Spaces:
Sleeping
Sleeping
Meet Patel
Refactor TutorX MCP server to integrate Mistral OCR for document processing, update concept graph tools for LLM-driven responses, and enhance learning path generation with Gemini. Transitioned various tools to utilize LLM for improved educational interactions and streamlined API responses.
a806ca2
| """ | |
| Concept-related MCP tools for TutorX. | |
| """ | |
| import random | |
| from typing import Dict, Any, Optional | |
| from datetime import datetime, timezone | |
| import sys | |
| import os | |
| from pathlib import Path | |
| import json | |
| # Add the parent directory to the Python path | |
| current_dir = Path(__file__).parent | |
| parent_dir = current_dir.parent.parent | |
| sys.path.insert(0, str(parent_dir)) | |
| import sys | |
| import os | |
| from pathlib import Path | |
| # Add the parent directory to the Python path | |
| current_dir = Path(__file__).parent | |
| parent_dir = current_dir.parent | |
| sys.path.insert(0, str(parent_dir)) | |
| # Import from local resources | |
| from resources.concept_graph import get_concept, get_all_concepts | |
| # Import MCP | |
| from mcp_server.mcp_instance import mcp | |
| from mcp_server.model.gemini_flash import GeminiFlash | |
| MODEL = GeminiFlash() | |
| async def get_concept_tool(concept_id: str = None) -> dict: | |
| """ | |
| Get a specific concept or all concepts from the knowledge graph, fully LLM-driven. | |
| If a concept_id is provided, use Gemini to generate a JSON object with explanation, key points, and example. | |
| """ | |
| if not concept_id: | |
| return {"error": "concept_id is required for LLM-driven mode"} | |
| prompt = ( | |
| f"Explain the concept '{concept_id}' in detail. " | |
| f"Return a JSON object with fields: explanation (string), key_points (list of strings), and example (string)." | |
| ) | |
| llm_response = await MODEL.generate_text(prompt) | |
| try: | |
| data = json.loads(llm_response) | |
| except Exception: | |
| data = {"llm_raw": llm_response, "error": "Failed to parse LLM output as JSON"} | |
| return data | |
| async def assess_skill_tool(student_id: str, concept_id: str) -> dict: | |
| """ | |
| Assess a student's understanding of a specific concept, fully LLM-driven. | |
| Use Gemini to generate a JSON object with a score (0-1), feedback, and recommendations. | |
| """ | |
| prompt = ( | |
| f"A student (ID: {student_id}) is being assessed on the concept '{concept_id}'. " | |
| f"Generate a JSON object with: score (float 0-1), feedback (string), and recommendations (list of strings)." | |
| ) | |
| llm_response = await MODEL.generate_text(prompt) | |
| try: | |
| data = json.loads(llm_response) | |
| except Exception: | |
| data = {"llm_raw": llm_response, "error": "Failed to parse LLM output as JSON"} | |
| return data | |