diff --git a/.env.example b/.env.example
new file mode 100644
index 0000000..13104a9
--- /dev/null
+++ b/.env.example
@@ -0,0 +1,35 @@
+# taiji-AI-PAD 环境变量配置模板
+# 复制此文件为 .env 并填写实际的密钥值
+# cp .env.example .env
+
+# ========== 数据库配置 ==========
+POSTGRES_DB=taiji_db
+POSTGRES_USER=taiji_user
+POSTGRES_PASSWORD=taiji_pass
+
+# ========== LiteLLM 网关配置 ==========
+LITELLM_MASTER_KEY=sk-taiji-master-key
+
+# ========== OpenRouter 配置 ==========
+OPENROUTER_API_KEY=your-openrouter-api-key-here
+OPENROUTER_BASE_URL=https://openrouter.ai/api/v1
+
+# ========== RapidAPI 配置 ==========
+RAPIDAPI_KEY=your-rapidapi-key-here
+RAPIDAPI_HOST=rapidapi.com
+
+# ========== OpenAI 配置（可选）==========
+# OPENAI_API_KEY=your-openai-api-key-here
+
+# ========== Anthropic 配置（可选）==========
+# ANTHROPIC_API_KEY=your-anthropic-api-key-here
+
+# ========== Langfuse 配置（可选，用于监控）==========
+# LANGFUSE_PUBLIC_KEY=your-langfuse-public-key
+# LANGFUSE_SECRET_KEY=your-langfuse-secret-key
+# LANGFUSE_HOST=https://cloud.langfuse.com
+
+# ========== 其他服务配置 ==========
+REDIS_URL=redis://redis:6379
+NATS_URL=nats://nats:4222
+DATABASE_URL=postgresql://taiji_user:taiji_pass@postgres:5432/taiji_db
diff --git a/Docs/环境变量配置说明.md b/Docs/环境变量配置说明.md
new file mode 100644
index 0000000..eb15f58
--- /dev/null
+++ b/Docs/环境变量配置说明.md
@@ -0,0 +1,166 @@
+# taiji-AI-PAD 环境变量配置说明
+
+## 📋 概述
+
+为了简化配置管理，所有 API Key 和敏感信息现在统一在 `.env` 文件中管理。您只需要在一个地方填写所有密钥，无需在多个配置文件中重复填写。
+
+## 🚀 快速开始
+
+### 1. 创建环境变量文件
+
+如果项目中没有 `.env` 文件，请复制模板：
+
+```bash
+cp .env.example .env
+```
+
+### 2. 编辑 `.env` 文件
+
+打开 `.env` 文件，填写您的实际 API Key：
+
+```bash
+nano .env
+# 或使用您喜欢的编辑器
+```
+
+### 3. 配置项说明
+
+#### 必需配置项
+
+| 配置项 | 说明 | 示例 |
+|--------|------|------|
+| `OPENROUTER_API_KEY` | OpenRouter API 密钥 | `sk-or-v1-...` |
+| `RAPIDAPI_KEY` | RapidAPI 密钥 | `33902cc39dmsh...` |
+| `LITELLM_MASTER_KEY` | LiteLLM 主密钥 | `sk-taiji-master-key` |
+
+#### 可选配置项
+
+| 配置项 | 说明 | 何时需要 |
+|--------|------|----------|
+| `OPENAI_API_KEY` | OpenAI API 密钥 | 使用 OpenAI 模型时 |
+| `ANTHROPIC_API_KEY` | Anthropic API 密钥 | 使用 Claude 模型时 |
+| `LANGFUSE_*` | Langfuse 监控配置 | 启用监控功能时 |
+
+## 📝 当前配置的密钥位置
+
+### ✅ 已统一管理的密钥
+
+以下密钥现在都在 `.env` 文件中：
+
+1. **OpenRouter API Key** - 用于模型网关
