import { GetPromptResult } from "@modelcontextprotocol/sdk/types";
import { generateVariableTypeContent } from "../generateVariableTypeContent";
import { handleAiActionsInvoke } from "../ai-actions-invoke";
import { handleAiActionsOverview } from "../ai-actions-overview";
/**
* Handler for AI Actions create prompt
* @param args Optional arguments for the prompt
* @returns Prompt result with messages
*/
export function handleAiActionsCreate(args?: Record<string, string>): GetPromptResult {
return {
messages: [
{
role: "assistant",
content: {
type: "text",
text: "I'm your Contentful AI Actions specialist. I can help you design, create, and configure AI Actions that integrate AI capabilities into your content workflows. I understand templates, variables, conditions, and model settings.",
},
},
{
role: "user",
content: {
type: "text",
text: `I want to create an AI Action for this use case: ${args?.useCase || "my Contentful project"}. ${args?.modelType ? `I'm planning to use the ${args?.modelType} model.` : "Please recommend appropriate model types."} Guide me through the creation process, including template design, variable configuration, and testing.`,
},
},
{
role: "assistant",
content: {
type: "text",
text: `# Creating an AI Action for ${args?.useCase || "your Contentful project"}
I'll guide you through creating an effective AI Action for your use case. Here's a step-by-step approach:
## 1. Define your AI Action
Let's start by clearly defining what your AI Action will do:
- **Name**: Choose a descriptive name that reflects the action's purpose
- **Description**: Write a clear description of what the AI Action does and when to use it
## 2. Choose the right model
${args?.modelType ? `You've mentioned using the ${args?.modelType} model, which is a good choice for this use case.` : `For your use case, consider these models:
- **GPT-4** (gpt-4): Best for complex reasoning and high-quality content generation
- **Claude 3 Opus** (claude-3-opus): Excellent for nuanced understanding and creative content
- **Claude 3 Sonnet** (claude-3-5-sonnet): Good balance of quality and performance
- **GPT-3.5 Turbo** (gpt-3.5-turbo): Fastest option for simpler tasks`}
**Temperature setting**:
- Lower (0.0-0.3): More consistent, predictable outputs
- Medium (0.4-0.7): Balance of creativity and consistency
- Higher (0.8-1.0): More creative, varied outputs
## 3. Design your template
The template is the prompt that will be sent to the AI model, with placeholders for variables. Here are some best practices:
- Start with clear instructions about the task
- Define the desired tone, style, and format
- Include context about how the output will be used
- Use variable placeholders with double curly braces: {{variable_name}}
Example template structure:
\`\`\`
You are helping create content for a Contentful entry.
TASK: [Clear description of what to generate]
CONTEXT: The content will be used for {{purpose}}.
TONE: {{tone}}
BASED ON THIS INFORMATION: {{input_content}}
GENERATE: [Specific output instructions]
\`\`\`
## 4. Define your variables
For each placeholder in your template, you'll need to define a variable:
### Common variable types:
1. **StandardInput**: For primary text input
2. **Text**: For simple text fields
3. **FreeFormInput**: For custom text input
4. **StringOptionsList**: For selecting from predefined options
5. **Reference**: For linking to other Contentful entries
6. **MediaReference**: For linking to assets (images, videos, etc.)
7. **Locale**: For specifying language/region
For each variable, define:
- ID: Internal identifier
- Name: User-friendly display name
- Description: Clear explanation of what the variable is for
- Type: One of the types above
- Configuration: Type-specific settings (e.g., option values for StringOptionsList)
## 5. Implementation steps
To create this AI Action using the MCP tools:
1. Use the **create_ai_action** tool with these parameters:
- spaceId, environmentId
- name, description
- instruction (containing template and variables)
- configuration (model type and temperature)
2. Test your AI Action:
- Add test cases with sample values
- Verify outputs match expectations
3. Publish your AI Action:
- Use the **publish_ai_action** tool to make it available to content editors
Would you like me to help you structure a specific template and variables for your use case?`,
},
},
],
};
}
/**
* Handler for AI Actions variables prompt
* @param args Optional arguments for the prompt
* @returns Prompt result with messages
*/
export function handleAiActionsVariables(args?: Record<string, string>): GetPromptResult {
return {
messages: [
{
role: "assistant",
content: {
type: "text",
text: "I'm your Contentful AI Actions variables expert. I can explain how to configure variables for AI Actions, including different types, configurations, and best practices for different scenarios.",
},
},
{
role: "user",
content: {
type: "text",
text: `${args?.variableType ? `Explain how to use and configure the ${args?.variableType} variable type in AI Actions.` : "Explain the different variable types available in AI Actions, their use cases, and how to configure them effectively."} Include examples and best practices for template integration.`,
},
},
{
role: "assistant",
content: {
type: "text",
text: `# AI Action Variables${args?.variableType ? `: ${args?.variableType} Type` : " Overview"}
${generateVariableTypeContent(args?.variableType)}`,
},
},
],
};
}
/**
* Export all AI Actions related handlers
*/
export const aiActionsHandlers = {
"ai-actions-overview": () => handleAiActionsOverview(),
"ai-actions-create": (args?: Record<string, string>) => handleAiActionsCreate(args),
"ai-actions-variables": (args?: Record<string, string>) => handleAiActionsVariables(args),
"ai-actions-invoke": (args?: Record<string, string>) => handleAiActionsInvoke(args?.actionId, args?.details),
};
