voyage-context-4 takes the chunking decision out of RAG indexing
- Source
- x.com
- Date
You should not have to think about chunking. voyage-context-4 makes that possible. With auto-chunking, no 32K ceiling, and built-in overlap support, it delivers 2.1% better retrieval than voyage-context-3 at 33% lower cost. A strong fit for retrieval over long documents and agent context engineering. Learn more:
Context
Voyage AI's voyage--4 moves document splitting into the service. Developers can submit a whole document and request separate, context-aware representations for its passages, extending the approach introduced in voyage-context-3. The release article explains that documents beyond 32,000 are split and processed automatically.
The removed ceiling therefore concerns what the service accepts, rather than an unlimited amount of text processed together in one model pass. Developers can still specify target passage size and overlap, or supply their own pre-split passages. The change reduces the separate preparation code needed to index long documents while retaining those controls.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
- embedding — A list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Checking sign-in…
Loading comments…