Request body size limit

maxRequestBodyMB caps the size, in megabytes, of a request body sent to a model endpoint such as /v1/chat/completions, /v1/embeddings or /v1/audio/transcriptions. The default is 250.

maxRequestBodyMB: 1024

A request over the limit fails with HTTP 413 and the message request body exceeds the 250 MB limit (showing your configured value). The request never reaches the model and no model is loaded for it.

The limit counts the bytes actually received, so it also applies to chunked uploads that send no Content-Length header. A request with a Content-Length over the limit is rejected before any of the body is read.

What goes wrong

• Large uploads fail with 413. Long audio files and requests with many base64-encoded images are the usual cause. Base64 makes an image about a third larger than the file on disk. Raise maxRequestBodyMB.
• Memory use grows with the limit. llama-swap reads request bodies into memory to find the model field and to apply filters, so one request can use several times its own size in RAM. Raising the limit lets each request use more memory, and concurrent large requests add up.
• 0 or a negative value is rejected when the config loads. There is no setting for "no limit".
• A 413 from somewhere else. A reverse proxy in front of llama-swap (nginx client_max_body_size, for example) or the upstream server can have its own, smaller limit. If the error message is not the one above, look there.