@moorchehai/mcp 1.3.2 → 1.3.4

package/src/server/utils/prompts.js

@@ -1,359 +1,359 @@
-import { z } from 'zod';
-
-// Search optimization prompt
-export const searchOptimizationPrompt = {
-  name: 'search-optimization',
-  description: 'Tips for optimizing search queries in Moorcheh',
-  argsSchema: {
-    search_type: z.enum(['text','vector']).optional().describe('Type of search (text or vector)'),
-    domain: z.string().optional().describe('Domain or topic area of your content')
-  },
-  handler: async (args) => {
-    const searchType = args?.search_type || 'text';
-    const domain = args?.domain || 'general';
-    
-    const text = `# Moorcheh Search Optimization Guide
-
-## Optimizing ${searchType} search for ${domain} content
-
-### Request Format Examples:
-
-**Text Search Request:**
-\`\`\`json
-{
-  "query": "your search text here",
-  "namespaces": ["your-namespace"],
-  "top_k": 10,
-  "kiosk_mode": false
-}
-\`\`\`
-
-**Vector Search Request:**
-\`\`\`json
-{
-  "query": [0.1, 0.2, 0.3, ..., 0.768],
-  "namespaces": ["vector-embeddings"],
-  "top_k": 5,
-  "kiosk_mode": true,
-  "threshold": 0.1
-}
-\`\`\`
-
-**cURL Example for Vector Search:**
-\`\`\`bash
-curl -X POST "https://api.moorcheh.ai/v1/search" \\
-  -H "Content-Type: application/json" \\
-  -H "x-api-key: your-api-key-here" \\
-  -d '{
-    "query": [0.1, 0.2, 0.3, ..., 0.768],
-    "namespaces": ["vector-embeddings"],
-    "top_k": 5,
-    "kiosk_mode": true,
-    "threshold": 0.1
-  }'
-\`\`\`
-
-### Search Strategy:
-
-${searchType === 'text' ? `
-**Text Search Best Practices:**
-- Use specific, descriptive terms rather than generic ones
-- Include key domain terminology from your ${domain} field
-- Try both short keywords and longer phrases
-- Use synonyms if initial searches don't return good results
-- Natural language queries work well for semantic search
-
-**Query Examples:**
-- Instead of: "help"
-- Try: "troubleshooting guide", "how to resolve", "step by step"
-- Use complete sentences: "How do I configure authentication?"
-- Include context: "Python authentication setup for web applications"
-` : `
-**Vector Search Best Practices:**
-- Ensure your query embeddings use the same model as your stored vectors
-- Vector search works well with semantic similarity
-- Natural language queries often work better than keywords
-- Consider the context and intent behind your search
-- Match vector dimensions with your namespace configuration
-
-**Query Optimization:**
-- Use complete sentences or questions for better semantic matching
-- Include relevant context in your query
-- Test with variations of the same concept
-- Ensure vector dimensions match your namespace (e.g., 768 for many models)
-`}
-
-### Parameter Tuning:
-
-1. **top_k**: Number of results to return (default: 10)
-   - Start with 5-10 for most use cases
-   - Increase for broader exploration
-   - Decrease for highly targeted results
-
-2. **threshold**: Similarity threshold (0-1, optional)
-   - 0.7-0.8: High similarity, fewer but more relevant results
-   - 0.5-0.7: Moderate similarity, balanced results
-   - 0.3-0.5: Lower similarity, more comprehensive results
-   - **Important**: When kiosk_mode is true, threshold becomes mandatory
-
-3. **kiosk_mode**: Boolean (default: false)
-   - true: Restrict search to specific namespace(s) with threshold filtering
-   - false: Search across all specified namespaces
-   - **Note**: When kiosk_mode is enabled, threshold parameter is required
-
-### Kiosk Mode Requirements:
-
-When using kiosk_mode=true:
-- **threshold parameter is mandatory**
-- Results are filtered by the specified threshold
-- Useful for production environments with strict relevance requirements
-- Example: kiosk_mode=true with threshold=0.7 ensures only high-confidence results
-
-### Domain-Specific Tips for ${domain}:
-- Use terminology specific to ${domain}
-- Consider the typical language patterns in your content
-- Test with actual user queries from your ${domain} context
-- Monitor search performance and adjust parameters accordingly
-- For technical content, include specific technology names and versions
-
-### Troubleshooting:
-- If no results: Lower threshold, check spelling, try synonyms
-- If too many irrelevant results: Raise threshold, use more specific terms
-- If results seem random: Check your embeddings model compatibility
-- For vector search: Verify vector dimensions match namespace configuration
-- For kiosk mode: Ensure threshold is provided when kiosk_mode=true
-
-### Response Format:
-The search returns results with:
-- **id**: Document identifier
-- **score**: Similarity score (0-1, higher is better)
-- **label**: Relevance label (Close Match, Very High Relevance, etc.)
