RAG

RAG is the way for the model to answer about what it did not see in training. You upload documents to a knowledge base, the platform splits them, indexes them, and returns the chunks relevant to each question — which you inject into the prompt, or let an assistant retrieve on its own.

From document to answer

1. 1

   Create the knowledge base

   A POST https://api.hinow.ai/v1/knowledge-bases sets the embedding model and the chunk size.

2. 2

   Upload documents

   PDF, Markdown, HTML, or text. Indexing is asynchronous.

3. 3

   Search

   A POST https://api.hinow.ai/v1/knowledge-bases/{id}/search returns the chunks with the similarity score.

4. 4

   Answer

   Inject the chunks into the prompt, or attach the knowledge base to an assistant and let the search happen on its own.