil {
-									if mapped, okMap := util.OpenAIThinkingBudgetToEffort(normalizedModel, *budget); okMap && mapped != "" {
+									if mapped, okMap := util.ThinkingBudgetToEffort(normalizedModel, *budget); okMap && mapped != "" {
 										mapped = strings.ToLower(strings.TrimSpace(mapped))
 										if normalized, okLevel := util.NormalizeReasoningEffortLevel(normalizedModel, mapped); okLevel {
 											return true, normalized, false
@@ -609,7 +609,7 @@ func TestRawPayloadThinkingConversions(t *testing.T) {
 										return true, normalized, false
 									}
 									if budget, ok := cs.thinkingParam.(int); ok {
-										if mapped, okM := util.OpenAIThinkingBudgetToEffort(model, budget); okM && mapped != "" {
+										if mapped, okM := util.ThinkingBudgetToEffort(model, budget); okM && mapped != "" {
 											return true, mapped, false
 										}
 									}
@@ -625,7 +625,7 @@ func TestRawPayloadThinkingConversions(t *testing.T) {
 									return false, "", true // invalid level
 								}
 								if budget, ok := cs.thinkingParam.(int); ok {
-									if mapped, okM := util.OpenAIThinkingBudgetToEffort(model, budget); okM && mapped != "" {
+									if mapped, okM := util.ThinkingBudgetToEffort(model, budget); okM && mapped != "" {
 										// Check if the mapped effort is valid for this model
 										if _, validLevel := util.NormalizeReasoningEffortLevel(model, mapped); !validLevel {
 											return true, mapped, true // expect validation error
@@ -646,7 +646,7 @@ func TestRawPayloadThinkingConversions(t *testing.T) {
 									return false, "", true
 								}
 								if budget, ok := cs.thinkingParam.(int); ok {
-									if mapped, okM := util.OpenAIThinkingBudgetToEffort(model, budget); okM && mapped != "" {
+									if mapped, okM := util.ThinkingBudgetToEffort(model, budget); okM && mapped != "" {
 										// Check if the mapped effort is valid for this model
 										if _, validLevel := util.NormalizeReasoningEffortLevel(model, mapped); !validLevel {
 											return true, mapped, true // expect validation error
@@ -721,7 +721,7 @@ func TestRawPayloadThinkingConversions(t *testing.T) {
 	}
 }
 
-func TestOpenAIThinkingBudgetToEffortRanges(t *testing.T) {
+func TestThinkingBudgetToEffort(t *testing.T) {
 	cleanup := registerCoreModels(t)
 	defer cleanup()
 
@@ -733,7 +733,7 @@ func TestOpenAIThinkingBudgetToEffortRanges(t *testing.T) {
 		ok     bool
 	}{
 		{name: "dynamic-auto", model: "gpt-5", budget: -1, want: "auto", ok: true},
-		{name: "zero-none", model: "gpt-5", budget: 0, want: "none", ok: true},
+		{name: "zero-none", model: "gpt-5", budget: 0, want: "minimal", ok: true},
 		{name: "low-min", model: "gpt-5", budget: 1, want: "low", ok: true},
 		{name: "low-max", model: "gpt-5", budget: 1024, want: "low", ok: true},
 		{name: "medium-min", model: "gpt-5", budget: 1025, want: "medium", ok: true},
@@ -741,14 +741,14 @@ func TestOpenAIThinkingBudgetToEffortRanges(t *testing.T) {
 		{name: "high-min", model: "gpt-5", budget: 8193, want: "high", ok: true},
 		{name: "high-max", model: "gpt-5", budget: 24576, want: "high", ok: true},
 		{name: "over-max-clamps-to-highest", model: "gpt-5", budget: 64000, want: "high", ok: true},
-		{name: "over-max-xhigh-model", model: "gpt-5.2", budget: 50000, want: "xhigh", ok: true},
+		{name: "over-max-xhigh-model", model: "gpt-5.2", budget: 64000, want: "xhigh", ok: true},
 		{name: "negative-unsupported", model: "gpt-5", budget: -5, want: "", ok: false},
 	}
 
 	for _, cs := range cases {
 		cs := cs
 		t.Run(cs.name, func(t *testing.T) {
-			got, ok := util.OpenAIThinkingBudgetToEffort(cs.model, cs.budget)
+			got, ok := util.ThinkingBudgetToEffort(cs.model, cs.budget)
 			if ok != cs.ok {
 				t.Fatalf("ok mismatch for model=%s budget=%d: expect %v got %v", cs.model, cs.budget, cs.ok, ok)
 			}
@@ -758,3 +758,41 @@ func TestOpenAIThinkingBudgetToEffortRanges(t *testing.T) {
 		})
 	}
 }
+
+func TestThinkingEffortToBudget(t *testing.T) {
+	cleanup := registerCoreModels(t)
+	defer cleanup()
+
+	cases := []struct {
+		name   string
+		model  string
+		effort string
+		want   int
+		ok     bool
+	}{
+		{name: "none", model: "gemini-2.5-pro", effort: "none", want: 0, ok: true},
+		{name: "auto", model: "gemini-2.5-pro", effort: "auto", want: -1, ok: true},
+		{name: "minimal", model: "gemini-2.5-pro", effort: "minimal", want: 512, ok: true},
+		{name: "low", model: "gemini-2.5-pro", effort: "low", want: 1024, ok: true},
+		{name: "medium", model: "gemini-2.5-pro", effort: "medium", want: 8192, ok: true},
+		{name: "high", model: "gemini-2.5-pro", effort: "high", want: 24576, ok: true},
+		{name: "xhigh", model: "gemini-2.5-pro", effort: "xhigh", want: 32768, ok: true},
+		{name: "empty-unsupported", model: "gemini-2.5-pro", effort: "", want: 0, ok: false},
+		{name: "invalid-unsupported", model: "gemini-2.5-pro", effort: "ultra", want: 0, ok: false},
+		{name: "case-insensitive", model: "gemini-2.5-pro", effort: "LOW", want: 1024, ok: true},
+		{name: "case-insensitive-medium", model: "gemini-2.5-pro", effort: "MEDIUM", want: 8192, ok: true},
+	}
+
+	for _, cs := range cases {
+		cs := cs
+		t.Run(cs.name, func(t *testing.T) {
+			got, ok := util.ThinkingEffortToBudget(cs.model, cs.effort)
+			if ok != cs.ok {
+				t.Fatalf("ok mismatch for model=%s effort=%s: expect %v got %v", cs.model, cs.effort, cs.ok, ok)
+			}
+			if got != cs.want {
+				t.Fatalf("value mismatch for model=%s effort=%s: expect %d got %d", cs.model, cs.effort, cs.want, got)
+			}
+		})
+	}
+}