chatnct / implementation_plan.md
Mina Diaa
Deploy to Hugging Face Space
3c3b1f4
|
Raw History Blame Contribute Delete
1.16 kB

Academic Analyzer - Implementation Plan

Current State

  • academic_analyzer agent exists with 2 tools: get_student_attendance_summary and get_course_session_log
  • Database already has: profiles, courses, attendance_sessions, attendance tables
  • Server already injects [STUDENT_CODE: xxx] into messages (line 129 in server.py)
  • Agent already extracts student code from the prefix

What Needs to Be Done

1. New Database Tables (Migrations)

  • lectures — maps each attendance_session to a lecture with title/topic
  • lecture_summaries — stores AI-generated summaries for each lecture

2. New Tools

  • get_missed_lectures — returns list of lectures the student missed
  • get_lecture_summaries — returns summaries for missed lectures

3. Security Hardening

  • Add explicit student_code validation in every tool
  • Block any user-supplied student codes (only use injected ones)
  • Add guardrails in agent instructions

4. Enhanced Agent Instructions

  • Egyptian dialect refinements
  • Deprivation warnings (< 75% threshold)
  • Cross-agent referrals (study_agent for summaries)

5. Updated Migrations in database.py