Latitude MCP Server

/** * PromptL Documentation Content * * Comprehensive documentation for all PromptL syntax features. * Each topic is exported as a markdown string constant. * * Topics: * - overview: Introduction to PromptL * - structure: Config + Messages * - variables: {{ }} syntax * - conditionals: if/else/endif * - loops: for/each iteration * - references: <prompt> tag * - tools: Function calling * - chains: <step> multi-step * - agents: Multi-agent orchestration * - techniques: Few-shot, CoT, ToT, Role prompting */ import type { DocsTopic } from '../types/docs.types.js'; import { DOCS_CONFIG_JSON_OUTPUT, DOCS_CONFIG_GENERATION, DOCS_TOOLS_BUILTIN, DOCS_TECHNIQUE_ROLE, DOCS_RECIPE_CLASSIFICATION, DOCS_RECIPE_EXTRACTION, DOCS_CONVERSATION_HISTORY, DOCS_GUIDE_DEBUGGING, } from './phase1.docs.js'; import { DOCS_CONFIG_BASICS, DOCS_CONFIG_ADVANCED, DOCS_MESSAGES_ROLES, DOCS_MESSAGES_MULTIMODAL, DOCS_TOOLS_CUSTOM, DOCS_TOOLS_SCHEMA, DOCS_TECHNIQUE_FEW_SHOT, DOCS_TECHNIQUE_COT, DOCS_TECHNIQUE_TOT, DOCS_RECIPE_GENERATION, DOCS_RECIPE_CHATBOT, DOCS_GUIDE_SAFETY, DOCS_GUIDE_PERFORMANCE, } from './phase2.docs.js'; import { DOCS_RECIPE_RAG, DOCS_AGENT_PATTERNS, DOCS_GUIDE_TESTING, DOCS_GUIDE_VERSIONING, DOCS_PROVIDERS_OPENAI, DOCS_PROVIDERS_ANTHROPIC, DOCS_PROVIDERS_GOOGLE, } from './phase3.docs.js'; // ============================================================================ // HELP CONTENT - Server Overview // ============================================================================ export const HELP_CONTENT = `# Latitude MCP Server Your AI-powered interface for managing prompts on Latitude.so. ## Available Tools (19 total) ### 📚 Documentation | Tool | Description | |------|-------------| | \`latitude_help\` | This help menu - start here! | | \`latitude_find_docs\` | Search docs by describing your goal | | \`latitude_get_docs\` | Get PromptL documentation by topic | ### 📁 Projects | Tool | Description | |------|-------------| | \`latitude_list_projects\` | List all projects in workspace | | \`latitude_create_project\` | Create a new project | ### 🔀 Versions | Tool | Description | |------|-------------| | \`latitude_list_versions\` | List versions (commits) for a project | | \`latitude_get_version\` | Get version details | | \`latitude_create_version\` | Create a draft version | | \`latitude_publish_version\` | Publish draft to live | ### 📝 Prompts | Tool | Description | |------|-------------| | \`latitude_list_prompts\` | List prompts in a version | | \`latitude_get_prompt\` | Get prompt content by path | | \`latitude_push_prompt\` | Push prompt content to draft | | \`latitude_push_prompt_from_file\` | Push from local file | | \`latitude_push_changes\` | Push multiple changes at once | ### 🤖 Execution | Tool | Description | |------|-------------| | \`latitude_run_prompt\` | Execute prompt with parameters | | \`latitude_chat\` | Continue a conversation | | \`latitude_get_conversation\` | Get conversation history | | \`latitude_stop_conversation\` | Stop ongoing generation | | \`latitude_create_log\` | Log execution for analytics | --- ## 📖 Documentation Topics (38 total) Call \`latitude_get_docs\` with one of these topics: ### Core Syntax (11) | Topic | What You'll Learn | |-------|------------------| | \`overview\` | What is PromptL, getting started | | \`structure\` | Config section + message types | | \`variables\` | \`{{ }}\` syntax, expressions | | \`conditionals\` | \`if/else/endif\` logic | | \`loops\` | \`for/each\` iteration | | \`references\` | Include prompts with \`<prompt>\` | | \`tools\` | Function calling basics | | \`chains\` | Multi-step with \`<step>\` | | \`agents\` | Multi-agent orchestration | | \`techniques\` | Prompting techniques overview | | \`agent-patterns\` | Agent collaboration patterns | ### Configuration (7) | Topic | What You'll Learn | |-------|------------------| | \`config-basics\` | Provider, model essentials | | \`config-generation\` | temperature, maxTokens | | \`config-json-output\` | Structured JSON output | | \`config-advanced\` | maxSteps, retries, seed | | \`providers-openai\` | GPT-4, GPT-3.5 specifics | | \`providers-anthropic\` | Claude specifics | | \`providers-google\` | Gemini specifics | ### Messages (2) | Topic | What You'll Learn | |-------|------------------| | \`messages-roles\` | System, user, assistant, tool | | \`messages-multimodal\` | Images, files, rich content | ### Tools (4) | Topic | What You'll Learn | |-------|------------------| | \`tools-builtin\` | latitude/search, /code, /extract | | \`tools-custom\` | Define custom tools | | \`tools-schema\` | JSON Schema deep dive | ### Techniques (4) | Topic | What You'll Learn | |-------|------------------| | \`technique-role\` | Expert roles, personas | | \`technique-few-shot\` | Examples for learning | | \`technique-cot\` | Chain-of-Thought reasoning | | \`technique-tot\` | Tree-of-Thoughts branching | ### Recipes (5) | Topic | What You'll Learn | |-------|------------------| | \`recipe-classification\` | Sentiment, intent | | \`recipe-extraction\` | Entity extraction | | \`recipe-generation\` | Content creation | | \`recipe-chatbot\` | Conversational AI | | \`recipe-rag\` | Retrieval Augmented Generation | ### Guides (6) | Topic | What You'll Learn | |-------|------------------| | \`conversation-history\` | Multi-turn context | | \`guide-debugging\` | Common errors, fixes | | \`guide-safety\` | Guardrails, filtering | | \`guide-performance\` | Token optimization | | \`guide-testing\` | Testing strategies | | \`guide-versioning\` | Draft vs live workflow | --- ## 🚀 Quick Start Workflow \`\`\` 1. latitude_list_projects → Find your project 2. latitude_get_docs("structure") → Learn PromptL basics 3. latitude_create_version → Create a draft 4. latitude_push_prompt → Add your prompt 5. latitude_run_prompt → Test it! 6. latitude_publish_version → Go live \`\`\` --- ## 📚 Resources (Read-Only) | Resource URI | Description | |-------------|-------------| | \`latitude://projects\` | List all projects | | \`latitude://projects/{id}/versions\` | List versions | | \`latitude://projects/{id}/versions/{uuid}/prompts\` | List prompts | | \`docs://latitude/{topic}\` | Documentation by topic | --- ## 💡 Tips - **Start with \`latitude_get_docs("overview")\`** to understand PromptL - **Use "live" for production**, create drafts for testing - **Can only push to draft versions**, not "live" - **File push auto-derives path** from filename **Need help with a specific topic?** Call \`latitude_get_docs\` with the topic name. `; // ============================================================================ // TOPIC: OVERVIEW // ============================================================================ const DOCS_OVERVIEW = `# PromptL Overview > The prompt templating language for Latitude.so ## What is PromptL? PromptL is a powerful templating language designed for creating dynamic, reusable AI prompts. It combines: - **YAML configuration** for model settings - **XML-like message tags** for conversation structure - **Mustache-style variables** for dynamic content - **Logic constructs** for conditionals and loops ## Why Use PromptL? | Feature | Benefit | |---------|---------| | **Version Control** | Track prompt changes like code | | **Reusability** | Share prompts across projects | | **Dynamic Content** | Inject variables at runtime | | **Provider Agnostic** | Works with OpenAI, Anthropic, Google, etc. | | **Testability** | Run prompts directly from API | ## Quick Example \`\`\`promptl --- provider: OpenAI model: gpt-4o temperature: 0.7 --- You are a helpful assistant that speaks {{ language || "English" }}. <user> {{ user_question }} </user> \`\`\` ## Core Concepts ### 1. Config Section YAML between \`---\` delimiters. Sets provider, model, and parameters. ### 2. Messages Chat-based structure with roles: \`system\`, \`user\`, \`assistant\`, \`tool\`. ### 3. Variables Dynamic content with \`{{ variable }}\` syntax. ### 4. Logic Conditionals (\`if/else\`) and loops (\`for/each\`) for dynamic prompts. ### 5. Advanced Features - **References**: Include other prompts - **Chains**: Multi-step reasoning - **Agents**: Multi-agent orchestration - **Tools**: Function calling --- ## Getting Started 1. **Create a project**: \`latitude_create_project\` 2. **Create a draft**: \`latitude_create_version\` 3. **Push a prompt**: \`latitude_push_prompt\` 4. **Test it**: \`latitude_run_prompt\` 5. **Go live**: \`latitude_publish_version\` --- ## Next Steps - **Learn structure**: \`latitude_get_docs({ topic: "structure" })\` - **Master variables**: \`latitude_get_docs({ topic: "variables" })\` - **Add logic**: \`latitude_get_docs({ topic: "conditionals" })\` `; // ============================================================================ // TOPIC: STRUCTURE // ============================================================================ const DOCS_STRUCTURE = `# Prompt Structure > Config section + Messages = Complete prompt ## Overview Every PromptL prompt has two sections: 1. **Config Section** (optional) - YAML settings between \`---\` 2. **Messages** - Conversation flow with role tags \`\`\`promptl --- provider: OpenAI model: gpt-4o temperature: 0.7 --- You are a helpful assistant. <user> Hello, how can you help me? </user> \`\`\` --- ## Config Section The config section is **optional** but recommended. It uses YAML format: \`\`\`yaml --- provider: OpenAI model: gpt-4o-mini temperature: 0.6 top_p: 0.9 maxTokens: 1000 --- \`\`\` ### Required Settings | Setting | Description | Example | |---------|-------------|---------| | \`provider\` | AI provider name | \`OpenAI\`, \`Anthropic\`, \`Google\` | | \`model\` | Model identifier | \`gpt-4o\`, \`claude-3-opus\`, \`gemini-pro\` | ### Generation Parameters | Parameter | Description | Range | |-----------|-------------|-------| | \`temperature\` | Randomness (lower = focused) | 0.0 - 2.0 | | \`maxTokens\` | Max response length | 1 - model max | | \`topP\` | Nucleus sampling | 0.0 - 1.0 | | \`topK\` | Top-K sampling | 1 - vocabulary size | | \`presencePenalty\` | Avoid repetition | -2.0 - 2.0 | | \`frequencyPenalty\` | Avoid frequent words | -2.0 - 2.0 | ### Advanced Settings | Setting | Description | |---------|-------------| | \`seed\` | For reproducible outputs | | \`stopSequences\` | Stop generation at these strings | | \`maxSteps\` | Max steps for chains/agents (default: 20) | | \`maxRetries\` | Retry failed calls (default: 2) | | \`schema\` | JSON Schema for structured output | | \`tools\` | Available functions for model to call | --- ## Messages Messages define the conversation. Use XML-like tags: ### System Message (Default) Text without tags is a **system message**: \`\`\`promptl You are a helpful assistant. \`\`\` ### User Message \`\`\`promptl <user> What's the weather like today? </user> \`\`\` ### Assistant Message \`\`\`promptl <assistant> I'd be happy to help! What city are you in? </assistant> \`\`\` ### Tool Message \`\`\`promptl <tool id="call_123" name="get_weather"> {"temperature": 72, "condition": "sunny"} </tool> \`\`\` --- ## Complete Example \`\`\`promptl --- provider: OpenAI model: gpt-4o temperature: 0.7 maxTokens: 500 --- You are a knowledgeable travel assistant. Help users plan trips. Be concise but informative. Always ask clarifying questions. <user> I want to plan a vacation. </user> <assistant> I'd love to help you plan a vacation! To give you the best recommendations: 1. What's your budget range? 