Create Answer Engine

Create a new Answer Engine and index the provided documents.

This endpoint creates a new answer engine that can be used to answer questions based on the provided documents.

Form Fields (Required):

• answer_engine_name: Name for the engine
• model_llm: LLM model for answer generation (e.g. llama-v3p3-70b-instruct)
• model_embedding: Embedding model for retrieval (e.g. nugen-flash-embed)

Form Fields (Optional):

• model_reranker: Reranker model (e.g. bge-reranker-v2-m3)
• system_prompt: Custom system prompt for the engine
• temperature: LLM temperature (default: 0.2)
• max_tokens: Max tokens for LLM response (default: 600)
• documents: One or more files to index
• s3_uris: List of S3 URIs to index
• s3_prefix: S3 prefix to list and index JSON files from

Note: At least one of documents, s3_uris, or s3_prefix must be provided.

Returns:

• data: Newly created engine object containing:
  • answer_engine_id: Unique answer engine identifier (e.g. ae_8c625d9d71)
  • status: Engine status (processing while indexing, ready when done)
  • answer_engine_name: Name of the engine
  • user_id: User identifier
  • org_id: Organisation identifier
  • document_count: Number of indexed documents
  • chunk_count: Number of indexed chunks
  • error_message: Error message if indexing failed (null if no error)
  • created_at: ISO 8601 timestamp of creation

Errors:

• 400: Must provide at least one of documents, s3_uris, or s3_prefix
• 422: Missing required field

Example Request:

curl -X POST https://api.nugen.in/api/v3/answer-engine \
  -H "Authorization: Bearer <api_key>" \
  -F answer_engine_name="Test_quer_round" \
  -F model_llm="llama-v3p3-70b-instruct" \
  -F model_embedding="nugen-flash-embed" \
  -F model_reranker="bge-reranker-v2-m3" \
  -F system_prompt="You are a helpful assistant that answers strictly from the provided documents." \
  -F temperature="0.2" \
  -F max_tokens="600" \
  -F documents=@handbook.pdf \
  -F s3_uris="s3://<bucket>/<file_path>.json" \
  -F s3_prefix="s3://<bucket>/<prefix>/"

Example Response:

{
  "data": {
    "answer_engine_id": "ae_8c625d9d71",
    "status": "ready",
    "answer_engine_name": "Test_quer_round",
    "user_id": "user_415",
    "org_id": "org_415",
    "document_count": 1,
    "chunk_count": 5,
    "error_message": null,
    "created_at": "2026-03-13T15:11:51.105262+05:30"
  }
}

Notes:

• Engine status will be processing while documents are being indexed
• Poll GET /api/v3/answer-engine/{id} to check when status becomes ready
• Once ready, the engine can be used for chat and retrieval