RAGFlow Knowledge Base API Handbook

Problem Being Solved

When integrating RAGFlow, the harder part is not calling one endpoint, but keeping parameters consistent across knowledge bases, documents, chunks, retrieval, metadata, and parser settings. Manually scanning docs can miss key fields such as parser_config, chunk_method, and authentication differences, causing upload parsing failures or retrieval results that do not match expectations.

How the Skill Works

This RAGFlow Knowledge Base API Handbook organizes the API reference by workflow:
- Knowledge base management: covers creation, listing, editing, and deletion, with settings such as embedding_model, permission, and chunk_method.
- Document management: describes file upload, URL import, listing, editing, downloading, and deletion, emphasizing dataset_id, document_id, and processing status run.
- Parsing and chunking: explains chunking methods such as naive, paper, table, qa, and parent_child, plus parser_config fields for token size, delimiters, layout recognition, and table/image context.
- Retrieval and stats: includes knowledge-base retrieval, usage statistics, re-embedding, and error-code notes for debugging recall and parsing issues.

Boundaries and Cautions

The material is an API reference and does not replace server-side permission governance. Some endpoints use session or beta token authentication, so confirm the credential type before calling. Changes to chunk_method or parser_config often require re-parsing to take effect; deleting a knowledge base cascades to documents, chunks, files, and vector indexes, so use caution in production.