+2. **RapidAPI Key** - 用于数据接入服务
+3. **LiteLLM Master Key** - 用于模型网关认证
+
+### 📍 配置文件位置
+
+- **`.env`** - 实际环境变量文件（包含真实密钥，已加入 .gitignore）
+- **`.env.example`** - 配置模板文件（可提交到 Git）
+- **`docker-compose.yml`** - 使用 `${VAR}` 语法引用环境变量
+
+## 🔧 如何添加新的 API Key
+
+### 步骤 1: 在 `.env` 文件中添加
+
+```bash
+# 在 .env 文件中添加
+NEW_API_KEY=your-new-api-key-here
+```
+
+### 步骤 2: 在 `docker-compose.yml` 中引用
+
+```yaml
+services:
+  your-service:
+    environment:
+      - NEW_API_KEY=${NEW_API_KEY}
+```
+
+### 步骤 3: 在代码中读取
+
+```python
+import os
+api_key = os.getenv("NEW_API_KEY", "")
+```
+
+## 🔒 安全注意事项
+
+1. **⚠️ 永远不要提交 `.env` 文件到 Git**
+   - `.env` 文件已在 `.gitignore` 中
+   - 只提交 `.env.example` 作为模板
+
+2. **生产环境建议**
+   - 使用密钥管理服务（如 AWS Secrets Manager）
+   - 使用环境变量注入（如 Kubernetes Secrets）
+   - 定期轮换 API Key
+
+3. **权限控制**
+   - 确保 `.env` 文件权限为 `600`（仅所有者可读写）
+   ```bash
+   chmod 600 .env
+   ```
+
+## 📊 配置项清单
+
+### 当前已配置的密钥
+
+- ✅ OpenRouter API Key
+- ✅ RapidAPI Key
+- ✅ LiteLLM Master Key
+
+### 可选配置的密钥
+
+- ⚪ OpenAI API Key（如需要直接使用 OpenAI）
+- ⚪ Anthropic API Key（如需要直接使用 Anthropic）
+- ⚪ Langfuse 监控密钥（如需要启用监控）
+
+## 🧪 验证配置
+
+配置完成后，验证环境变量是否正确加载：
+
+```bash
+# 检查环境变量
+docker-compose config | grep -E "OPENROUTER|RAPIDAPI|LITELLM"
+
+# 重启服务以应用新配置
+docker-compose restart litellm-gateway data-ingestion
+
+# 检查服务健康状态
+curl http://localhost:4000/health  # LiteLLM Gateway
+curl http://localhost:8001/health  # Data Ingestion
+```
+
+## 📚 相关文档
+
+- [Docker Compose 环境变量文档](https://docs.docker.com/compose/environment-variables/)
+- [项目 README](../README.md)
+- [测试准备说明](./测试准备说明.md)
+
+## ❓ 常见问题
+
+### Q: 为什么需要 `.env` 文件？
+
+A: `.env` 文件可以：
+- 集中管理所有密钥
+- 避免在代码中硬编码敏感信息
+- 方便不同环境使用不同配置
+- 提高安全性（不提交到 Git）
+
+### Q: 如何在不同环境使用不同配置？
+
+A: 可以创建多个环境文件：
+- `.env.development` - 开发环境
+- `.env.production` - 生产环境
+- `.env.testing` - 测试环境
+
+然后使用：
+```bash
+docker-compose --env-file .env.production up
+```
+
+### Q: 忘记填写某个 Key 会怎样？
+
+A: 如果某个环境变量未设置，Docker Compose 会使用空字符串或默认值。服务可能会启动失败或功能受限。请检查服务日志：
+```bash
+docker-compose logs service-name
+```
+
diff --git a/docker-compose.yml b/docker-compose.yml
index c2613a5..f6ae889 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -1,5 +1,8 @@
 version: '3.8'
 
