PLAN.md

Python Search Engine – Build Plan

Duration: 45 Days

---

V3 - Core Features

Phase 1

• Create separate embedding service
• Explore and benchmark other sentence transformer models for later update
• Add separate search service
• Implement persistent Pinecone index that supports high speed & automatic recovery
• Implement search with URL sync and cached results in frontend
• Create user handler
• Use linting to ensure standard practices
• Benchmark current performance

Phase 2

• Add caching with FastAPI
• Add batch embeding

Phase 3

• Add analytics dashboard

Phase 4

• Add database for search

Phase 5

• Update the embedding model considering quality & speed

---

V4 - Go & Dockerization

Phase 1

• Convert user handler to Go

Phase 2

• Convert caching handler to Go

Phase 3

• Dockerize user handler
• Dockerize embedding service

Phase 4

• Dockerize caching + search
• Ensure ephemeral nature of images

---

Final Step

• Deploy to K8s or MicroK3s (V4.5)
• Redesign update client (V5)