do not look bad until the answer comes out (classic OCR is the textbook case: EasyOCR recovers the words and quietly drops the table structure around them, and the answer reads fine until you check it against the page). A deterministic check catches the obvious failures for free, but a table that parsed into plausible-looking…
(A production RAG pipeline for PDFs: relational parsing, TOC retrieval, typed answers) we upgraded each of the four bricks: document parsing, question parsing, retrieval, and generation, and wired them into one clean, linear pipeline. A PDF goes in, and a typed, cited answer comes out. That was one paper and one question. The real test…
A user asks the pipeline: “what is the premium?” on a fifty-page insurance policy. The parser looks at the doc profile it just got from the parsing brick: doc_type: insurance_policy n_pages: 47 toc_df names General Information, Coverages, Exclusions, Endorsements. No section called Premium. Naive top-k embeds “premium” and scans the whole document. It returns exclusion…
and expensive. The fast parser misses the table the answer lives in. Run the heavy one on every page of a 200-page report and you pay for 200 pages to rescue three. Run the cheap one everywhere and the single flattened table sinks the answer. Neither setting is right for the whole document, because the…
talk about context engineering today, they usually mean retrieving the right context from a document: chunk selection, hybrid search, reranking, TOC-aware retrieval. That is one half of the story. The other half is that the question itself is context the LLM will see, and the question deserves the same treatment as the retrieved passage. Take…
then you’ve already been through this situation when it looks like your RAG system is working fine but it is still wrong. The retrieval fetches some chunks, passes it to the generative model, and it writes a fluent answer. Nothing throws an error, and yet the answer might be built on the wrong document, or…
a confident wrong answer, the instinct is to reach for the generation knobs: a stricter prompt, a bigger model, a temperature of zero. That treats a retrieval failure as a generation failure. The model was handed the wrong pages and did its job on them. Retrieval decides what the model can invent, because it decides…
application we build is a RAG app. The recipe is simple: chunk, embed, retrieve, then answer. It looks clean on paper. But once you use it on real cases, things get messy very quickly: similarity search finds similar wordings but not necessarily useful chunks, the right evidence never shows up in the retrieved context as…
, I have written quite a lot about RAG, starting with the Hitchhiker’s Guide to RAG with ChatGPT API and LangChain, and then exploring various topics related to RAG and AI, like chunking, hybrid search, reranking, contextual retrieval, and a three-part series on evaluating retrieval quality. In other words, we have covered a lot of…
Retrieval-Augmented Generation (RAG) has evolved rapidly in recent times. The original RAG paradigm was designed to be straightforward: retrieve the most relevant chunks from a corpus and use them to answer a question. While effective for many document-centric queries, this approach begins to struggle when questions extend beyond the boundaries of a single document. Or…