Support private deployment for OpenGuardrails Dify Plugin

Author: ThomasLWang

README.md

@@ -1,7 +1,7 @@
 # OpenGuardrails
 
 **Author:** openguardrails  
-**Version:** 0.0.3
+**Version:** 0.0.4
 **Type:** tool  
 
 ## Description  
@@ -17,7 +17,7 @@ OpenGuardrails focuses on:
 - **Content Safety & Compliance** (based on **Rich risk categories**) 
 - **Sensitive Data Leakage Prevention** PII, Business Secrets
 - **Contextual Semantic Understanding** for precise detection  
-- **thomas Protection** with dual deployment modes:  
+- **Runtime Protection** with dual deployment modes:  
   - **API Detection Mode** – flexible integration and precise control  
   - **Security Gateway Mode** – WAF-like transparent proxy, zero code change   
 
@@ -112,12 +112,13 @@ score:
 To use OpenGuardrails, you need an **API Key**:  
 
 1. Register at the OpenGuardrails Platform: [https://OpenGuardrails.com/platform/](https://OpenGuardrails.com/platform/)  
-2. Log in and go to **Account Management** → Get your API Key  
+2. Log in and go to **Application Management** → Get your API Key for free
+![API Key](_assets/account.png)  
 3. Add the API Key to your Dify plugin configuration  
 
 ## Example Usage
 
-Use OpenGuardrails’ **check_prompt** and **check_response_ctx** tools to protect the input and output of large language models.  
+Use OpenGuardrails’ **OGR Check Prompt** and **OGR Check Response Contextual** tools to protect the input and output of large language models.  
 
 ![workflow](_assets/workflow.png)  
 

_assets/account.jpg

@@ -1,4 +1,14 @@
 from dify_plugin import Plugin, DifyPluginEnv
+import logging
+
+# Configure logging to show detailed debug information
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
+    handlers=[
+        logging.StreamHandler()
+    ]
+)
 
 plugin = Plugin(DifyPluginEnv(MAX_REQUEST_TIMEOUT=120))
 

_assets/account.png

@@ -1,11 +1,11 @@
-version: 0.0.3
+version: 0.0.4
 type: plugin
 author: openguardrails
 name: openguardrails
 label:
-  en_US: OpenGuardrails - Open Source, Free, Contextual
+  en_US: OpenGuardrails - Free Open-source Contextual
 description:
-  en_US: Open source and free AI guardrails for contextually protecting against prompt attacks, content safety and sensitive data leakage. 
+  en_US: Free Open-source AI guardrails for contextually protecting against prompt attacks, content safety and sensitive data leakage. 
 icon: icon.png
 icon_dark: icon-dark.png
 resource:

_assets/workflow.png

@@ -1,31 +1,42 @@
 from typing import Any
+import logging
 
 from dify_plugin import ToolProvider
 from dify_plugin.errors.tool import ToolProviderCredentialValidationError
 from openguardrails import OpenGuardrails
 
+logger = logging.getLogger(__name__)
+
 
 class OpenGuardrailsProvider(ToolProvider):
 
     def _validate_credentials(self, credentials: dict[str, Any]) -> None:
         try:
-            # 获取API密钥
+            # Get API key
             api_key = credentials.get("api_key")
             if not api_key:
                 raise ToolProviderCredentialValidationError("API key is required")
 
-            # 验证API密钥有效性，通过一个简单的测试调用
-            client = OpenGuardrails(api_key)
-            # 使用一个简单的测试来验证API密钥
+            # Get base_url, if not provided, use default value
+            base_url = credentials.get("base_url")
+
+            if base_url:
+                client = OpenGuardrails(api_key, base_url=base_url)
+            else:
+                client = OpenGuardrails(api_key)
+
+            # Use a simple test to validate API key
             test_result = client.check_prompt("test")
 
-            # 如果调用成功并返回了结果，则验证通过
+            # If the call is successful and returns a result, the validation is successful
             if hasattr(test_result, 'suggest_action'):
-                return  # 验证成功
+                return  # Validate Successfully
             else:
+                logger.error(f"Invalid response format, missing suggest_action attribute")
                 raise ToolProviderCredentialValidationError("Invalid API key response format")
 
         except Exception as e:
+            logger.error(f"Credential validation error: {str(e)}", exc_info=True)
             if "API key" in str(e).lower() or "auth" in str(e).lower():
                 raise ToolProviderCredentialValidationError("Invalid API key")
             else:

main.py

@@ -2,9 +2,9 @@ identity:
   author: "openguardrails"
   name: "openguardrails"
   label:
-    en_US: "OpenGuardrails - Open Source, Free, Contextual"
+    en_US: "OpenGuardrails - Free Open-source Contextual"
   description:
-    en_US: "Open source and free contextual AI guardrails for protecting against prompt attacks, content safety and sensitive data leakage."
+    en_US: "Free Open source contextual AI guardrails for protecting against prompt attacks, content safety and sensitive data leakage."
   icon: "icon.png"
   icon_dark: "icon-dark.png"
 
