RAG QA

Perform retrieval-augmented generation QA based on uploaded knowledge base documents. The endpoint retrieves relevant document chunks, constructs context, and then calls the specified large language model to generate answers.

Try It

POST/search/chat_messagehttps://api-platform.ope.ai

Authentication

Uses Bearer Token authentication.

• Header: Authorization: Bearer <token>
• Example: Authorization: Bearer sk-xxxxxx

Request Body (application/json)

Field	Type	Required	Description	Default / Range
question	string	Yes	User question, used for document retrieval and answer generation	-
llm_model_name	string	Yes	Large language model name used to generate answers	-
embedding_model_name	string	Yes	Embedding model name used for text vectorization and retrieval	-
session_id	string	Yes	Session ID, used to associate conversation history	-
top_k	integer	No	Number of similar documents returned by retrieval	Default 5
stream	boolean	No	Whether to use SSE streaming response	Default true
return_sources	boolean	No	Whether to return relevant source documents	Default true
return_history	boolean	No	Whether to return conversation history records	Default false
extra_prompt	string	No	Additional prompt to supplement answer constraints	Default None
example_prompt	string	No	Example prompt to constrain response style	Default example
fallback_context	string	No	Backup context when no relevant content is found in the knowledge base	Default empty string
cosine_radius	number	No	Cosine similarity threshold radius	Default 0.55

Request Examples

Non-streaming cURL

curl -X POST "https://api-platform.ope.ai/search/chat_message" \
  -H "Authorization: Bearer $OPEAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "What is RAG?",
    "llm_model_name": "deepseek-v4-flash",
    "embedding_model_name": "bge-m3",
    "top_k": 5,
    "session_id": "session_1",
    "stream": false,
    "return_sources": true
  }'

Streaming cURL

curl -N -X POST "https://api-platform.ope.ai/search/chat_message" \
  -H "Authorization: Bearer $OPEAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "Summarize product advantages based on the knowledge base",
    "llm_model_name": "deepseek-v4-flash",
    "embedding_model_name": "bge-m3",
    "session_id": "session_1",
    "stream": true,
    "return_sources": true
  }'

JavaScript

const res = await fetch("https://api-platform.ope.ai/search/chat_message", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.OPEAI_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    question: "What is RAG?",
    llm_model_name: "deepseek-v4-flash",
    embedding_model_name: "bge-m3",
    top_k: 5,
    session_id: "session_1",
    stream: false,
    return_sources: true,
  }),
});

console.log(await res.json());

Response Example

Non-streaming

{
  "message": "RAG is Retrieval-Augmented Generation, which retrieves relevant content from the knowledge base and generates answers combined with the model.",
  "sources": [
    {
      "doc_id": "doc_1",
      "chunk_id": "chunk_1",
      "content": "RAG combines retrieval with generation."
    }
  ],
  "info": {
    "top_k": 5
  },
  "history": []
}

Streaming

event: message
data: {"content":"RAG is"}

event: sources
data: [{"doc_id":"doc_1","chunk_id":"chunk_1"}]

event: end
data: {}

Error Responses

Status Code	Scenario	Description
400	Parameter validation error	Missing or incorrect format for required fields like question, model, or session
401	Authentication failed	API Key is missing, invalid, or has insufficient permissions
422	Validation failed	Parameter structure validation failed
500	Processing failed	Database, retrieval, model call, or other server errors