Read web pages

Use this pattern for lightweight research loops where you need to fetch a few web pages, clean the text, and summarize them quickly without building a full retrieval system.

It is a good fit for rapid validation, but also a useful reminder of when an ad-hoc workflow has reached the point where it should become a proper index.

What to focus on

• Getting clean page text (noise removal matters)
• Summarization constraints (length, citations, extraction)
• When to upgrade to a real RAG index instead of ad-hoc reads

When it works well

• You need fast external reading rather than a permanent knowledge base
• The number of pages is still small and manageable
• You want to test whether a web-based workflow is worth operationalizing

If you want to reproduce it locally

After your local Cookbook environment is ready, you can validate the matching example with:

make test-notebooks NOTEBOOK=misc/read_web_pages_with_haiku.ipynb