-- **text**: Original text content (for text namespaces)
-- **metadata**: Associated metadata
-`;
-
-    return {
-      messages: [
-        {
-          role: 'assistant',
-          content: { type: 'text', text }
-        }
-      ]
-    };
-  }
-};
-
-// Data organization prompt
-export const dataOrganizationPrompt = {
-  name: 'data-organization',
-  description: 'Best practices for organizing data in Moorcheh namespaces',
-  argsSchema: {
-    content_type: z.string().optional().describe("Type of content you're organizing"),
-    team_size: z.enum(['small','large']).optional().describe('Size of your team using this data')
-  },
-  handler: async (args) => {
-    const contentType = args?.content_type || 'documents';
-    const teamSize = args?.team_size || 'small';
-    
-    const text = `# Moorcheh Data Organization Guide
-
-## Organizing ${contentType} for a ${teamSize} team
-
-### Namespace Strategy:
-
-${teamSize === 'large' ? `
-**Large Team Considerations:**
-- Create separate namespaces by department or project
-- Use consistent naming conventions across teams
-- Implement clear metadata standards
-- Consider access control and permissions
-` : `
-**Small Team Approach:**
-- Fewer namespaces with more content per namespace
-- Flexible organization that can evolve
-- Focus on clear naming and good metadata
-- Regular cleanup and maintenance
-`}
-
-### Content Organization for ${contentType}:
-
-1. **Metadata Schema:**
-   \`\`\`json
-   {
-     "category": "primary classification",
-     "tags": ["tag1", "tag2", "tag3"],
-     "author": "content creator",
-     "created_date": "2024-01-01",
-     "department": "relevant team",
-     "priority": "high/medium/low",
-     "status": "draft/review/published"
-   }
-   \`\`\`
-
-2. **Document ID Conventions:**
-   - Use descriptive, hierarchical IDs
-   - Include date/version when relevant
-   - Examples: "support-faq-login-2024", "product-spec-v2-auth"
-
-3. **Content Chunking:**
-   - Break large documents into logical sections
-   - Each chunk should be self-contained
-   - Include context in metadata
-
-### Maintenance Workflow:
-
-1. **Regular Reviews:**
-   - Monthly metadata cleanup
-   - Remove outdated content
-   - Update tags and categories
-
-2. **Quality Control:**
-   - Standardize formatting before upload
-   - Validate metadata completeness
-   - Test search functionality after major updates
-
-3. **Growth Planning:**
-   - Monitor namespace size and performance
-   - Plan for content archival strategy
-   - Consider splitting namespaces as they grow
-
-### Search-Friendly Organization:
-- Use consistent terminology in your ${contentType}
-- Include alternative phrasings in metadata
-- Create logical content hierarchies
-- Tag content with multiple relevant categories
-
-This organization will help your ${teamSize} team efficiently manage and search through your ${contentType}.
