Latitude MCP Server

/** * Recipes Documentation * Complete implementation patterns for common use cases */ // ============================================================================ // RECIPE-ANALYSIS // ============================================================================ export const DOCS_RECIPE_ANALYSIS = `# Analysis Recipe > Multi-step data analysis workflow ## Overview Analyze data through structured steps: gather, process, interpret, conclude. Perfect for reports, insights, and decision support. --- ## Quick Reference \`\`\` Gather data → Analyze → Synthesize → Recommend \`\`\` --- ## Basic Analysis \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <step as="data_summary"> Summarize the key data points: {{ data }} </step> <step as="analysis"> Analyze patterns and trends: {{ data_summary }} </step> <step as="insights"> What are the top 3 insights? {{ analysis }} </step> <step> ## Analysis Report ### Data Summary {{ data_summary }} ### Key Insights {{ insights }} ### Recommendations Based on the analysis, here are my recommendations: </step> \`\`\` --- ## Structured Analysis Report \`\`\`promptl --- provider: OpenAI model: gpt-4o schema: type: object properties: executive_summary: type: string key_findings: type: array items: type: object properties: finding: type: string evidence: type: string impact: type: string enum: [high, medium, low] required: [finding, evidence, impact] recommendations: type: array items: type: object properties: action: type: string priority: type: string enum: [immediate, short-term, long-term] rationale: type: string required: [action, priority] risks: type: array items: type: string required: [executive_summary, key_findings, recommendations] --- <system> You are a business analyst. Provide structured, actionable analysis. </system> <user> Analyze this data and provide a comprehensive report: {{ data }} Focus areas: {{ focus_areas }} </user> \`\`\` --- ## Comparative Analysis \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <step as="option_a"> Analyze Option A: {{ option_a }} - Pros: - Cons: - Key metrics: </step> <step as="option_b"> Analyze Option B: {{ option_b }} - Pros: - Cons: - Key metrics: </step> <step> ## Comparison | Aspect | Option A | Option B | |--------|----------|----------| | ... | ... | ... | ### Recommendation Based on {{ criteria }}, I recommend... </step> \`\`\` --- ## With Research \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent tools: - latitude/search - latitude/extract --- <system> You are a research analyst. 1. Gather relevant data 2. Analyze patterns 3. Provide evidence-based conclusions </system> <user> Research and analyze: {{ topic }} Deliverable: {{ report_type }} </user> \`\`\` --- ## Best Practices - ✅ **DO**: Break analysis into clear steps - ✅ **DO**: Cite data for every claim - ✅ **DO**: Provide actionable recommendations - ✅ **DO**: Acknowledge limitations - ❌ **DON'T**: Make claims without evidence - ❌ **DON'T**: Overcomplicate simple analyses --- ## Next Steps - **Chains** → \`docs({ action: "get", topic: "chains" })\` - **RAG Recipe** → \`docs({ action: "get", topic: "recipe-rag" })\` `; // ============================================================================ // RECIPE-MODERATION // ============================================================================ export const DOCS_RECIPE_MODERATION = `# Content Moderation Recipe > Detect and filter inappropriate content ## Overview Automatically detect harmful, inappropriate, or policy-violating content. Essential for user-generated content platforms. --- ## Quick Reference \`\`\`promptl --- provider: OpenAI model: gpt-4o schema: type: object properties: safe: type: boolean categories: type: array items: type: string confidence: type: number required: [safe, categories, confidence] --- Analyze content for policy violations. \`\`\` --- ## Basic Moderation \`\`\`promptl --- provider: OpenAI model: gpt-4o temperature: 0 schema: type: object properties: decision: type: string enum: [approve, flag, reject] reason: type: string categories: type: array items: type: string enum: [safe, spam, hate_speech, harassment, violence, sexual, self_harm, illegal, misinformation] confidence: type: number required: [decision, categories, confidence] --- <system> You are a content moderator. Analyze content against these policies: PROHIBITED: - Hate speech or discrimination - Harassment or bullying - Violence or threats - Sexual content (if inappropriate for platform) - Self-harm promotion - Illegal