2. Do you prefer beaches, cities, or nature? 3. When are you planning to travel? </assistant> <user> {{ user_input }} </user> \`\`\` --- ## Tips - **Temperature 0** enables response caching - **Use either** \`temperature\` OR \`topP\`, not both - **System messages** set the AI's behavior and context - **Message order matters** - it's a conversation flow --- ## Next Steps - **Add variables**: \`latitude_get_docs({ topic: "variables" })\` - **Add logic**: \`latitude_get_docs({ topic: "conditionals" })\` - **Try tools**: \`latitude_get_docs({ topic: "tools" })\` `; // ============================================================================ // TOPIC: VARIABLES // ============================================================================ const DOCS_VARIABLES = `# Variables & Expressions > Dynamic content with \`{{ }}\` syntax ## Overview Variables let you inject dynamic content into prompts at runtime. Use double curly braces: \`{{ variable_name }}\` \`\`\`promptl Hello, {{ user_name }}! How can I help you today? \`\`\` --- ## Syntax Reference ### Basic Interpolation \`\`\`promptl {{ variable_name }} \`\`\` ### Variable Assignment \`\`\`promptl {{ greeting = "Hello" }} {{ count = 42 }} {{ is_premium = true }} \`\`\` ### Default Values \`\`\`promptl {{ name || "Guest" }} {{ config.timeout || 30 }} {{ items || [] }} \`\`\` ### Expressions \`\`\`promptl {{ price * quantity }} {{ firstName + " " + lastName }} {{ age >= 18 }} {{ items.length }} \`\`\` ### Object Access \`\`\`promptl {{ user.name }} {{ user.profile.email }} {{ data['key-with-dashes'] }} \`\`\` ### Array Access \`\`\`promptl {{ items[0] }} {{ users[index].name }} {{ results[results.length - 1] }} \`\`\` --- ## Examples ### Example 1: User Greeting \`\`\`promptl --- provider: OpenAI model: gpt-4o --- {{ greeting = "Hello" }} {{ name = user.firstName || "there" }} You are a friendly assistant. <user> {{ greeting }}, {{ name }}! I need help with {{ topic }}. </user> \`\`\` **Parameters:** \`\`\`json { "user": { "firstName": "Alice" }, "topic": "coding" } \`\`\` ### Example 2: Dynamic Pricing \`\`\`promptl {{ subtotal = items.length * pricePerItem }} {{ discount = subtotal > 100 ? 0.1 : 0 }} {{ total = subtotal * (1 - discount) }} Your order summary: - Items: {{ items.length }} - Subtotal: \${{ subtotal }} - Discount: {{ discount * 100 }}% - Total: \${{ total }} \`\`\` ### Example 3: Conditional Content \`\`\`promptl {{ language = user.language || "English" }} {{ formality = user.isPremium ? "formal" : "casual" }} You are a {{ formality }} assistant that responds in {{ language }}. \`\`\` --- ## Operators | Operator | Example | Description | |----------|---------|-------------| | \`+\` | \`a + b\` | Addition / Concatenation | | \`-\` | \`a - b\` | Subtraction | | \`*\` | \`a * b\` | Multiplication | | \`/\` | \`a / b\` | Division | | \`%\` | \`a % b\` | Modulo | | \`==\` | \`a == b\` | Equality | | \`!=\` | \`a != b\` | Inequality | | \`<\` | \`a < b\` | Less than | | \`>\` | \`a > b\` | Greater than | | \`<=\` | \`a <= b\` | Less or equal | | \`>=\` | \`a >= b\` | Greater or equal | | \`&&\` | \`a && b\` | Logical AND | | \`\\|\\|\` | \`a \\|\\| b\` | Logical OR / Default | | \`!\` | \`!a\` | Logical NOT | | \`? :\` | \`a ? b : c\` | Ternary | --- ## Common Patterns | Pattern | Syntax | Use Case | |---------|--------|----------| | **Fallback** | \`{{ x \\|\\| "default" }}\` | Handle missing values | | **Nested access** | \`{{ user.profile.name }}\` | Deep object access | | **Array item** | \`{{ items[0] }}\` | Get first item | | **Array length** | \`{{ items.length }}\` | Count items | | **Ternary** | \`{{ x ? "yes" : "no" }}\` | Inline conditional | | **Math** | \`{{ a + b * c }}\` | Calculations | --- ## Tips - **Always provide defaults** for optional variables: \`{{ x || "default" }}\` - **Use descriptive names**: \`{{ user_email }}\` not \`{{ e }}\` - **Keep expressions simple** - complex logic should use conditionals - **Test with edge cases** - undefined, null, empty strings --- ## Next Steps - **Add conditions**: \`latitude_get_docs({ topic: "conditionals" })\` - **Add loops**: \`latitude_get_docs({ topic: "loops" })\` - **Push a prompt**: \`latitude_push_prompt\` `; // ============================================================================ // TOPIC: CONDITIONALS // ============================================================================ const DOCS_CONDITIONALS = `# Conditional Statements > Dynamic content based on conditions with \`if/else/endif\` ## Overview Conditionals let you include or exclude content based on runtime conditions. Use \`{{ if }}\`, \`{{ else }}\`, and \`{{ endif }}\`. \`\`\`promptl {{ if user.isPremium }} You have access to all features! {{ else }} Upgrade to premium for more features. {{ endif }} \`\`\` --- ## Syntax ### Basic If \`\`\`promptl {{ if condition }} Content when true {{ endif }} \`\`\` ### If-Else \`\`\`promptl {{ if condition }} Content when true {{ else }} Content when false {{ endif }} \`\`\` ### If-Else If-Else \`\`\`promptl {{ if condition1 }} Content for condition1 {{ else if condition2 }} Content for condition2 {{ else }} Default content {{ endif }} \`\`\` --- ## Examples ### Example 1: Age Verification \`\`\`promptl {{ if age < 18 }} Sorry, you must be 18 or older to use this service. {{ else }} <user> {{ question }} </user> {{ endif }} \`\`\` ### Example 2: Role-Based System Message \`\`\`promptl {{ if role == "admin" }} <system> The user is an admin with full access to all data and operations. </system> {{ else if role == "moderator" }} <system> The user is a moderator who can manage content but not system settings. </system> {{ else }} <system> The user has standard access with limited permissions. </system> {{ endif }} \`\`\` ### Example 3: Nested Conditions \`\`\`promptl {{ if user.isPremium }} <system>Premium user with enhanced features.