+# Docker Compose 会自动从 .env 文件读取环境变量
+# 所有 ${VAR} 形式的变量都会从 .env 文件中获取
+
 services:
   # 数据库服务
   postgres:
@@ -54,9 +57,12 @@ services:
     ports:
       - "4000:4000"
     environment:
-      - LITELLM_MASTER_KEY=sk-taiji-master-key
-      - DATABASE_URL=postgresql://taiji_user:taiji_pass@postgres:5432/taiji_db
-      - REDIS_URL=redis://redis:6379
+      - LITELLM_MASTER_KEY=${LITELLM_MASTER_KEY}
+      - DATABASE_URL=${DATABASE_URL}
+      - REDIS_URL=${REDIS_URL}
+      # OpenRouter 配置
+      - OPENROUTER_API_KEY=${OPENROUTER_API_KEY}
+      - OPENROUTER_BASE_URL=${OPENROUTER_BASE_URL}
     volumes:
       - ./services/model-gateway/config:/app/config
       - ./logs:/app/logs
@@ -76,9 +82,12 @@ services:
     ports:
       - "8001:8000"
     environment:
-      - DATABASE_URL=postgresql://taiji_user:taiji_pass@postgres:5432/taiji_db
-      - REDIS_URL=redis://redis:6379
-      - NATS_URL=nats://nats:4222
+      - DATABASE_URL=${DATABASE_URL}
+      - REDIS_URL=${REDIS_URL}
+      - NATS_URL=${NATS_URL}
+      # RapidAPI 配置
+      - RAPIDAPI_KEY=${RAPIDAPI_KEY}
+      - RAPIDAPI_HOST=${RAPIDAPI_HOST}
     volumes:
       - ./services/data-ingestion:/app
       - ./logs:/app/logs
diff --git a/services/model-gateway/Dockerfile b/services/model-gateway/Dockerfile
index cc51546..9db1760 100644
--- a/services/model-gateway/Dockerfile
+++ b/services/model-gateway/Dockerfile
@@ -19,11 +19,15 @@ WORKDIR /app
         npm config set registry https://registry.npmmirror.com
 
