prompt-plus-plus-mcp/dist/llm-client.js

Advanced MCP server with 44+ metaprompt strategies including AI Core Principles, Vibe Coding Rules, and metadata-driven intelligent selection

import { logger } from './logger.js';
/**
 * LLMClient prepares prompts for Claude processing through MCP.
 * The actual LLM interaction happens through the MCP protocol when
 * Claude receives these prepared prompts.
 */
export class LLMClient {
    constructor() {
        logger.info('LLM Client initialized for MCP prompt preparation');
    }
    /**
     * Prepares a prompt for Claude processing.
     * In the MCP context, this returns the formatted prompt that Claude
     * will process when the user interacts with the MCP server.
     */
    async sendPrompt(prompt) {
        try {
            logger.debug('Preparing prompt for Claude', { promptLength: prompt.length });
            // Return the formatted prompt for Claude to process
            // The actual LLM processing happens when Claude receives this through MCP
            return {
                text: prompt,
                rawResponse: { prompt, prepared: true }
            };
        }
        catch (error) {
            logger.error('Failed to prepare prompt', {
                error: error instanceof Error ? error.message : String(error)
            });
            throw error;
        }
    }
    /**
     * Formats a prompt with additional context for better Claude processing
     */
    formatPromptForClaude(prompt, context) {
        const formattedPrompt = `${prompt}

---
Note: This prompt has been prepared by the prompt-plus-plus-mcp server for enhanced processing.`;
        if (context?.strategy) {
            return `${formattedPrompt}\nStrategy: ${context.strategy.name}`;
        }
        return formattedPrompt;
    }
    async processMetaprompt(userPrompt, strategy, type = 'refine') {
        let fullPrompt;
        if (type === 'refine') {
            const metapromptTemplate = strategy.template.replace('[Insert initial prompt here]', userPrompt);
            fullPrompt = `You are an expert prompt engineer. Apply the '${strategy.name}' meta-prompt template to refine the following user prompt.

${strategy.description}

Process the meta-prompt completely and return a JSON response with:
1. initial_prompt_evaluation: Analysis of the original prompt's strengths and weaknesses
2. refined_prompt: The enhanced version
3. explanation_of_refinements: What was improved and why

Meta-prompt template:
${metapromptTemplate}

Remember to return your response in valid JSON format.`;
        }
        else {
            // Auto-refine logic
            fullPrompt = `You are an expert prompt engineer. Analyze this prompt and select the best strategy to enhance it:

User Prompt: ${userPrompt}

Available Strategies: [Strategies would be listed here]

Select the most appropriate strategy and apply it to create an enhanced version of the prompt.`;
        }
        return await this.sendPrompt(fullPrompt);
    }
}
// Singleton instance
export const llmClient = new LLMClient();
//# sourceMappingURL=llm-client.js.map