</system> {{ if user.hasApiAccess }} API access is enabled. You can help with API-related questions. {{ endif }} {{ else }} <system>Standard user. Suggest premium upgrade when relevant.</system> {{ endif }} \`\`\` ### Example 4: Variable Existence Check \`\`\`promptl {{ if lastName }} {{ fullName = firstName + " " + lastName }} {{ else }} {{ fullName = firstName }} {{ endif }} <user> Hi! My name is {{ fullName }}. </user> \`\`\` ### Example 5: Complex Expressions \`\`\`promptl {{ if items.length > 0 && user.isLoggedIn }} <assistant> You have {{ items.length }} items in your cart. Ready to checkout? </assistant> {{ else if !user.isLoggedIn }} <assistant> Please log in to view your cart. </assistant> {{ else }} <assistant> Your cart is empty. Let's find something for you! </assistant> {{ endif }} \`\`\` --- ## Condition Operators | Operator | Example | Description | |----------|---------|-------------| | \`==\` | \`role == "admin"\` | Equals | | \`!=\` | \`status != "blocked"\` | Not equals | | \`<\` | \`age < 18\` | Less than | | \`>\` | \`price > 100\` | Greater than | | \`<=\` | \`count <= 10\` | Less or equal | | \`>=\` | \`score >= 80\` | Greater or equal | | \`&&\` | \`a && b\` | Both true | | \`\\|\\|\` | \`a \\|\\| b\` | Either true | | \`!\` | \`!isBlocked\` | Not | --- ## Truthiness These are **falsy** (evaluate to false): - \`false\` - \`null\` - \`undefined\` - \`0\` - \`""\` (empty string) Everything else is **truthy**. \`\`\`promptl {{ if userName }} Hello, {{ userName }}! {{ else }} Hello, guest! {{ endif }} \`\`\` --- ## Best Practices | Do | Don't | |----|----| | ✅ \`{{ if user.isLoggedIn }}\` | ❌ \`{{ if x }}\` (unclear) | | ✅ Keep conditions simple | ❌ Deep nesting (3+ levels) | | ✅ Use defaults for optional vars | ❌ Assume variables exist | | ✅ Test edge cases | ❌ Only test happy path | --- ## Debugging Add debug output to check values: \`\`\`promptl Debug: role={{ role }}, isPremium={{ isPremium }} {{ if role == "admin" && isPremium }} ... {{ endif }} \`\`\` --- ## Next Steps - **Add loops**: \`latitude_get_docs({ topic: "loops" })\` - **Learn variables**: \`latitude_get_docs({ topic: "variables" })\` - **Try techniques**: \`latitude_get_docs({ topic: "techniques" })\` `; // ============================================================================ // TOPIC: LOOPS // ============================================================================ const DOCS_LOOPS = `# Loops & Iteration > Repeat content with \`for/each\` over arrays ## Overview Loops let you iterate over arrays to generate dynamic content. Perfect for few-shot examples, lists, and repeated structures. \`\`\`promptl {{ for item in items }} - {{ item }} {{ endfor }} \`\`\` --- ## Syntax ### Basic Loop \`\`\`promptl {{ for item in array }} Content with {{ item }} {{ endfor }} \`\`\` ### Loop with Index \`\`\`promptl {{ for item, index in array }} {{ index + 1 }}. {{ item }} {{ endfor }} \`\`\` ### Object Properties \`\`\`promptl {{ for value, key in object }} {{ key }}: {{ value }} {{ endfor }} \`\`\` --- ## Examples ### Example 1: Few-Shot Learning \`\`\`promptl --- provider: OpenAI model: gpt-4o --- You are a sentiment classifier. Classify text as positive, negative, or neutral. {{ for example in examples }} <user>{{ example.text }}</user> <assistant>{{ example.sentiment }}</assistant> {{ endfor }} <user>{{ input_text }}</user> \`\`\` **Parameters:** \`\`\`json { "examples": [ { "text": "I love this product!", "sentiment": "positive" }, { "text": "Terrible experience.", "sentiment": "negative" }, { "text": "It's okay I guess.", "sentiment": "neutral" } ], "input_text": "This is amazing!" } \`\`\` ### Example 2: Numbered List \`\`\`promptl Here are your tasks for today: {{ for task, index in tasks }} {{ index + 1 }}. {{ task.title }} (Priority: {{ task.priority }}) {{ endfor }} \`\`\` ### Example 3: Conversation History \`\`\`promptl {{ for message in history }} <{{ message.role }}> {{ message.content }} </{{ message.role }}> {{ endfor }} <user>{{ new_message }}</user> \`\`\` ### Example 4: Product Catalog \`\`\`promptl Available products: {{ for product in products }} **{{ product.name }}** - Price: \${{ product.price }} - In Stock: {{ product.inStock ? "Yes" : "No" }} {{ endfor }} Which product interests you? \`\`\` ### Example 5: Nested Loops \`\`\`promptl {{ for category in categories }} ## {{ category.name }} {{ for item in category.items }} - {{ item }} {{ endfor }} {{ endfor }} \`\`\` --- ## Common Patterns ### Dynamic Examples (Few-Shot) \`\`\`promptl {{ for ex in examples }} Input: {{ ex.input }} Output: {{ ex.output }} {{ endfor }} Now classify: {{ user_input }} \`\`\` ### Building Context \`\`\`promptl <system> You have access to these documents: {{ for doc in documents }} - {{ doc.title }}: {{ doc.summary }} {{ endfor }} </system> \`\`\` ### Conditional in Loop \`\`\`promptl {{ for user in users }} {{ if user.isActive }} - {{ user.name }} (Active) {{ endif }} {{ endfor }} \`\`\` --- ## Loop Variables Inside a loop, you have access to: | Variable | Description | |----------|-------------| | \`item\` | Current item | | \`index\` | Current index (0-based) | | \`length\` | Array