@@ -19,6 +19,16 @@ credentials_for_provider:
     help:
       en_US: "Get your free API key from OpenGuardrails management console: https://OpenGuardrails.com/platform/"
     url: https://OpenGuardrails.com/platform/
+  base_url:
+    type: text-input
+    required: false
+    label:
+      en_US: Base URL
+    placeholder:
+      en_US: Leave empty to use the free cloud API (https://api.openguardrails.com/v1).
+    help:
+      en_US: "If you have a private deployment, enter your private OpenGuardrails server URL (e.g., http://localhost:5001/v1)."
+    url: https://OpenGuardrails.com/platform/
 
 tools:
   - tools/check_prompt.yaml

manifest.yaml

@@ -1,10 +1,13 @@
 from collections.abc import Generator
 from typing import Any
+import logging
 
 from dify_plugin import Tool
 from dify_plugin.entities.tool import ToolInvokeMessage
 from openguardrails import OpenGuardrails
 
+logger = logging.getLogger(__name__)
+
 class CheckPromptTool(Tool):
     def _invoke(self, tool_parameters: dict[str, Any]) -> Generator[ToolInvokeMessage]:
         try:
@@ -24,39 +27,57 @@ def _invoke(self, tool_parameters: dict[str, Any]) -> Generator[ToolInvokeMessag
                 yield self.create_text_message("Error: API key is required.")
                 return
 
+            # Get base_url from credentials
+            base_url = self.runtime.credentials.get("base_url")
+
             # Create OpenGuardrails client and check user input
-            client = OpenGuardrails(api_key)
-            result = client.check_prompt(prompt, user_id=user_id)
+            try:
+                if base_url:
+                    client = OpenGuardrails(api_key, base_url=base_url)
+                else:
+                    client = OpenGuardrails(api_key)
+
+                result = client.check_prompt(prompt, user_id=user_id)
+            except Exception as client_error:
+                logger.error(f"OpenGuardrails client error: {str(client_error)}", exc_info=True)
+                raise
 
             # Extract categories field: from compliance and security not equal to "no_risk" categories list first item
-            categories = []
-            if result.result.compliance.risk_level != "no_risk" and result.result.compliance.categories:
-                categories.append(result.result.compliance.categories[0])
-            elif result.result.security.risk_level != "no_risk" and result.result.security.categories:
-                categories.append(result.result.security.categories[0])
-            elif result.result.data.risk_level != "no_risk" and result.result.data.categories:
-                categories.append(result.result.data.categories[0])
-            
-            categories_str = ", ".join(categories)
-            if categories_str:
-                categories_str = f"{categories_str}"
-            else:
-                categories_str = ""
-
-            # Process suggest_answer field, if not exist, set to empty string
-            suggest_answer = ""
-            if result.suggest_answer:
-                suggest_answer = result.suggest_answer
-
-            # Use custom variable to return result
-            yield self.create_variable_message("id", result.id)
-            yield self.create_variable_message("overall_risk_level", result.overall_risk_level)
-            yield self.create_variable_message("suggest_action", result.suggest_action)
-            yield self.create_variable_message("suggest_answer", suggest_answer)
-            yield self.create_variable_message("categories", categories_str)
-            yield self.create_variable_message("score", result.score)
+            try:
+                categories = []
+                if result.result.compliance.risk_level != "no_risk" and result.result.compliance.categories:
+                    categories.append(result.result.compliance.categories[0])
+                elif result.result.security.risk_level != "no_risk" and result.result.security.categories:
+                    categories.append(result.result.security.categories[0])
+                elif result.result.data.risk_level != "no_risk" and result.result.data.categories:
+                    categories.append(result.result.data.categories[0])
+
+                categories_str = ", ".join(categories)
+                if categories_str:
+                    categories_str = f"{categories_str}"
+                else:
+                    categories_str = ""
+
+                # Process suggest_answer field, if not exist, set to empty string
+                suggest_answer = ""
+                if result.suggest_answer:
+                    suggest_answer = result.suggest_answer
+
+                # Use custom variable to return result
+                yield self.create_variable_message("id", result.id)
+                yield self.create_variable_message("overall_risk_level", result.overall_risk_level)
+                yield self.create_variable_message("suggest_action", result.suggest_action)
+                yield self.create_variable_message("suggest_answer", suggest_answer)
+                yield self.create_variable_message("categories", categories_str)
+                # Ensure score is a number, not None
+                score_value = result.score if result.score is not None else 0.0
+                yield self.create_variable_message("score", score_value)
+            except Exception as result_error:
+                logger.error(f"Error processing result: {str(result_error)}", exc_info=True)
+                raise
 
         except Exception as e:
             # 错误处理
+            logger.error(f"CheckPromptTool error: {str(e)}", exc_info=True)
             yield self.create_text_message(f"Error: {str(e)}")
             yield self.create_json_message({"error": str(e)})

provider/openguardrails.py

@@ -2,7 +2,7 @@ identity:
   name: "check_prompt"
   author: "openguardrails"
   label:
-    en_US: "OpenGuardrails - Check Prompt"
+    en_US: "OGR Check Prompt"
 description:
   human:
     en_US: "Detect user input for prompt attacks, jailbreaks, malicious operations and content safety issues based on OWASP TOP 10 LLM Applications and GB/T45654-2025 standards"