Shape your requests
Learn how arguments, pagination, filtering, field selection, responses, and errors work with the Langfuse API on MindCloud.
Once your first Langfuse request works, use this page to shape real production requests. MindCloud keeps the app-specific API details behind one REST format, so you can control arguments, pagination, filtering, fields, and errors without learning a separate convention for every provider. If you have not connected Langfuse yet, start with Connect to Langfuse.
Arguments
Arguments are the inputs an action needs. Pass them as flat fields beside the Universal API fields: query parameters for read and delete actions, JSON body fields for create and update actions. Each Langfuse action page documents its exact argument keys, types, and which ones are required. Keys are case-sensitive, and requests with missing or invalid required arguments fail instead of guessing.
Request format
The REST format is intentionally plain. Choose a Langfuse action, call its Universal API URL, authenticate with your MindCloud API Key, and include the connection and action fields in the correct lane for the HTTP method.
| Piece | Where it goes |
|---|---|
| Action fields | Query parameters for GET and DELETE; JSON body fields for POST, PUT, and PATCH. |
connectionId | Query string for reads and deletes; body for creates and updates. |
limit, offset, where, sort, fields | Universal API controls; always query parameters, never body fields. |
Pagination
Paginated list actions accept limit and offset as query parameters. Some providers use page numbers, some use cursors; MindCloud translates your request into whatever Langfuse expects, so you page through results the same way on every app. Start at offset=0, add limit to the offset after each page, and stop when a page returns fewer rows than you asked for.
curl --request GET \
--url "https://connect.mindcloud.co/v1/universal/langfuse/latest/actions/get-annotation-queue" \
--header "Authorization: Bearer $MINDCLOUD_API_KEY" \
--get \
--data-urlencode "connectionId=$CONNECTION_ID" \
--data-urlencode "limit=25" \
--data-urlencode "offset=0"These Langfuse actions support pagination:
- List Annotation Queue Items
- List Annotation Queues
- List Comments
- List Dataset Items
- List Dataset Run Items
- List Dataset Runs
- List Datasets
- List Models
- List Prompts
- List Score Configs
- List Scores
- List Sessions
- List Traces
Field selection
Use the fields query parameter to return only the response fields your code reads, with dot notation for nested objects: fields=id,name,profile.email. Selection is applied to each row in data after the response is mapped, and id is kept whenever it exists so rows stay identifiable.
Responses and errors
Every Langfuse response is mapped into the same public envelope before it reaches your code: success, your rows in a data array (single-row actions return an array with one item), and optional meta. Failed requests return success: false with a stable code to branch on and a human-readable message:
{
"success": false,
"code": "CONNECTION_APP_MISMATCH",
"message": "Connection \"conn_123\" is not configured for this app."
}Check the HTTP status first: 401 or 403 for authentication and authorization, 400 for invalid request shape, and 5xx when MindCloud cannot complete the action upstream. Most failures come from a missing API Key, a missing or mismatched connection, or arguments that do not match the action page.