length | \`\`\`promptl {{ for item, index in items }} Item {{ index + 1 }} of {{ items.length }}: {{ item }} {{ endfor }} \`\`\` --- ## Tips - **Use for few-shot examples** - dynamically inject training examples - **Combine with conditionals** - filter items in loops - **Keep iterations reasonable** - large loops increase token count - **Index is 0-based** - add 1 for human-readable numbering --- ## Next Steps - **Learn techniques**: \`latitude_get_docs({ topic: "techniques" })\` (few-shot) - **Try conditionals**: \`latitude_get_docs({ topic: "conditionals" })\` - **Build chains**: \`latitude_get_docs({ topic: "chains" })\` `; // ============================================================================ // TOPIC: REFERENCES // ============================================================================ const DOCS_REFERENCES = `# Prompt References > Include and chain prompts with \`<prompt>\` tag ## Overview References let you include one prompt inside another, enabling modular and reusable prompt design. \`\`\`promptl <prompt path="shared/system-instructions" /> <user>{{ user_question }}</user> \`\`\` --- ## Syntax ### Basic Reference \`\`\`promptl <prompt path="prompt-name" /> \`\`\` ### With Parameters \`\`\`promptl <prompt path="greeting" name="Alice" style="formal" /> \`\`\` ### Relative Paths \`\`\`promptl <prompt path="../shared/common" /> <prompt path="./helpers/formatter" /> \`\`\` --- ## Examples ### Example 1: Shared System Instructions **Main prompt:** \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <prompt path="system/base-assistant" /> <user>{{ question }}</user> \`\`\` **system/base-assistant:** \`\`\`promptl You are a helpful, accurate, and friendly AI assistant. Always be concise. Ask clarifying questions when needed. Never make up information - say "I don't know" if unsure. \`\`\` ### Example 2: Parameterized Reference **Main prompt:** \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <prompt path="templates/persona" role="financial advisor" tone="professional" expertise="retirement planning" /> <user>{{ user_question }}</user> \`\`\` **templates/persona:** \`\`\`promptl You are a {{ role }} with expertise in {{ expertise }}. Your communication style is {{ tone }}. Provide accurate, helpful advice within your domain. \`\`\` ### Example 3: Composable Prompts \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <prompt path="components/system-rules" /> <prompt path="components/output-format" format="json" /> <prompt path="components/examples" count="3" /> <user>{{ input }}</user> \`\`\` --- ## Use Cases | Use Case | Description | |----------|-------------| | **Shared system prompts** | Consistent behavior across prompts | | **Reusable personas** | Same character in different contexts | | **Common examples** | Few-shot examples used in multiple prompts | | **Output formatters** | Consistent JSON/markdown formatting | | **Guardrails** | Safety instructions everywhere | --- ## Path Resolution | Path | Resolves To | |------|-------------| | \`helper\` | Same folder: \`/folder/helper\` | | \`./helper\` | Same folder explicitly | | \`../shared\` | Parent folder: \`/shared\` | | \`templates/base\` | From project root | --- ## Tips - **Keep referenced prompts focused** - single responsibility - **Pass only needed parameters** - don't over-parameterize - **Document dependencies** - note which prompts reference others - **Test references independently** - ensure they work alone --- ## Next Steps - **Build chains**: \`latitude_get_docs({ topic: "chains" })\` - **Create agents**: \`latitude_get_docs({ topic: "agents" })\` - **Push prompts**: \`latitude_push_prompt\` `; // ============================================================================ // TOPIC: TOOLS // ============================================================================ const DOCS_TOOLS = `# Tool Use & Function Calling > Enable AI to call external functions ## Overview Tools let you give the AI model access to external functions. Define tools in the config, and the model can call them during execution. \`\`\`promptl --- provider: OpenAI model: gpt-4o tools: - get_weather: description: Get current weather for a location parameters: type: object properties: location: type: string description: City name, e.g., "San Francisco, CA" required: [location] --- <user>What's the weather in Tokyo?</user> \`\`\` --- ## Syntax ### Tool Definition \`\`\`yaml tools: - tool_name: description: What the tool does parameters: type: object properties: param1: type: string description: Parameter description required: [param1] \`\`\` ### Built-in Latitude Tools \`\`\`yaml tools: - latitude/search # Web search - latitude/code # Code execution - latitude/extract # Data extraction \`\`\` --- ## JSON Schema for Parameters Parameters use JSON Schema format: ### String Parameter \`\`\`yaml location: type: string description: City and state \`\`\` ### Number Parameter \`\`\`yaml temperature: type: number description: Temperature in Celsius \`\`\` ### Enum Parameter \`\`\`yaml unit: type: string enum: [celsius, fahrenheit] description: Temperature unit \`\`\` ### Array Parameter \`\`\`yaml tags: type: array items: type: string description: List of tags \`\`\` ### Object Parameter \`\`\`yaml user: type: object properties: name: type: string age: type: integer required: [name] \`\`\` --- ## Examples ### Example 1: Weather Tool \`\`\`promptl --- provider: OpenAI model: gpt-4o tools: - get_weather: description: Get current weather for a specified location parameters: type: object properties: location: type: string description: City and state, e.g., San Francisco, CA unit: type: string enum: [celsius, fahrenheit] description: Temperature unit required: [location] --- You are a weather assistant. <user>What's the weather like in {{ location }}?