Skip to content

TECH_STACK.md

Latest commit

 

History

History
39 lines (32 loc) · 1.31 KB

TECH_STACK.md

File metadata and controls

39 lines (32 loc) · 1.31 KB
Raw

Tech Stack

Frontend

  • Next.js (App Router)
  • React
  • TypeScript
  • Tailwind CSS
  • Lightweight local UI components

Backend

  • Python
  • FastAPI
  • Pydantic (request/response models)
  • SQLAlchemy (ORM)
  • Alembic (migrations; likely added later)

Database

  • PostgreSQL
  • pgvector (vector column + similarity search)

RAG / AI layer (current)

  • Embeddings provider: OpenAI embeddings (text-embedding-3-small)
  • Chunking: server-side utilities to split text into retrieval-friendly segments
  • Retrieval: pgvector similarity search (cosine similarity, top-K, optional source filtering)
  • Modes: retrieval-only chunk inspection and grounded answer generation
  • Grounding format: answer + sources (retrieved chunk references)

Local development

  • Docker Compose (PostgreSQL + pgvector, backend, frontend)
  • Makefile or npm scripts (optional convenience; add later if useful)

Future optional AI integration layer

  • Provider-agnostic design to support:
    • Local embeddings/LLMs (e.g., via Ollama) as a future extension

Why Python/FastAPI for this RAG project

  • Python has the strongest ecosystem for embeddings, chunking, and evaluation tooling
  • FastAPI is lightweight, fast to iterate on, and encourages typed, well-defined APIs
  • Clear separation between web/API concerns and RAG domain logic keeps the project reviewable