RAG pipeline with your own documents

From PDF upload to chunking and semantic retrieval: full setup guide for a Mycelis knowledge base.

Upload and structure

1. Group documents by topic.
2. Split large files into logical units.
3. Introduce versioning for document revisions.

Chunking

1. Match chunk size to content type.
2. Keep overlap only as high as needed.
3. Regularly review retrieval quality.

Retrieval and answer quality

1. Test top-k and reranking.
2. Return source references with answers.
3. Explicitly state when information is missing.