-`;
-
-    return {
-      messages: [
-        {
-          role: 'assistant',
-          content: { type: 'text', text }
-        }
-      ]
-    };
-  }
-};
-
-// AI answer setup prompt
-export const aiAnswerSetupPrompt = {
-  name: 'ai-answer-setup',
-  description: 'Guide for configuring AI-powered answers in Moorcheh',
-  argsSchema: {
-    answer_style: z.enum(['concise','detailed','technical','friendly','balanced']).optional().describe('Desired style for AI answers'),
-    context_type: z.string().optional().describe('Type of context/domain for answers')
-  },
-  handler: async (args) => {
-    const answerStyle = args?.answer_style || 'balanced';
-    const contextType = args?.context_type || 'general';
-    
-    const text = `# Moorcheh AI Answer Configuration Guide
-
-## Setting up ${answerStyle} AI answers for ${contextType} content
-
-### Header Prompt Configuration:
-
-\`\`\`
-You are an AI assistant specialized in ${contextType}. 
-Your role is to provide ${answerStyle} answers based on the provided context.
-
-Style Guidelines:
-${answerStyle === 'concise' ? `
-- Keep answers brief and to the point
-- Use bullet points for multiple items
-- Avoid unnecessary explanations
-- Focus on actionable information
-` : answerStyle === 'detailed' ? `
-- Provide comprehensive explanations
-- Include relevant background information
-- Use examples where helpful
-- Structure answers with clear sections
-` : answerStyle === 'technical' ? `
-- Use precise technical terminology
-- Include code examples when relevant
-- Explain complex concepts step-by-step
-- Reference specific technical details
-` : answerStyle === 'friendly' ? `
-- Use conversational, approachable language
-- Include encouraging phrases
-- Explain concepts in simple terms
-- Ask clarifying questions when needed
-` : `
-- Balance brevity with completeness
-- Use clear, professional language
-- Structure information logically
-- Include key details without overwhelming
-`}
-
-Context: You have access to ${contextType} documentation and resources.
-\`\`\`
-
-### Footer Prompt Configuration:
-
-\`\`\`
-Additional Guidelines:
-- Always cite specific sources when possible
-- If information is incomplete, acknowledge limitations
-- Suggest follow-up questions for complex topics
-- ${contextType === 'technical' ? 'Include relevant code snippets or examples' : 'Provide practical next steps'}
-
-Response Format:
-- Start with a direct answer to the question
-- Support with relevant context from the knowledge base
-- End with helpful next steps or related information
-\`\`\`
-
-### Parameter Recommendations:
-
-1. **Temperature**: 
-   ${answerStyle === 'technical' ? '0.3-0.5 (more precise, less creative)' : answerStyle === 'friendly' ? '0.7-0.9 (more conversational)' : '0.5-0.7 (balanced)'}
-
-2. **top_k**: 
-   - 3-5 for focused answers
-   - 5-8 for comprehensive responses
-   - 8-10 for exploratory questions
-
-3. **threshold**: 
-   - 0.7+ for high confidence answers
-   - 0.5-0.7 for broader context inclusion
-
-### Chat History Usage:
-
-Include previous conversation context to:
-- Maintain conversation flow
-- Reference previous questions
-- Build on established context
-- Avoid repeating information
-
-### Example Usage:
-
-\`\`\`javascript
-{
-  "namespace": "your-namespace",
-  "query": "How do I configure authentication?",
-  "header_prompt": "You are a technical assistant...",
-  "footer_prompt": "Always include code examples...",
-  "temperature": 0.5,
-  "top_k": 5,
-  "chat_history": [
-    {"role": "user", "content": "Previous question..."},
-    {"role": "assistant", "content": "Previous answer..."}
-  ]
-}
-\`\`\`
-
-This configuration will provide ${answerStyle} answers tailored to your ${contextType} use case.
