---
title: "Discussion on \"In, Store, Retrieve: The Three Places RAG Quietly Fails — And How We Fixed Each One\" | Hashnode"
description: "Discussion on \"In, Store, Retrieve: The Three Places RAG Quietly Fails — And How We Fixed Each One\". What five months of building a private AI document intelligence system taught me about retrieval quality, LLM reliability, and the gap between \"it works\" and \"it works well\""
url: "https://hashnode.com/posts/in-store-retrieve-the-three-places-rag-quietly-fails-and-how-we-fixed-each-one/69b94746c22d3eeb8aa93763"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [In, Store, Retrieve: The Three Places RAG Quietly Fails — And How We Fixed Each One](https://herald-ai.hashnode.dev/in-store-retrieve-the-three-places-rag-quietly-fails-and-how-we-fixed-each-one?utm_source=hashnode&utm_medium=feed)

What five months of building a private AI document intelligence system taught me about retrieval quality, LLM reliability, and the gap between "it works" and "it works well"

Mar 17 · 18 min read · herald-ai.hashnode.dev  
Read the article

-   [#ai-infrastructure](https://hashnode.com/tag/ai-infrastructure)
-   [#rag](https://hashnode.com/tag/rag)
-   [#llm](https://hashnode.com/tag/llm)
-   [#document-ai](https://hashnode.com/tag/document-ai)
-   [#docker](https://hashnode.com/tag/docker)
-   [#vector-database](https://hashnode.com/tag/vector-database)
-   [#vector-embeddings](https://hashnode.com/tag/vector-embeddings)
-   [#vectorsearch](https://hashnode.com/tag/vectorsearch)
-   [#vectors](https://hashnode.com/tag/vectors)
-   [#homelab](https://hashnode.com/tag/homelab)
-   [#machine-learning](https://hashnode.com/tag/machine-learning)
-   [#machinelearning](https://hashnode.com/tag/machinelearning)
-   [#bm25](https://hashnode.com/tag/bm25)
-   [#hybrid-rag-systems](https://hashnode.com/tag/hybrid-rag-systems)
-   [#retrieval-augmented-generation](https://hashnode.com/tag/retrieval-augmented-generation)

## Written by

[Ravi Mohta](https://hashnode.com/@ravimohta)

@ravimohta · London, UK

  
[View profile](https://hashnode.com/@ravimohta)

AI Architect/Engineer | Building private, local AI systems | RAG, LLMs, Vector Databases, Agents

## Responses

Markdown supported · ⌘↵ to post

No responses yet.