rowbull/mcp-prompt-server

Quin MCP Prompt Server

This server manages conversation prompts and localization for Quin's financial advisor AI. It decouples conversation logic from the core LLM, enabling global scalability.

Running the Server

1. Install dependencies:

   pip install -r requirements.txt
   

2. Run the Flask development server:

   python app.py
   

The server will be available at http://127.0.0.1:5000.

API Endpoints

Health Check

• URL: /health
• Method: GET
• Success Response:
  • Code: 200
  • Content: {"status": "ok"}

Get Prompt Instructions

This is the main endpoint for the Quin Edge Function to call.

• URL: /mcp/getPromptInstructions
• Method: POST
• Data Params:

  {
    "country": "US",
    "language": "en",
    "sessionType": "session1",
    "userContext": {}
  }
  

• Success Response:
  • Code: 200
  • Content: A JSON object with the full conversation instructions.

Test Endpoint

A simple GET endpoint for quick testing.

• URL: /test/prompt?country=US&language=en&session=session1
• Method: GET
• Success Response:
  • Code: 200
  • Content: A JSON object with the full conversation instructions.

Integration Documentation

To integrate this MCP server with the Quin Supabase Edge Function, the function should be modified to make a POST request to the /mcp/getPromptInstructions endpoint instead of using a hardcoded prompt.

Example Edge Function (JavaScript):

import { serve } from "https://deno.land/std@0.168.0/http/server.ts"

serve(async (req) => {
  // Assuming you have user data and session info
  const userProfile = { ... };
  const sessionType = 'session1';
  const userLanguage = 'en'; // or 'es'
  const userCountry = 'US';

  const mcpResponse = await fetch('YOUR_MCP_SERVER_URL/mcp/getPromptInstructions', {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
    },
    body: JSON.stringify({
      country: userCountry,
      language: userLanguage,
      sessionType: sessionType,
      userContext: userProfile,
    }),
  });

  const promptData = await mcpResponse.json();

  // Now use promptData to construct the prompt for the LLM
  // e.g., pass promptData.instructions to the LLM

  return new Response(
    JSON.stringify(promptData),
    { headers: { "Content-Type": "application/json" } },
  )
})

Expansion Guide

Adding a new country or language is straightforward.

Adding a new Language

1. Create Prompt File: Add a new prompt file in mcp_server/prompts/. For example, for French in the US, create us_fr_session1.txt.
2. Update Available Languages: In mcp_server/logic.py, update the get_available_languages function to include the new language for the respective country.

Adding a new Country

1. Create Context File: Add a new financial context file in mcp_server/context/. For example, for Canada, create ca_context.json.
2. Create Prompt Files: Add prompt files for the new country in mcp_server/prompts/. For example, ca_en_session1.txt.
3. Update Available Languages: In mcp_server/logic.py, add the new country and its supported languages to the get_available_languages function.