LLM-based chunking (Roadmap)

Important

This chunker is planned and not available in the current public API.

Goal

LLM-based chunking is intended for advanced segmentation where chunk boundaries depend on instruction-following or task-specific semantic interpretation.

Planned behavior

  • Use model-guided boundary proposals.

  • Support strategy prompts for domain-specific chunking.

  • Provide deterministic fallbacks when model access is unavailable.

Current status