     # 安装LiteLLM和prisma (使用国内源)
+    # 注意：Prisma Python 客户端 0.12.0 需要 Prisma CLI 5.8.0
+    # 需要同时安装 Prisma CLI 和 Python 客户端
     RUN pip install --no-cache-dir -i https://pypi.tuna.tsinghua.edu.cn/simple \
         litellm[proxy]==1.17.0 \
         redis==5.0.1 \
-        prometheus-client==0.19.0 && \
-        npm install -g prisma
+        prometheus-client==0.19.0 \
+        prisma==0.12.0 && \
+        (rm -f /usr/local/bin/prisma || true) && \
+        npm install -g prisma@5.8.0
 
 # 复制配置文件
 COPY config/ ./config/
diff --git a/services/model-gateway/config/litellm.yaml b/services/model-gateway/config/litellm.yaml
index 5682546..29f80c5 100644
--- a/services/model-gateway/config/litellm.yaml
+++ b/services/model-gateway/config/litellm.yaml
@@ -136,6 +136,71 @@ model_list:
       input_cost_per_token: 0.0  # 本地模型无成本
       output_cost_per_token: 0.0
 
+  # OpenRouter 模型组 - 通过 OpenRouter 访问多种模型
+  - model_name: "openrouter-gpt-4"
+    litellm_params:
+      model: "openrouter/openai/gpt-4"
+      api_key: "os.environ/OPENROUTER_API_KEY"
+      api_base: "os.environ/OPENROUTER_BASE_URL"
+      max_tokens: 8000
+      temperature: 0.7
+    model_info:
+      mode: "chat"
+      supports_function_calling: true
+      supports_vision: false
+      max_input_tokens: 8192
+      max_output_tokens: 8192
+      input_cost_per_token: 0.00003
+      output_cost_per_token: 0.00006
+      
+  - model_name: "openrouter-gpt-3.5-turbo"
+    litellm_params:
+      model: "openrouter/openai/gpt-3.5-turbo"
+      api_key: "os.environ/OPENROUTER_API_KEY"
+      api_base: "os.environ/OPENROUTER_BASE_URL"
+      max_tokens: 4000
+      temperature: 0.7
+    model_info:
+      mode: "chat"
+      supports_function_calling: true
+      supports_vision: false
+      max_input_tokens: 16385
+      max_output_tokens: 4096
+      input_cost_per_token: 0.0000015
+      output_cost_per_token: 0.000002
+
+  - model_name: "openrouter-claude-3.5-sonnet"
+    litellm_params:
+      model: "openrouter/anthropic/claude-3.5-sonnet"
+      api_key: "os.environ/OPENROUTER_API_KEY"
+      api_base: "os.environ/OPENROUTER_BASE_URL"
+      max_tokens: 4000
+      temperature: 0.7
+    model_info:
+      mode: "chat"
+      supports_function_calling: true
+      supports_vision: true
+      max_input_tokens: 200000
+      max_output_tokens: 4096
+      input_cost_per_token: 0.000003
+      output_cost_per_token: 0.000015
+
+  - model_name: "openrouter-claude-3-opus"
+    litellm_params:
+      model: "openrouter/anthropic/claude-3-opus"
+      api_key: "os.environ/OPENROUTER_API_KEY"
+      api_base: "os.environ/OPENROUTER_BASE_URL"
+      max_tokens: 4000
+      temperature: 0.7
+    model_info:
+      mode: "chat"
+      supports_function_calling: true
+      supports_vision: true
+      max_input_tokens: 200000
+      max_output_tokens: 4096
+      input_cost_per_token: 0.000015
+      output_cost_per_token: 0.000075
+
 # 路由器配置
 router_settings:
   routing_strategy: "least-busy"  # 路由策略: least-busy, round-robin, latency-based
@@ -183,13 +248,24 @@ router_settings:
           weight: 0.3
         - model_name: "claude-3-sonnet"
           weight: 0.3
+      
+    - group_name: "openrouter-group"
+      models:
+        - model_name: "openrouter-gpt-4"
+          weight: 0.3
+        - model_name: "openrouter-gpt-3.5-turbo"
+          weight: 0.3
+        - model_name: "openrouter-claude-3.5-sonnet"
+          weight: 0.25
+        - model_name: "openrouter-claude-3-opus"
+          weight: 0.15
 
 # 用户和权限配置
 litellm_settings:
   # API密钥管理
   api_keys:
     - key: "sk-taiji-mcp-server"
-      models: ["gpt-3.5-turbo", "gpt-4", "claude-3-haiku", "claude-3-sonnet"]
+      models: ["gpt-3.5-turbo", "gpt-4", "claude-3-haiku", "claude-3-sonnet", "openrouter-gpt-4", "openrouter-gpt-3.5-turbo", "openrouter-claude-3.5-sonnet"]
       max_budget: 100.0
       budget_duration: "1d"
       metadata:
@@ -197,7 +273,7 @@ litellm_settings:
         service: "mcp-server"
         
     - key: "sk-taiji-data-ingestion"
-      models: ["gpt-3.5-turbo", "claude-3-haiku", "llama-3-8b"]
+      models: ["gpt-3.5-turbo", "claude-3-haiku", "llama-3-8b", "openrouter-gpt-3.5-turbo", "openrouter-claude-3.5-sonnet"]
       max_budget: 50.0
       budget_duration: "1d"
       metadata:
@@ -205,7 +281,7 @@ litellm_settings:
         service: "data-ingestion"
         
     - key: "sk-taiji-agent-dev"
-      models: ["gpt-3.5-group", "claude-group", "fast-models"]
+      models: ["gpt-3.5-group", "claude-group", "fast-models", "openrouter-group"]
       max_budget: 20.0
       budget_duration: "1d"
       metadata:
@@ -213,7 +289,7 @@ litellm_settings:
         service: "agent-development"
         
