name: personalize-chapter
description: Generates a personalized version of any curriculum chapter adapted to the learner's skill level, professional background, and language preference. Uses the learner's profile from Neon Postgres combined with Qdrant-retrieved chapter content.
license: MIT
allowed-tools: neon_postgres_read qdrant_search gemini_generate redis_cache
metadata:
author: Ayesha Khalid
version: "1.0.0"
category: education
domain: personalization
requires_auth: "true"
cache_ttl: "3600"
Personalize Chapter Skill
Purpose
Rewrite any curriculum chapter to match the learner's background so it feels written specifically for them. A software engineer gets code-heavy analogies; a hardware engineer gets circuit-level comparisons; a student gets first-principles explanations.
Prerequisites
- Learner must be authenticated (valid RS256 JWT)
- Learner profile must exist in Neon Postgres (
users table)
Instructions
- Validate JWT via the JWKS endpoint (
/.well-known/jwks.json)
- Fetch learner profile from Neon Postgres:
skill_level: beginner | intermediate | advancedbackground: software_engineer | hardware_engineer | roboticist | studentpreferred_language: en | ur | (other)
- Retrieve chapter chunks from Qdrant (filter by
chapter_id metadata field, top-k=20)
- Check Redis cache — key:
personalize:{user_id}:{chapter_id}:{profile_hash} (TTL: 1 hour). Return cached version if hit.
- Build the personalization prompt:
You are Aria, personalizing a robotics curriculum chapter.
The learner is a {background} with {skill_level} level.
Language: {language}.
Rewrite the following chapter content so it:
- Uses analogies from {background} domain
- Adjusts technical depth for {skill_level}
- Keeps all code blocks unchanged
- If language is Urdu: translate prose to Urdu RTL, keep code in English
[CHAPTER CONTENT]
{chapter_chunks}
- Stream the personalized chapter via SSE
- Cache the result in Redis (TTL: 1 hour)
Supported Personalizations
| Background |
Adaptation Strategy |
software_engineer |
Emphasize API contracts, async patterns, type systems |
hardware_engineer |
Connect to electrical signals, timing, I/O interfaces |
roboticist |
Use kinematics/dynamics framing, reference URDF/SDF directly |
student |
First-principles explanations, more worked examples, slower pacing |
Constraints
- Code blocks (
\```) are NEVER modified or translated
- ROS 2 package names, topic names, and CLI commands stay in English always
- Personalized output must include
skill_level and background in the response metadata
Error Handling
| Scenario |
Response |
| Unauthenticated request |
401 + login modal redirect |
| Profile not found |
Use default: beginner + student + English |
| Chapter not in Qdrant |
"This chapter has not been indexed yet. Try the standard version." |
| Gemini failure |
Fall back to OpenAI; if both fail, return the original chapter with a notice |
By