Building a Multilingual RAG Document Assistant with FastAPI, FAISS and Ollama
imABDRAOUF · reddit · 2026-09-12
To learn RAG systems, the author built and open-sourced CloudRAG (GitHub: abderaoufsec/CloudRAG).
Flow: documents → chunking → embeddings → FAISS → retrieved chunks → Ollama → answer. It supports PDF, TXT and DOCX and returns the sources used. Stack: Python, FastAPI, React, SentenceTransformers, FAISS, Ollama, SQLite, Docker.
The author added retrieval evaluation and testing to check retrieval quality rather than just whether an answer was produced; a Qdrant Cloud deployment failed, so the working version uses local FAISS. Biggest takeaway: RAG isn't just the LLM — chunking, embeddings and especially retrieval quality matter enormously.
More from coding & agent
- Sourcegraph CEO Praises Meta's Muse Spark Coding Model and CLI — zeeg · 2026-09-12
- ClaudeAI Weekly Survival Guide: GPT-6 Astra, Usage Resets, and Context Rot Rules — ClaudeAI-mod-bot · 2026-09-12
- Everyone who tried the AI agent tool spawn this week is hooked, devs report — TAbrodi · 2026-09-12
- Veteran Dev With ADD Says AI Pair Coding Finally Fits His Brain — Interesting-Town-433 · 2026-09-12
- This Founder Runs a Business With 5 Grok Bots and a Chief Organizer Bot — cyberlyons · 2026-09-12
- 'Staff' once meant your own walking stick — AI agents should be loyal to you, not your company — granawkins · 2026-09-12