activities - Misinformation Flag content that needs human review. Reject clearly violating content. Approve safe content. </system> <user> Moderate this content: {{ content }} </user> \`\`\` --- ## Multi-Category Moderation \`\`\`promptl --- provider: OpenAI model: gpt-4o temperature: 0 schema: type: object properties: overall_safe: type: boolean categories: type: object properties: hate_speech: type: object properties: detected: type: boolean severity: type: string enum: [none, low, medium, high] evidence: type: string harassment: type: object properties: detected: type: boolean severity: type: string enum: [none, low, medium, high] evidence: type: string violence: type: object properties: detected: type: boolean severity: type: string enum: [none, low, medium, high] evidence: type: string action: type: string enum: [approve, flag_for_review, auto_reject] required: [overall_safe, categories, action] --- Analyze content for each category of potential violation. Content: {{ content }} \`\`\` --- ## With Context \`\`\`promptl --- provider: OpenAI model: gpt-4o temperature: 0 --- <system> Content moderator for {{ platform_type }}. Platform rules: {{ platform_rules }} Age restriction: {{ age_rating }} Consider context - something appropriate for one platform may not be for another. </system> <user> Content: {{ content }} Author history: {{ author_context }} Thread context: {{ thread_context }} </user> \`\`\` --- ## Moderation Pipeline \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <step as="quick_check" temperature={{ 0 }}> Quick safety check - any obvious violations? Content: {{ content }} Answer: safe/unsafe </step> {{ if quick_check == "unsafe" }} <step as="detailed" temperature={{ 0 }}> Detailed analysis of flagged content: {{ content }} Specific violations: Severity: Recommended action: </step> {{ endif }} <step> {{ if quick_check == "safe" }} APPROVED - No violations detected {{ else }} {{ detailed }} {{ endif }} </step> \`\`\` --- ## Best Practices - ✅ **DO**: Use temperature 0 for consistency - ✅ **DO**: Require confidence scores - ✅ **DO**: Consider context - ✅ **DO**: Have human review for edge cases - ❌ **DON'T**: Rely solely on AI moderation - ❌ **DON'T**: Over-moderate (false positives) --- ## Next Steps - **Safety guide** → \`docs({ action: "get", topic: "guide-safety" })\` - **Classification** → \`docs({ action: "get", topic: "recipe-classification" })\` `; // ============================================================================ // RECIPE-SUPPORT // ============================================================================ export const DOCS_RECIPE_SUPPORT = `# Customer Support Recipe > AI support agent with tools and escalation ## Overview Build a customer support agent that can answer questions, look up information, and escalate to humans when needed. --- ## Quick Reference \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent tools: - search_knowledge_base - get_order_status - escalate_to_human --- Customer support agent with tools. \`\`\` --- ## Basic Support Agent \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <system> You are a customer support agent for {{ company }}. Guidelines: - Be empathetic and professional - Solve problems step by step - Offer alternatives if first solution doesn't work - Apologize for inconvenience when appropriate - Know when to escalate {{ if customer_tier }} Customer tier: {{ customer_tier }} {{ endif }} </system> {{ for msg in history }} <{{ msg.role }}>{{ msg.content }}</{{ msg.role }}> {{ endfor }} <user>{{ message }}</user> \`\`\` --- ## Support Agent with Tools \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent maxSteps: 10 tools: - search_knowledge_base: description: Search help articles and FAQs parameters: type: object properties: query: type: string required: [query] - get_order_status: description: Look up order status by order ID parameters: type: object properties: order_id: type: string required: [order_id] - get_account_info: description: Get customer account information parameters: type: object properties: customer_id: type: string required: [customer_id] - create_ticket: description: Create support ticket for complex issues parameters: type: object properties: subject: type: string priority: type: string enum: [low, medium, high, urgent] description: type: string required: [subject, priority, description] - escalate_to_human: description: Transfer to human agent parameters: type: object properties: reason: type: string context: type: string