OpenRAG is a comprehensive, single package Retrieval-Augmented Generation platform built on Langflow, Docling, and Opensearch.
Python
from utils.embedding_fields import build_knn_vector_field, get_embedding_field_name
from utils.logging_config import get_logger
logger = get_logger(__name__)
async def create_index_body(
embedding_model: str | None = None,
embedding_dimensions: int | None = None,
*,
embedding_provider: str | None = None,
embedding_space_id: str | None = None,
) -> dict:
"""Create a static index body configuration.
Returns:
OpenSearch index body configuration
"""
from config.embedding_constants import get_declared_default_embedding_model
from config.settings import (
ACL_PRINCIPAL_LABELS_MAPPING,
OPENSEARCH_NUMBER_OF_REPLICAS,
OPENSEARCH_NUMBER_OF_SHARDS,
VECTOR_DIM,
get_openrag_config,
)
config = get_openrag_config()
resolved_embedding_model = (
embedding_model
or config.knowledge.embedding_model
or get_declared_default_embedding_model(config.knowledge.embedding_provider)
)
properties = {
"document_id": {"type": "keyword"},
"filename": {"type": "keyword"},
"mimetype": {"type": "keyword"},
"page": {"type": "integer"},
"text": {"type": "text"},
# Legacy field - kept for backward compatibility and for clusters where
# Langflow cannot perform mapping updates with a DLS-filtered JWT.
"chunk_embedding": build_knn_vector_field(VECTOR_DIM),
# Track which embedding model was used for this chunk
"embedding_model": {"type": "keyword"},
"embedding_provider": {"type": "keyword"},
"embedding_space_id": {"type": "keyword"},
"embedding_dimensions": {"type": "integer"},
"source_url": {"type": "keyword"},
"connector_type": {"type": "keyword"},
"record_kind": {"type": "keyword"},
"web_source_id": {"type": "keyword"},
"web_page_id": {"type": "keyword"},
"web_page_depth": {"type": "integer"},
"web_child_count": {"type": "integer"},
"root_source_url": {"type": "keyword"},
"canonical_url": {"type": "keyword"},
"status": {"type": "keyword"},
"error": {"type": "text"},
"ingest_run_id": {"type": "keyword"},
# Object-store entity tag of the source file, for sync change detection.
"content_etag": {"type": "keyword"},
"owner": {"type": "keyword"},
"owner_email": {"type": "keyword"},
"allowed_users": {"type": "keyword"},
"allowed_groups": {"type": "keyword"},
"allowed_principals": {"type": "keyword"},
"allowed_principal_labels": ACL_PRINCIPAL_LABELS_MAPPING,
"created_time": {"type": "date"},
"modified_time": {"type": "date"},
"indexed_time": {"type": "date"},
"metadata": {"type": "object"},
}
if embedding_dimensions:
field_identity = embedding_space_id
if not field_identity and embedding_provider:
from utils.embedding_fields import get_embedding_space_id
field_identity = get_embedding_space_id(embedding_provider, resolved_embedding_model)
properties[get_embedding_field_name(field_identity or resolved_embedding_model)] = (
build_knn_vector_field(embedding_dimensions)
)
return {
"settings": {
"index": {"knn": True},
"number_of_shards": OPENSEARCH_NUMBER_OF_SHARDS,
"number_of_replicas": OPENSEARCH_NUMBER_OF_REPLICAS,
},
"mappings": {"properties": properties},
}