The error "<https://arize.com/docs/phoenix/datasets-and-experiments/how-to-datasets/creating-datasets|HTTP/1.1 422 Unprocessable Entity]" when uploading a CSV dataset to Phoenix via Docker is likely due to the content type of your file upload being set as "application/vnd.ms-excel" instead of the expected "text/csv" or "multipart/form-data". Phoenix expects CSV uploads to use the correct content type, and mismatches can cause a 422 error. This is a known issue, especially when CSVs are created or saved with certain editors or when the browser or client sets the wrong MIME type during upload. Ensure your upload uses "text/csv" and, if possible, try uploading via the Phoenix UI or Python client, which handle content types correctly [[source 1>], <https://github.com/Arize-ai/phoenix/issues/3045|[source 2>].
If you continue to see this error, try re-saving your CSV with a different editor (e.g., Excel, VSCode, or a plain text editor with UTF-8 encoding) and ensure the file extension is ".csv". Also, uploading via the Python client with Pandas DataFrame or using the UI's file upload dialog (which now prioritizes CSV selection and auto-detects columns) can help avoid content-type issues <https://github.com/Arize-ai/phoenix/pull/8421|[source 3>].
Would you like more detail on how to upload via the Python client or troubleshoot further?
Sources:
- <https://arize.com/docs/phoenix/datasets-and-experiments/how-to-datasets/creating-datasets|source 1>
- <https://github.com/Arize-ai/phoenix/issues/3045|source 2>
- <https://github.com/Arize-ai/phoenix/pull/8421|source 3>
