Files, images and audio

Private files

Files supports upload, metadata, content retrieval and deletion for the caller's own files. Use the returned identifier in protocols supporting that attachment type.

curl --fail-with-body https://api.llm.netter.io/v1/files \
  -H "Authorization: Bearer $NETTER_API_KEY" \
  -F purpose=assistants \
  -F file=@document.txt

Uploading does not make a file public or establish that every model can read it. Resources retain tenant and ownership boundaries.

Text documents

UTF-8 text, text-based PDF and DOCX extraction happens locally in supported flows. PII protection applies to processed text fields. PDFs without a text layer are refused as text documents rather than automatically passed through OCR.

Images

Images require a compatible vision model and mapping for the selected Chat Completions, Messages or Responses format. Image bytes are forwarded unchanged and are not anonymized.

Verify vision support specifically; it does not follow from a model's presence in the text catalog.

Audio

Operation	Endpoint	Expected pricing basis
Transcription	/v1/audio/transcriptions	Audio duration verified locally
Translation	/v1/audio/translations	Audio duration verified locally
Speech synthesis	/v1/audio/speech	Text characters according to mapping

Audio is not anonymized. With PII enabled, TTS text containing recognized personal data is refused before dispatch: those values cannot be restored in remotely generated audio without resynthesizing it.

Limits

Surface	Documented limit
API files	20 MB
Portal upload	32 MB
Consulting ZIP	5,000 files, 64 MB expanded, 4 MB per file
ZIP paths	No symlinks or paths outside the archive

Individual mapping context/output limits and request size checks also apply. Do not bypass a limit by splitting content that the model needs to interpret together.

Read Privacy and the API contract before sending media.