Tag: rag
All the articles with the tag "rag".
-
The Full Flow of RAG Data Preparation — Selection, Cleansing, Chunking, Metadata, and Evaluation Sets
While working on the FEMS project, I got curious about 'what and how should go into a vector DB.' Key insight — data should be selected backward from 'questions that need to be retrieved,' not from the 'domain' criterion. Also covers cleansing / the effect of metadata (document title, section path) before chunks / OCR preprocessing for analog data / building an evaluation set — the full flow of the RAG data pipeline.
-
FEMS Project #2 — A Real 230-Page Corpus + Chunk Quality Gates + the Chinese-Character Mixing Trap in a Local LLM
Built a corpus of roughly 230 pages / 35,000 rows from the Korea BEMS Association guides, Korea Energy Agency materials, and the UCI Steel dataset. After paragraph-based chunking (target 800 chars) + bge-m3 + Chroma indexing, a quality gate (ratio of complete Hangul/ASCII characters) excluded 7 chunks from table-of-contents pages. Then I hit a trap — qwen2.5:7b mixed in Chinese characters on the second question and suffered generation collapse (spitting out unrelated Chinese city coordinates as GeoJSON). Partially fixed with temperature / system prompt → ultimately switched to exaone3.5:7.8b for clean handling.
-
FEMS Project #1 — Comparing Low-Spec Local Setup (Ollama + bge-m3 + Chroma) vs Claude API RAG
Building a RAG prototype while studying the FEMS (Factory Energy Management System) domain. Comparing local LLM inference (Ollama) on a low-spec environment (GTX1660 Super, 6GB VRAM) against calling the Claude / OpenAI APIs. Using bge-m3 for embeddings (strong Korean support) and Chroma as the vector DB. Ollama's cold start of 95 seconds dropped to 10 seconds after warm-up, with accuracy matching the cloud.
-
Building a RAG Q&A System with Chroma + My Blog — From Parts to System
Yesterday I played with the parts of RAG (embedding / cosine similarity / chunks). Today, I put them together into a system. I introduced the Chroma vector DB, indexed 270 chunks from my blog, and built a Q&A system with OpenAI + Claude. The most memorable moment was when it answered 'I don't know' to a question about information not covered on my blog.