DevBook

About

DevBook is a personal software engineering knowledge base for learning and interview preparation. Notes are written in Obsidian and organized around programming, computer science, data, architecture, security, cloud, AI, and other engineering topics. Each note aims to explain an idea in plain language, then ground it with examples, tradeoffs, pitfalls, questions, and references.

The same Markdown collection also supports a public Quartz website and a local .NET retrieval proof of concept. This keeps the writing useful as ordinary notes while making it possible to explore publishing and retrieval over real, maintained material.

Highlights

• One Markdown source supports both the Obsidian vault and the published site.
• Quartz provides static pages, navigation, search, backlinks, and theme support.
• The .NET platform explores ingestion, chunking, embeddings, vector search, reranking, grounded answers, and retrieval evaluation.
• Retrieval configurations can be compared against a shared golden dataset instead of judged from a few hand-picked queries.

How it works

The repository has three main parts: Vault/ contains the authored notes, Web/ publishes them with Quartz, and Platform/ contains the local ASP.NET Core RAG API and evaluation projects.

For retrieval, published Markdown is ingested, divided by a selected chunking strategy, embedded, and stored in MongoDB Atlas. A search request embeds the query, retrieves candidates through vector search, and applies the configured reranker. The ask endpoint reuses that retrieval path before passing the selected sources to an answer agent.

Project details

• Content: Markdown notes authored in Obsidian
• Website: Quartz static site
• Retrieval platform: ASP.NET Core, MongoDB Atlas, and Microsoft.Extensions.AI
• Evaluation: retrieval-focused datasets, metrics, and Microsoft.Extensions.AI HTML reports
• License: MIT