A Processing Profile determines how documents are chunked and embedded during the semantic processing phase.
Chunking Strategies
- Recursive — Splits by paragraphs, then sentences
- Sentence — Splits at sentence boundaries
- Fixed Size — Splits at fixed character count
Embedding Strategies
- Hash — Deterministic, no API key required
- OpenAI — text-embedding-3-small via AI SDK