Appearance

RAG Configuration ​

Per-document configuration for the retrieval pipeline. Controls how documents are chunked, how retrieval queries are executed, and how results are reranked. See the Chunking method, Retrieval mode, and Reranker type values in the overview.

Dry-run chunking preview from raw text (no persistence)​

POST
/v1/rag/chunks/preview

Chunk raw text with the supplied settings and return the resulting chunks
without storing anything. Use it to tune chunk size and overlap before
committing a configuration to a document.

Authorizations​

HTTPBearer
Type
HTTP (bearer)

Request Body​

application/json

Per-document chunking and embedding settings.

Chunking settings used by document ingestion and chunk preview.

Per-document embedding model selection.

Min Length1

Responses​

Successful Response

application/json
JSON
{
  
"chunks": [
  
  
{
  
  
  
"chunk_kind": "string",
  
  
  
"content": "string",
  
  
  
"contextual_text": "string",
  
  
  
"index": 0,
  
  
  
"token_count": 0
  
  
}
  
],
  
"total": 0
}

Samples​

Dry-run chunking preview for a stored document (no persistence)​

POST
/v1/rag/documents/{document_id}/chunks/preview

Same chunking behavior as POST /rag/chunks/preview, but content is read from the document and the body is just the chunking settings.

Authorizations​

HTTPBearer
Type
HTTP (bearer)

Parameters​

Path Parameters

document_id*
Type
string
Required
Format
"uuid"

Request Body​

application/json

Chunking settings used by document ingestion and chunk preview.

Chunking strategy: fixed-size token windows, or structure-aware
splitting that follows the document's own layout.

Valid values"token""semantic"
Default"token"
Default200
Minimum0
Default512
Minimum1

Responses​

Successful Response

application/json
JSON
{
  
"chunks": [
  
  
{
  
  
  
"chunk_kind": "string",
  
  
  
"content": "string",
  
  
  
"contextual_text": "string",
  
  
  
"index": 0,
  
  
  
"token_count": 0
  
  
}
  
],
  
"total": 0
}

Samples​

Chunk and embed a document​

POST
/v1/rag/ingest/{document_id}

Queues chunking then embedding for the document. By default (redo=false) work resumes from existing chunks and unembedded rows. Set redo=true to delete all chunks and re-run from scratch (e.g. after a chunking config change). Requires write access to the document.

Authorizations​

HTTPBearer
Type
HTTP (bearer)

Parameters​

Path Parameters

document_id*
Type
string
Required
Format
"uuid"

Request Body​

application/json

Options for POST /rag/ingest/{document_id}.

Default{ "redo": false }

When false (default), chunking and embedding resume from existing DB progress (partial chunks or missing embeddings). When true, delete all chunks and clear ingestion hashes before queueing work.

Defaultfalse

Responses​

Successful Response

application/json
JSON
{
  
"chunking_complete": false,
  
"document_id": "string",
  
"embedding_complete": false,
  
"ingest_job_id": "string",
  
"queued": true,
  
"upload_complete": false
}

Samples​