     - key: "sk-taiji-premium"
-      models: ["premium-models", "gpt-4-group"]
+      models: ["premium-models", "gpt-4-group", "openrouter-group", "openrouter-gpt-4", "openrouter-claude-3-opus", "openrouter-claude-3.5-sonnet"]
       max_budget: 200.0
       budget_duration: "1d"
       metadata:
@@ -278,9 +354,13 @@ error_handling:
   fallback:
     enabled: true
     fallback_models:
-      "gpt-4": ["gpt-4-turbo", "claude-3-sonnet"]
-      "claude-3-opus": ["claude-3-sonnet", "gpt-4"]
-      "gpt-3.5-turbo": ["claude-3-haiku", "llama-3-8b"]
+      "gpt-4": ["openrouter-gpt-4", "gpt-4-turbo", "claude-3-sonnet", "openrouter-claude-3-sonnet"]
+      "claude-3-opus": ["openrouter-claude-3-opus", "claude-3-sonnet", "openrouter-claude-3-sonnet", "gpt-4", "openrouter-gpt-4"]
+      "gpt-3.5-turbo": ["openrouter-gpt-3.5-turbo", "claude-3-haiku", "llama-3-8b"]
+      "openrouter-gpt-4": ["gpt-4", "gpt-4-turbo", "openrouter-claude-3.5-sonnet"]
+      "openrouter-gpt-3.5-turbo": ["gpt-3.5-turbo", "claude-3-haiku"]
+      "openrouter-claude-3.5-sonnet": ["claude-3-sonnet", "gpt-4", "openrouter-gpt-4"]
+      "openrouter-claude-3-opus": ["claude-3-opus", "claude-3-sonnet", "openrouter-claude-3.5-sonnet"]
 
 # 日志配置
 logging:
diff --git a/services/model-gateway/config/litellm_simple.yaml b/services/model-gateway/config/litellm_simple.yaml
index b3d5c01..765bbbf 100644
--- a/services/model-gateway/config/litellm_simple.yaml
+++ b/services/model-gateway/config/litellm_simple.yaml
@@ -13,7 +13,7 @@ general_settings:
   # 安全设置
   allowed_ips: ["127.0.0.1", "172.20.0.0/16", "0.0.0.0/0"]  # 允许所有IP用于测试
 
-# 模型配置 - 使用测试模型
+# 模型配置 - 使用测试模型和 OpenRouter
 model_list:
   # 使用HuggingFace的免费模型进行测试
   - model_name: "test-model"
@@ -25,6 +25,43 @@ model_list:
       supports_function_calling: false
       supports_vision: false
 
+  # OpenRouter 模型 - 通过 OpenRouter 访问
+  - model_name: "openrouter-gpt-4o-mini"
+    litellm_params:
+      model: "openrouter/openai/gpt-4o-mini"
+      api_key: "os.environ/OPENROUTER_API_KEY"
+      api_base: "os.environ/OPENROUTER_BASE_URL"
+      max_tokens: 1000
+      temperature: 0.7
+    model_info:
+      mode: "chat"
+      supports_function_calling: true
+      supports_vision: false
+
+  - model_name: "openrouter-gpt-3.5-turbo"
+    litellm_params:
+      model: "openrouter/openai/gpt-3.5-turbo"
+      api_key: "os.environ/OPENROUTER_API_KEY"
+      api_base: "os.environ/OPENROUTER_BASE_URL"
+      max_tokens: 1000
+      temperature: 0.7
+    model_info:
+      mode: "chat"
+      supports_function_calling: true
+      supports_vision: false
+
+  - model_name: "openrouter-claude-3.5-sonnet"
+    litellm_params:
+      model: "openrouter/anthropic/claude-3.5-sonnet"
+      api_key: "os.environ/OPENROUTER_API_KEY"
+      api_base: "os.environ/OPENROUTER_BASE_URL"
+      max_tokens: 4000
+      temperature: 0.7
+    model_info:
+      mode: "chat"
+      supports_function_calling: true
+      supports_vision: true
+
 # 路由器配置
 router_settings:
   routing_strategy: "round-robin"
@@ -35,7 +72,7 @@ router_settings:
 litellm_settings:
   api_keys:
     - key: "sk-test-key"
-      models: ["test-model"]
+      models: ["test-model", "openrouter-gpt-4o-mini", "openrouter-gpt-3.5-turbo", "openrouter-claude-3.5-sonnet"]
       metadata:
         user_id: "test-user"
         service: "testing"