Skip to main content

Rate Limits

Request rate limits (QPS) for each RAG API endpoint and how to handle throttling.

Each endpoint has a request rate limit (QPS, requests per second) to protect service stability. When you go over, the endpoint returns HTTP 429 Too Many Requests.

Knowledge Base management endpoints

EndpointPathQPS limit
Retrieve/rag/index/retrieve2,000
Get Job Status/rag/index_job/status20
List Documents/rag/index/files15
Document Details/rag/list/index/file/details10
Create Knowledge Base and Import/rag/index/create_v210
Submit Import Job/rag/index/job/create10
List Knowledge Bases/rag/index/list10
Update Knowledge Base/rag/index/update10
Delete Knowledge Base/rag/index/delete10
Delete Document/rag/index/delete_file10
List Chunks/pipeline/{id}/chunklist10
Update Chunk/rag/index/chunk/update10
Delete Chunks/rag/index/chunk/delete10
Monitoring/rag/index/monitor1

Data Import endpoints

EndpointPathQPS limit
Describe File/describeFile10
Delete File/deleteFile10
Request Upload Lease/applyFileUploadLease10
Register File/addFile10
List Categories/listCategory5
Create Category/addCategory5
Delete Category/deleteCategory5
List Files/listFile5
Get Connector/getConnector5
Batch Update Tags/batchUpdateFileTag5
Create Connector/addConnector2
Import from OSS/addFilesFromAuthorizedOss5

Knowledge Search and Chat

Rate limits for Knowledge Search (/knowledge/search) and Knowledge Chat (/knowledge/chat) are not published yet — to be updated.

Handling rate limits

When you hit a rate limit, the endpoint returns:
{
  "code": "Throttling",
  "status_code": 429,
  "message": "Requests rate limit exceeded, please try again later.",
  "request_id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"
}
Recommendations:
  • On a 429, back off exponentially before retrying — for example, 1s → 2s → 4s.
  • Cap concurrency on batch operations so you don't burst past the limit.
  • Cache results client-side for low-QPS endpoints like Monitoring to cut down on calls.
If the default quota isn't enough for your workload, submit a ticket to request a higher limit.
Managed Agent API
Flow Agent API
Sandbox API
RAG API
Long-term Memory
ParseX API
Security API
Framework Integration