</user> \`\`\` ### Example 2: Calendar Tool \`\`\`promptl --- provider: OpenAI model: gpt-4o tools: - add_calendar_event: description: Add an event to the user's calendar parameters: type: object properties: title: type: string description: Event title date: type: string description: Date in YYYY-MM-DD format time: type: string description: Time in HH:MM format duration_minutes: type: integer description: Duration in minutes required: [title, date, time] --- You are a calendar assistant. Help users manage their schedule. <user>{{ request }}</user> \`\`\` ### Example 3: Multiple Tools \`\`\`promptl --- provider: OpenAI model: gpt-4o tools: - latitude/search - get_stock_price: description: Get current stock price parameters: type: object properties: symbol: type: string description: Stock ticker symbol (e.g., AAPL) required: [symbol] - calculate: description: Perform mathematical calculation parameters: type: object properties: expression: type: string description: Math expression to evaluate required: [expression] --- You are a financial assistant with access to web search, stock prices, and calculations. <user>{{ question }}</user> \`\`\` --- ## Tool Without Parameters For tools that don't need parameters: \`\`\`yaml tools: - get_random_number: description: Generate a random number parameters: type: object properties: {} required: [] \`\`\` --- ## Tips - **Write clear descriptions** - helps the model decide when to use the tool - **Mark required parameters** - prevents invalid calls - **Use enums for limited options** - guides the model's choices - **Test tool calls** - ensure parameters are correct --- ## Next Steps - **Build agents**: \`latitude_get_docs({ topic: "agents" })\` - **Add chains**: \`latitude_get_docs({ topic: "chains" })\` - **Run prompts**: \`latitude_run_prompt\` `; // ============================================================================ // TOPIC: CHAINS // ============================================================================ const DOCS_CHAINS = `# Chains & Multi-Step Prompts > Sequential reasoning with \`<step>\` tags ## Overview Chains let you break prompts into sequential steps. Each step runs, and its output is available to subsequent steps. Perfect for complex reasoning and multi-stage processing. \`\`\`promptl <step> Analyze the problem: {{ problem }} </step> <step> Based on the analysis, propose a solution. </step> <step> Evaluate the solution and provide final recommendations. </step> \`\`\` --- ## Syntax ### Basic Chain \`\`\`promptl <step> First step content </step> <step> Second step - can reference previous output </step> \`\`\` ### Accessing Previous Response The \`response\` variable contains the previous step's output: \`\`\`promptl <step> List 3 ideas for {{ topic }}. </step> <step> Based on these ideas: {{ response }} Evaluate each idea's feasibility. </step> \`\`\` --- ## Examples ### Example 1: Analysis Chain \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <step> # Step 1: Data Analysis Analyze this data: {{ data }} Extract key patterns and insights. </step> <step> # Step 2: Interpretation Based on the analysis: {{ response }} What do these patterns mean for the business? </step> <step> # Step 3: Recommendations Given the interpretation, provide 3 actionable recommendations. </step> \`\`\` ### Example 2: Chain of Thought \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <step> # Step 1: Understand the Problem Problem: {{ problem }} Let me break this down: 1. What is being asked? 2. What information do I have? 3. What approach should I take? </step> <step> # Step 2: Work Through the Solution Based on my understanding: {{ response }} Now let me solve this step by step... </step> <step> # Step 3: Verify and Conclude Solution work: {{ response }} Let me verify this is correct and provide the final answer. </step> \`\`\` ### Example 3: Content Generation Pipeline \`\`\`promptl --- provider: OpenAI model: gpt-4o --- <step> # Research Phase Topic: {{ topic }} Target audience: {{ audience }} Generate an outline with 5 key points to cover. </step> <step> # Draft Phase Using this outline: {{ response }} Write a detailed first draft. </step> <step> # Edit Phase Draft: {{ response }} Edit for clarity, engagement, and accuracy. Provide the final version. </step> \`\`\` --- ## Configuration ### Max Steps Limit the number of steps (default: 20, max: 150): \`\`\`yaml --- provider: OpenAI model: gpt-4o maxSteps: 50 --- \`\`\` ### Opt Out of maxSteps For simple prompts without steps: \`\`\`yaml --- provider: OpenAI model: gpt-4o type: prompt --- \`\`\` --- ## Chain with Tools Chains work with tools - each step can call tools: \`\`\`promptl --- provider: OpenAI model: gpt-4o tools: - latitude/search --- <step> Search for recent news about {{ topic }}. </step> <step> Based on the search results, summarize the key developments. </step> <step> Analyze the implications and provide insights. </step> \`\`\` --- ## Tips - **Keep steps focused** - one objective per step - **Use clear step titles** - helps with debugging - **Reference \`response\`** - connect steps logically - **Set appropriate maxSteps** - prevent runaway chains - **Test each step** - ensure logic flows correctly --- ## Next Steps - **Build agents**: \`latitude_get_docs({ topic: "agents" })\` - **Learn techniques**: \`latitude_get_docs({ topic: "techniques" })\` - **Push prompts**: \`latitude_push_prompt\` `; // ============================================================================ // TOPIC: AGENTS // ============================================================================ const DOCS_AGENTS = `# Multi-Agent Systems > Orchestrate multiple AI agents for complex tasks ## Overview Agents let you create specialized AI personas that can collaborate. Define agents in separate prompts and orchestrate them from a main prompt. \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent agents: - agents/researcher - agents/writer - agents/editor --- Coordinate the team to create content about {{ topic }}. \`\`\` --- ## Syntax ### Main Orchestrator \`\`\`yaml --- provider: OpenAI model: gpt-4o type: agent agents: - agents/agent1 - agents/agent2 --- Instructions for coordinating agents... \`\`\` ### Agent Definition Each agent is a separate prompt file: \`\`\`yaml --- provider: OpenAI model: gpt-4o type: agent path: agents/researcher --- You are a research specialist... \`\`\` --- ## Examples ### Example 1: Content Creation Team **Main orchestrator (content-team.promptl):** \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent agents: - agents/researcher - agents/writer - agents/editor --- # Content Creation Workflow Topic: {{ topic }} Target audience: {{ audience }} Tone: {{ tone }} Coordinate the team: 1. Researcher gathers information 2. Writer creates the draft 3. Editor polishes the final version Deliver high-quality content. \`\`\` **agents/researcher:** \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent path: agents/researcher --- # Research Agent You are a thorough research specialist. - Find key facts and statistics - Identify credible sources - Organize information logically - Note any gaps in available information Research thoroughly and present findings clearly. \`\`\` **agents/writer:** \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent path: agents/writer --- # Writer Agent You are a skilled content writer. - Create engaging, clear prose - Structure content logically - Match the target tone and audience - Incorporate research findings naturally Write compelling content. \`\`\` **agents/editor:** \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent path: agents/editor --- # Editor Agent You are a meticulous editor. - Check grammar and spelling - Improve clarity and flow - Ensure consistency - Verify facts and claims - Polish the final product Deliver publication-ready content. \`\`\` ### Example 2: Code Review Team \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent agents: - agents/security-reviewer - agents/performance-reviewer - agents/style-reviewer --- # Code Review Workflow Review this code: \`\`\` {{ code }} \`\`\` Each specialist should review from their perspective. Compile a comprehensive review report. \`\`\` ### Example 3: Decision Committee \`\`\`promptl --- provider: OpenAI model: gpt-4o type: agent agents: - agents/ceo-perspective - agents/cto-perspective - agents/cfo-perspective --- # Executive Decision Decision to evaluate: {{ decision }} Each executive provides their perspective. Synthesize into a unified recommendation. \`\`\` --- ## Agent Configuration ### Path Declaration Agents need their path declared: \`\`\`yaml --- type: agent path: agents/my-agent --- \`\`\` ### Agent with Tools Agents can have their own tools: \`\`\`yaml --- type: agent path: agents/researcher tools: - latitude/search --- \`\`\` --- ## Best Practices | Practice | Description | |----------|-------------| | **Specialized roles** | Each agent has clear expertise | | **Clear handoffs** | Define how agents pass work | | **Distinct personas** | Different voices for each agent | | **Focused scope** | Agents do one thing well | | **Orchestration logic** | Main prompt coordinates clearly | --- ## Tips - **Start simple** - 2-3 agents before scaling - **Test agents individually** - ensure each works alone - **Define clear interfaces** - what each agent inputs/outputs - **Use descriptive agent names** - self-documenting - **Monitor token usage** - multi-agent uses more tokens --- ## Next Steps - **Learn chains**: \`latitude_get_docs({ topic: "chains" })\` - **Add tools**: \`latitude_get_docs({ topic: "tools" })\` - **Try techniques**: \`latitude_get_docs({ topic: "techniques" })\` `; // ============================================================================ // TOPIC: TECHNIQUES // ============================================================================ const DOCS_TECHNIQUES = `# Prompting Techniques > Few-shot, Chain-of-Thought, Tree-of-Thoughts, Role Prompting ## Overview Advanced prompting techniques improve AI performance on complex tasks. Combine these with PromptL features for powerful prompts. --- ## 1. Few-Shot Learning Provide examples to guide the model's behavior. ### Basic Few-Shot \`\`\`promptl Classify the sentiment as positive, negative, or neutral. Text: I love this product! Sentiment: positive Text: Terrible experience. Sentiment: negative Text: It's okay. Sentiment: neutral Text: {{ input_text }} Sentiment: \`\`\` ### Dynamic Few-Shot with Loops \`\`\`promptl {{ for example in examples }} Input: {{ example.input }} Output: {{ example.output }} {{ endfor }} Input: {{ user_input }} Output: \`\`\` ### Few-Shot with Messages \`\`\`promptl {{ for ex in examples }} <user>{{ ex.question }}</user> <assistant>{{ ex.answer }}</assistant> {{ endfor }} <user>{{ user_question }}</user> \`\`\` --- ## 2. Chain-of-Thought (CoT) Guide the model through step-by-step reasoning. ### Simple CoT \`\`\`promptl Solve this problem step by step: {{ problem }} Let's think through this: 1. First, I need to understand... 