-`;
-
-    return {
-      messages: [
-        {
-          role: 'assistant',
-          content: { type: 'text', text }
-        }
-      ]
-    };
-  }
+import { z } from 'zod';
+
+// Search optimization prompt
+export const searchOptimizationPrompt = {
+  name: 'search-optimization',
+  description: 'Tips for optimizing search queries in Moorcheh',
+  argsSchema: {
+    search_type: z.enum(['text','vector']).optional().describe('Type of search (text or vector)'),
+    domain: z.string().optional().describe('Domain or topic area of your content')
+  },
+  handler: async (args) => {
+    const searchType = args?.search_type || 'text';
+    const domain = args?.domain || 'general';
+    
+    const text = `# Moorcheh Search Optimization Guide
+
+## Optimizing ${searchType} search for ${domain} content
+
+### Request Format Examples:
+
+**Text Search Request:**
+\`\`\`json
+{
+  "query": "your search text here",
+  "namespaces": ["your-namespace"],
+  "top_k": 10,
+  "kiosk_mode": false
+}
+\`\`\`
+
+**Vector Search Request:**
+\`\`\`json
+{
+  "query": [0.1, 0.2, 0.3, ..., 0.768],
+  "namespaces": ["vector-embeddings"],
+  "top_k": 5,
+  "kiosk_mode": true,
+  "threshold": 0.1
+}
+\`\`\`
+
+**cURL Example for Vector Search:**
+\`\`\`bash
+curl -X POST "https://api.moorcheh.ai/v1/search" \\
+  -H "Content-Type: application/json" \\
+  -H "x-api-key: your-api-key-here" \\
+  -d '{
+    "query": [0.1, 0.2, 0.3, ..., 0.768],
+    "namespaces": ["vector-embeddings"],
+    "top_k": 5,
+    "kiosk_mode": true,
+    "threshold": 0.1
+  }'
+\`\`\`
+
+### Search Strategy:
+
+${searchType === 'text' ? `
+**Text Search Best Practices:**
+- Use specific, descriptive terms rather than generic ones
+- Include key domain terminology from your ${domain} field
+- Try both short keywords and longer phrases
+- Use synonyms if initial searches don't return good results
+- Natural language queries work well for semantic search
+
+**Query Examples:**
+- Instead of: "help"
+- Try: "troubleshooting guide", "how to resolve", "step by step"
+- Use complete sentences: "How do I configure authentication?"
+- Include context: "Python authentication setup for web applications"
+` : `
+**Vector Search Best Practices:**
+- Ensure your query embeddings use the same model as your stored vectors
+- Vector search works well with semantic similarity
+- Natural language queries often work better than keywords
+- Consider the context and intent behind your search
+- Match vector dimensions with your namespace configuration
+
+**Query Optimization:**
+- Use complete sentences or questions for better semantic matching
+- Include relevant context in your query
+- Test with variations of the same concept
+- Ensure vector dimensions match your namespace (e.g., 768 for many models)
+`}
+
+### Parameter Tuning:
+
+1. **top_k**: Number of results to return (default: 10)
+   - Start with 5-10 for most use cases
+   - Increase for broader exploration
+   - Decrease for highly targeted results
+
+2. **threshold**: Similarity threshold (0-1, optional)
+   - 0.7-0.8: High similarity, fewer but more relevant results
+   - 0.5-0.7: Moderate similarity, balanced results
+   - 0.3-0.5: Lower similarity, more comprehensive results
+   - **Important**: When kiosk_mode is true, threshold becomes mandatory
+
+3. **kiosk_mode**: Boolean (default: false)
+   - true: Restrict search to specific namespace(s) with threshold filtering
+   - false: Search across all specified namespaces
+   - **Note**: When kiosk_mode is enabled, threshold parameter is required
+
+### Kiosk Mode Requirements:
+
+When using kiosk_mode=true:
+- **threshold parameter is mandatory**
+- Results are filtered by the specified threshold
+- Useful for production environments with strict relevance requirements
+- Example: kiosk_mode=true with threshold=0.7 ensures only high-confidence results
+
+### Domain-Specific Tips for ${domain}:
+- Use terminology specific to ${domain}
+- Consider the typical language patterns in your content
+- Test with actual user queries from your ${domain} context
+- Monitor search performance and adjust parameters accordingly
+- For technical content, include specific technology names and versions
+
+### Troubleshooting:
+- If no results: Lower threshold, check spelling, try synonyms
+- If too many irrelevant results: Raise threshold, use more specific terms
+- If results seem random: Check your embeddings model compatibility
+- For vector search: Verify vector dimensions match namespace configuration
+- For kiosk mode: Ensure threshold is provided when kiosk_mode=true
+
+### Response Format:
+The search returns results with:
+- **id**: Document identifier
+- **score**: Similarity score (0-1, higher is better)
+- **label**: Relevance label (Close Match, Very High Relevance, etc.)
