<div class="section-label"><span class="num">04</span> Payload Limits</div> <div class="section-title">Keep POST bodies under <em>10 MB</em>.</div> Every HTTP POST to the CSA Prediction Engine API must stay at or under **10 MB**. Larger bodies are rejected with **413 Payload Too Large**. This limit applies to direct integrations (`curl`, `requests`, and similar). The [[API/Python API Client/Introduction|v2 Python client]] can upload large input objects through an alternate channel before it posts the prediction task. > [!note] > If you call the API yourself, you are responsible for payload size. The Python client’s large-object path is not available to raw HTTP clients. ## Keep payloads small - Drop columns and rows you do not need for the task. - Encode compact JSON (no indentation, tight separators). - Prefer single-precision values — see [[API/Getting Started/Precision|Precision]]. ```python payload = json.dumps(variables, indent=None, separators=(',', ':'), cls=Float32Encoder) ``` ## If you receive 413 Confirm the encoded size before retrying. Trim the body, then post again. Watch payload size as N and K grow; a wide <span class="mono">X</span> is usually the dominant cost. ## Related - [[API/Introduction|API]] - [[API/Getting Started/Precision|Precision]] · [[API/Getting Started/Throttle Limits|Throttle Limits]] - [[API/Python API Client/Introduction|Python Client]]