draft of synthetic question generation

WIP but added a small section on NDCG and MRR + how to calculate them and the lancedb code walkthrough on how to retrieve chunks from the vector db.

Next few steps is to write up the code and walkthrough for a script to generate ~30 questions, run retrieval with semantic search and then from there run evaluations.

	:rocket: This PR description was created by Ellipsis for commit 7557e4ddf05baf48aa2c3144d07ca2be22d6e0a0.

Summary:

This PR adds a new blog post on evaluating the quality of Retrieval-Augmented Generation (RAG) applications, with practical code snippets and explanations on using Instructor for synthetic data generation, lancedb for chunking and embedding, and NDCG and MRR as performance metrics.

Key points:

Added a new blog post 'Retrieval is more than pure chance'.
The post discusses the evaluation of RAG applications, use of Instructor for synthetic data generation, chunking and embedding with lancedb, and the use of NDCG and MRR metrics.
The post includes practical code snippets.
Updated docs/index.md and docs/writing/index.md to include the new post.

Generated with :heart: by ellipsis.dev

Latest commit:	`7557e4d`
Status:	✅ Deploy successful!
Preview URL:	https://1f575d1a.blog-cbl.pages.dev
Branch Preview URL:	https://synthethic-data.blog-cbl.pages.dev

jxnl / blog

draft of synthetic question generation #48

Summary:

Deploying blog with Cloudflare Pages