+- **text**: Original text content (for text namespaces)
+- **metadata**: Associated metadata
+`;
+
+    return {
+      messages: [
+        {
+          role: 'assistant',
+          content: { type: 'text', text }
+        }
+      ]
+    };
+  }
+};
+
+// Data organization prompt
+export const dataOrganizationPrompt = {
+  name: 'data-organization',
+  description: 'Best practices for organizing data in Moorcheh namespaces',
+  argsSchema: {
+    content_type: z.string().optional().describe("Type of content you're organizing"),
+    team_size: z.enum(['small','large']).optional().describe('Size of your team using this data')
+  },
+  handler: async (args) => {
+    const contentType = args?.content_type || 'documents';
+    const teamSize = args?.team_size || 'small';
+    
+    const text = `# Moorcheh Data Organization Guide
+
+## Organizing ${contentType} for a ${teamSize} team
+
+### Namespace Strategy:
+
+${teamSize === 'large' ? `
+**Large Team Considerations:**
+- Create separate namespaces by department or project
+- Use consistent naming conventions across teams
+- Implement clear metadata standards
+- Consider access control and permissions
+` : `
+**Small Team Approach:**
+- Fewer namespaces with more content per namespace
+- Flexible organization that can evolve
+- Focus on clear naming and good metadata
+- Regular cleanup and maintenance
+`}
+
+### Content Organization for ${contentType}:
+
+1. **Metadata Schema:**
+   \`\`\`json
+   {
+     "category": "primary classification",
+     "tags": ["tag1", "tag2", "tag3"],
+     "author": "content creator",
+     "created_date": "2024-01-01",
+     "department": "relevant team",
+     "priority": "high/medium/low",
+     "status": "draft/review/published"
+   }
+   \`\`\`
+
+2. **Document ID Conventions:**
+   - Use descriptive, hierarchical IDs
+   - Include date/version when relevant
+   - Examples: "support-faq-login-2024", "product-spec-v2-auth"
+
+3. **Content Chunking:**
+   - Break large documents into logical sections
+   - Each chunk should be self-contained
+   - Include context in metadata
+
+### Maintenance Workflow:
+
+1. **Regular Reviews:**
+   - Monthly metadata cleanup
+   - Remove outdated content
+   - Update tags and categories
+
+2. **Quality Control:**
+   - Standardize formatting before upload
+   - Validate metadata completeness
+   - Test search functionality after major updates
+
+3. **Growth Planning:**
+   - Monitor namespace size and performance
+   - Plan for content archival strategy
+   - Consider splitting namespaces as they grow
+
+### Search-Friendly Organization:
+- Use consistent terminology in your ${contentType}
+- Include alternative phrasings in metadata
+- Create logical content hierarchies
+- Tag content with multiple relevant categories
+
+This organization will help your ${teamSize} team efficiently manage and search through your ${contentType}.