2. Then, I should consider... 3. Finally, I can conclude... \`\`\` ### CoT with Steps \`\`\`promptl <step> # Understand the Problem {{ problem }} What is being asked? What information do I have? </step> <step> # Develop Solution Based on my understanding, let me work through this... </step> <step> # Verify and Conclude Let me check my work and provide the final answer. </step> \`\`\` ### Zero-Shot CoT Simply add "Let's think step by step": \`\`\`promptl {{ problem }} Let's think step by step. \`\`\` --- ## 3. Tree-of-Thoughts (ToT) Explore multiple reasoning paths and evaluate them. ### Basic ToT \`\`\`promptl Problem: {{ problem }} ## Branch A: First Approach 1. Premise: ... 2. Reasoning: ... 3. Conclusion: ... Confidence: X/10 ## Branch B: Alternative Approach 1. Premise: ... 2. Reasoning: ... 3. Conclusion: ... Confidence: X/10 ## Branch C: Creative Approach 1. Premise: ... 2. Reasoning: ... 3. Conclusion: ... Confidence: X/10 ## Evaluation Compare branches and select the best solution. ## Final Answer Based on Branch [X]: ... \`\`\` ### ToT with Chains \`\`\`promptl <step> # Generate Branches Problem: {{ problem }} Generate 3 different approaches to solve this. </step> <step> # Evaluate Branches Based on the approaches above, evaluate each: - Feasibility (1-10) - Correctness (1-10) - Elegance (1-10) </step> <step> # Select and Develop Choose the best approach and develop it fully. </step> \`\`\` --- ## 4. Role Prompting Assign a specific persona or expertise. ### Expert Role \`\`\`promptl You are a senior financial advisor with 20 years of experience. - CFA certified - Specializes in retirement planning - Communication style: professional yet approachable {{ client_situation }} Provide expert advice: \`\`\` ### Dynamic Role \`\`\`promptl {{ if domain == "legal" }} You are a senior attorney specializing in {{ specialty }}. {{ else if domain == "medical" }} You are a board-certified physician in {{ specialty }}. {{ else if domain == "technical" }} You are a principal engineer with expertise in {{ specialty }}. {{ endif }} Question: {{ question }} \`\`\` ### Multi-Role with Agents \`\`\`promptl --- type: agent agents: - agents/optimist - agents/pessimist - agents/pragmatist --- Evaluate {{ proposal }} from multiple perspectives. \`\`\` --- ## 5. Combining Techniques ### Few-Shot + CoT \`\`\`promptl {{ for example in examples }} Problem: {{ example.problem }} Thinking: {{ example.reasoning }} Answer: {{ example.answer }} {{ endfor }} Problem: {{ user_problem }} Thinking: Let me work through this step by step... Answer: \`\`\` ### Role + Chain + Tools \`\`\`promptl --- provider: OpenAI model: gpt-4o tools: - latitude/search --- You are a research analyst at a top consulting firm. <step> Research: {{ topic }} Use search to find current information. </step> <step> Analyze findings and identify key trends. </step> <step> Provide executive summary with recommendations. </step> \`\`\` --- ## Quick Reference | Technique | When to Use | Key Feature | |-----------|-------------|-------------| | **Few-Shot** | Classification, formatting | Examples guide behavior | | **CoT** | Reasoning, math, logic | Step-by-step thinking | | **ToT** | Complex decisions, planning | Multiple paths evaluated | | **Role** | Domain expertise, persona | Specialized knowledge | --- ## Tips - **Combine techniques** - they work well together - **Use loops for few-shot** - dynamic example injection - **Use chains for CoT/ToT** - structured reasoning steps - **Use agents for roles** - multi-perspective analysis - **Test and iterate** - different techniques suit different tasks --- ## Next Steps - **Try loops**: \`latitude_get_docs({ topic: "loops" })\` - **Build chains**: \`latitude_get_docs({ topic: "chains" })\` - **Create agents**: \`latitude_get_docs({ topic: "agents" })\` - **Push prompts**: \`latitude_push_prompt\` `; // ============================================================================ // DOCS MAP - Topic to Content Mapping // ============================================================================ export const DOCS_MAP: Record<DocsTopic, string> = { // Core syntax (original) overview: DOCS_OVERVIEW, structure: DOCS_STRUCTURE, variables: DOCS_VARIABLES, conditionals: DOCS_CONDITIONALS, loops: DOCS_LOOPS, references: DOCS_REFERENCES, tools: DOCS_TOOLS, chains: DOCS_CHAINS, agents: DOCS_AGENTS, techniques: DOCS_TECHNIQUES, // Phase 1 - Tier 1 topics 'config-json-output': DOCS_CONFIG_JSON_OUTPUT, 'config-generation': DOCS_CONFIG_GENERATION, 'tools-builtin': DOCS_TOOLS_BUILTIN, 'technique-role': DOCS_TECHNIQUE_ROLE, 'recipe-classification': DOCS_RECIPE_CLASSIFICATION, 'recipe-extraction': DOCS_RECIPE_EXTRACTION, 'conversation-history': DOCS_CONVERSATION_HISTORY, 'guide-debugging': DOCS_GUIDE_DEBUGGING, // Phase 2 - Tier 2 topics 'config-basics': DOCS_CONFIG_BASICS, 'config-advanced': DOCS_CONFIG_ADVANCED, 'messages-roles': DOCS_MESSAGES_ROLES, 'messages-multimodal': DOCS_MESSAGES_MULTIMODAL, 'tools-custom': DOCS_TOOLS_CUSTOM, 'tools-schema': DOCS_TOOLS_SCHEMA, 'technique-few-shot': DOCS_TECHNIQUE_FEW_SHOT, 'technique-cot': DOCS_TECHNIQUE_COT, 'technique-tot': DOCS_TECHNIQUE_TOT, 'recipe-generation': DOCS_RECIPE_GENERATION, 'recipe-chatbot': DOCS_RECIPE_CHATBOT, 'guide-safety': DOCS_GUIDE_SAFETY, 'guide-performance': DOCS_GUIDE_PERFORMANCE, // Phase 3 - Tier 3 topics 'recipe-rag': DOCS_RECIPE_RAG, 'agent-patterns': DOCS_AGENT_PATTERNS, 'guide-testing': DOCS_GUIDE_TESTING, 'guide-versioning': DOCS_GUIDE_VERSIONING, 'providers-openai': DOCS_PROVIDERS_OPENAI, 'providers-anthropic': DOCS_PROVIDERS_ANTHROPIC, 'providers-google': DOCS_PROVIDERS_GOOGLE, };