Issue W2-MT-02 — Embeddings and Local Retrieval

[Abdully-dev]

📋 Overview
Create the local retrieval layer used by the Artificial Teacher.
🎯 Objectives
• Generate embeddings.
• Store embeddings locally.
• Configure vector store.
• Implement similarity retrieval.
• Retrieve by topic.
🎯 Scope
Retrieval only.
🔗 Call Chain
Processed Content
↓
Embeddings
↓
Vector Store
↓
Retrieval Query
↓
Relevant Chunks
🚫 Out of Scope
• Gemma response generation
• Student API
🔗 Dependencies
• W2-MT-01
✅ Acceptance Criteria
• Vector store works offline.
• Relevant chunks are retrieved.
• Retrieval returns metadata.
• Topic-based retrieval works.
🌿 Branch
offline-rag-retrieval
🚨 Files Concerned
model/
├── embeddings/
│ └── embedding_service.py
└── retrieval/
└── retrieval_pipeline.py

backend/
├── data/vector_store/
└── app/ai/
└── retrieval_service.py