Outline-first document retrieval beats chunking RAG in early tests
Linkly AI's Outlines Index offers an alternative to the standard RAG approach. Typical RAG splits documents into small chunks ahead of time, then feeds a batch of those chunks to the AI whenever a question comes in. Outlines Index instead mimics how a person reads: the agent first searches documents, checks the table of contents or outline, then reads only the relevant section.
This means the AI's context window is filled with targeted content rather than a pile of semi-random chunks. One user testing it hands-on found it noticeably cleaner and more satisfying than traditional chunking.
Key points
- Outlines Index skips upfront chunking; the agent reads the outline first, then fetches only the needed section
- Mirrors natural human reading behaviour rather than dumping random text chunks into the AI
- Reduces the amount of content loaded into the context window, cutting token costs
- Early hands-on impression rates it as cleaner than conventional RAG chunking
- Developed by Linkly AI; still at early-adopter evaluation stage
Quick term guide
- Outlines Index
- Linkly AI’s way of using a document’s structure to guide what the AI should read.
- table of contents
- A list of sections that helps you move through a long page or conversation.
- context window
- The amount of text an AI tool can remember and use in one chat.
- user testing
- Watching real users try a product so you can find what confuses them.
- token usage
- Token usage is a count of how much text an AI tool processes.
- performance
- How fast and smoothly a site loads and works.
- token costs
- Token costs are the fees paid for the text an AI model reads and writes.
- token cost
- The money or usage spent when sending text to an AI model and getting text back.