required: [reason] --- <system> You are a support agent for {{ company }}. WORKFLOW: 1. Understand the customer's issue 2. Search knowledge base first 3. Look up account/order info if relevant 4. Try to resolve the issue 5. Create ticket for complex issues 6. Escalate to human if: - Customer requests it - Issue is too complex - Customer is frustrated - Involves refunds > $100 - Security concerns TONE: - Professional but friendly - Empathetic - Solution-oriented Customer ID: {{ customer_id }} </system> {{ for msg in history }} <{{ msg.role }}>{{ msg.content }}</{{ msg.role }}> {{ endfor }} <user>{{ message }}</user> \`\`\` --- ## Issue Classification \`\`\`promptl --- provider: OpenAI model: gpt-4o temperature: 0 schema: type: object properties: category: type: string enum: [billing, technical, shipping, returns, account, general] priority: type: string enum: [low, medium, high, urgent] sentiment: type: string enum: [positive, neutral, frustrated, angry] needs_human: type: boolean summary: type: string required: [category, priority, sentiment, needs_human, summary] --- Classify this support request: {{ message }} \`\`\` --- ## Escalation Handling \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <step as="classification"> Classify and assess: - Issue type - Customer sentiment - Complexity - Requires human? (yes/no) Message: {{ message }} </step> {{ if classification requires_human }} <step> I understand this is a complex situation. Let me connect you with a specialist who can better assist you. [Creating ticket and escalating...] In the meantime, is there anything else I can help clarify? </step> {{ else }} <step> [Provide solution based on classification] </step> {{ endif }} \`\`\` --- ## Multi-Language Support \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <system> You are a multilingual support agent. Detect the customer's language and respond in the same language. If unsure, ask for confirmation. Supported languages: English, Spanish, French, German, Portuguese </system> <user>{{ message }}</user> \`\`\` --- ## Best Practices - ✅ **DO**: Search knowledge base before answering - ✅ **DO**: Track conversation context - ✅ **DO**: Have clear escalation criteria - ✅ **DO**: Be empathetic with frustrated customers - ❌ **DON'T**: Make promises you can't keep - ❌ **DON'T**: Argue with customers - ❌ **DON'T**: Ignore escalation signals --- ## Next Steps - **Chatbot recipe** → \`docs({ action: "get", topic: "recipe-chatbot" })\` - **Agents** → \`docs({ action: "get", topic: "agents" })\` `; // ============================================================================ // TOOLS-ORCHESTRATION // ============================================================================ export const DOCS_TOOLS_ORCHESTRATION = `# Tool Orchestration > Coordinate multiple tools in agents ## Overview Orchestrate multiple tools to accomplish complex tasks. Define when to use which tool and how to chain results. --- ## Quick Reference \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent tools: - tool_a - tool_b - tool_c --- Use tools in sequence or parallel to achieve goal. \`\`\` --- ## Sequential Orchestration \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent tools: - search_products: description: Search product catalog parameters: type: object properties: query: type: string required: [query] - get_inventory: description: Check inventory for product parameters: type: object properties: product_id: type: string required: [product_id] - create_order: description: Create order for product parameters: type: object properties: product_id: type: string quantity: type: integer required: [product_id, quantity] --- <system> Shopping assistant workflow: 1. Search for products matching request 2. Check inventory for selected product 3. Create order if in stock 4. Report result Always confirm inventory before ordering. </system> <user> {{ request }} </user> \`\`\` --- ## Conditional Tool Selection \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent tools: - search_internal_docs: description: Search company documentation parameters: type: object properties: query: type: string required: [query] - search_web: description: Search the public web parameters: type: object properties: query: type: string required: [query] - ask_expert: description: Escalate to human expert parameters: type: object properties: question: type: string required: [question] --- <system> Research assistant with tiered search: 1. FIRST: Search internal docs 2. IF not found: Search web 3. IF still unclear: Ask expert Don't search web for internal company questions. Don't ask expert for simple factual questions. </system> <user> {{ question }} </user> \`\`\` --- ## Parallel Tool Calls \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent tools: - get_weather: description: Get weather for city parameters: type: object properties: city: type: string required: [city] - get_events: description: Get events for city parameters: type: object properties: city: type: string required: [city] - get_restaurants: description: Get restaurants for city parameters: type: object properties: city: type: string cuisine: type: string required: [city] --- <system> Travel planner. When user asks about a city: - Get weather, events, AND restaurants simultaneously - Synthesize into travel recommendation </system> <user> Plan a day trip to {{ city }} </user> \`\`\` --- ## Error Handling \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent tools: - primary_api: description: Main data source parameters: type: object properties: query: type: string required: [query] - fallback_api: description: Backup data source parameters: type: object properties: query: type: string required: [query] --- <system> Data retrieval with fallback: 1. Try primary_api first 2. If it fails or returns empty, try fallback_api 3. If both fail, explain limitation to user Never return empty results without trying fallback. </system> <user> {{ query }} </user> \`\`\` --- ## Tool Selection Guide | Scenario | Strategy | |----------|----------| | Data gathering | Parallel calls | | Dependent data | Sequential calls | | Uncertain source | Conditional with fallback | | User workflow | Sequential with confirmation | --- ## Best Practices - ✅ **DO**: Define clear tool selection criteria - ✅ **DO**: Handle tool failures gracefully - ✅ **DO**: Limit number of tools (5-7 max) - ✅ **DO**: Use descriptive tool names - ❌ **DON'T**: Give agent too many tools - ❌ **DON'T**: Skip error handling --- ## Next Steps - **Agents** → \`docs({ action: "get", topic: "agents" })\` - **Custom tools** → \`docs({ action: "get", topic: "tools-custom" })\` `; // ============================================================================ // PROVIDERS-AZURE // ============================================================================ export const DOCS_PROVIDERS_AZURE = `# Azure OpenAI Provider > Enterprise Azure-hosted OpenAI models ## Overview Azure OpenAI provides OpenAI models through Microsoft Azure, offering enterprise compliance, regional deployment, and Azure integration. --- ## Quick Reference \`\`\`yaml --- provider: Azure model: gpt-4 endpoint: https://your-resource.openai.azure.com deployment: your-deployment-name --- \`\`\` --- ## Configuration \`\`\`yaml --- provider: Azure model: gpt-4 endpoint: {{ azure_endpoint }} deployment: {{ deployment_name }} api_version: 2024-02-15-preview --- \`\`\` --- ## Available Models Models are deployed to your Azure resource: | Azure Model | OpenAI Equivalent | |-------------|-------------------| | gpt-4 | GPT-4 | | gpt-4-32k | GPT-4 32k context | | gpt-4-turbo | GPT-4 Turbo | | gpt-35-turbo | GPT-3.5 Turbo | --- ## Deployment Setup 1. Create Azure OpenAI resource 2. Deploy a model 3. Get endpoint and deployment name 4. Configure in prompt --- ## Complete Example \`\`\`promptl --- provider: Azure model: gpt-4 endpoint: https://mycompany.openai.azure.com deployment: gpt4-production api_version: 2024-02-15-preview temperature: 0.7 --- <system> Enterprise assistant with Azure compliance. </system> <user> {{ question }} </user> \`\`\` --- ## When to Use Azure | Requirement | Use Azure? | |-------------|------------| | Enterprise compliance | Yes ✅ | | Data residency | Yes ✅ | | Azure integration | Yes ✅ | | Quick prototyping | No (use OpenAI directly) | | Cost-sensitive | Depends on agreement | --- ## Azure vs OpenAI Direct | Aspect | Azure | OpenAI Direct | |--------|-------|---------------| | Compliance | Enterprise | Standard | | Regions | Multiple | US-based | | SLA | Available | Limited | | Setup | More complex | Simple | | Pricing | Custom | Pay-as-you-go | --- ## Best Practices - ✅ **DO**: Use for enterprise workloads - ✅ **DO**: Set up proper Azure RBAC - ✅ **DO**: Monitor usage in Azure portal - ❌ **DON'T**: Hardcode credentials - ❌ **DON'T**: Use for quick prototypes --- ## Next Steps - **OpenAI** → \`docs({ action: "get", topic: "providers-openai" })\` - **Config basics** → \`docs({ action: "get", topic: "config-basics" })\` `;