+`;
+
+    return {
+      messages: [
+        {
+          role: 'assistant',
+          content: { type: 'text', text }
+        }
+      ]
+    };
+  }
+};
+
+// AI answer setup prompt
+export const aiAnswerSetupPrompt = {
+  name: 'ai-answer-setup',
+  description: 'Guide for configuring AI-powered answers in Moorcheh',
+  argsSchema: {
+    answer_style: z.enum(['concise','detailed','technical','friendly','balanced']).optional().describe('Desired style for AI answers'),
+    context_type: z.string().optional().describe('Type of context/domain for answers')
+  },
+  handler: async (args) => {
+    const answerStyle = args?.answer_style || 'balanced';
+    const contextType = args?.context_type || 'general';
+    
+    const text = `# Moorcheh AI Answer Configuration Guide
+
+## Setting up ${answerStyle} AI answers for ${contextType} content
+
+### Header Prompt Configuration:
+
+\`\`\`
+You are an AI assistant specialized in ${contextType}. 
+Your role is to provide ${answerStyle} answers based on the provided context.
+
+Style Guidelines:
+${answerStyle === 'concise' ? `
+- Keep answers brief and to the point
+- Use bullet points for multiple items
+- Avoid unnecessary explanations
+- Focus on actionable information
+` : answerStyle === 'detailed' ? `
+- Provide comprehensive explanations
+- Include relevant background information
+- Use examples where helpful
+- Structure answers with clear sections
+` : answerStyle === 'technical' ? `
+- Use precise technical terminology
+- Include code examples when relevant
+- Explain complex concepts step-by-step
+- Reference specific technical details
+` : answerStyle === 'friendly' ? `
+- Use conversational, approachable language
+- Include encouraging phrases
+- Explain concepts in simple terms
+- Ask clarifying questions when needed
+` : `
+- Balance brevity with completeness
+- Use clear, professional language
+- Structure information logically
+- Include key details without overwhelming
+`}
+
+Context: You have access to ${contextType} documentation and resources.
+\`\`\`
+
+### Footer Prompt Configuration:
+
+\`\`\`
+Additional Guidelines:
+- Always cite specific sources when possible
+- If information is incomplete, acknowledge limitations
+- Suggest follow-up questions for complex topics
+- ${contextType === 'technical' ? 'Include relevant code snippets or examples' : 'Provide practical next steps'}
+
+Response Format:
+- Start with a direct answer to the question
+- Support with relevant context from the knowledge base
+- End with helpful next steps or related information
+\`\`\`
+
+### Parameter Recommendations:
+
+1. **Temperature**: 
+   ${answerStyle === 'technical' ? '0.3-0.5 (more precise, less creative)' : answerStyle === 'friendly' ? '0.7-0.9 (more conversational)' : '0.5-0.7 (balanced)'}
+
+2. **top_k**: 
+   - 3-5 for focused answers
+   - 5-8 for comprehensive responses
+   - 8-10 for exploratory questions
+
+3. **threshold**: 
+   - 0.7+ for high confidence answers
+   - 0.5-0.7 for broader context inclusion
+
+### Chat History Usage:
+
+Include previous conversation context to:
+- Maintain conversation flow
+- Reference previous questions
+- Build on established context
+- Avoid repeating information
+
+### Example Usage:
+
+\`\`\`javascript
+{
+  "namespace": "your-namespace",
+  "query": "How do I configure authentication?",
+  "header_prompt": "You are a technical assistant...",
+  "footer_prompt": "Always include code examples...",
+  "temperature": 0.5,
+  "top_k": 5,
+  "chat_history": [
+    {"role": "user", "content": "Previous question..."},
+    {"role": "assistant", "content": "Previous answer..."}
+  ]
+}
+\`\`\`
+
+This configuration will provide ${answerStyle} answers tailored to your ${contextType} use case.
+`;
+
+    return {
+      messages: [
+        {
+          role: 'assistant',
+          content: { type: 'text', text }
+        }
+      ]
+    };
+  }
 };