[Feature] Support external dataset imports (.xlsx and .json)

[nodesagar]

Currently DataLoom only supports uploading .csv files. GSoC 2026 ideas explicitly state supporting multi-source data. Allowing basic parsing of Excel and JSON files directly into Pandas eliminates user friction of manually converting their tables before upload.

[VishakBaddur]

Good suggestion. Worth noting that the existing upload pipeline in projects.py reads the file directly into pandas via pd.read_csv(). Supporting .xlsx and .json would require detecting the file extension and routing to pd.read_excel() or pd.read_json() respectively before the DataFrame is saved as a CSV internally.

One thing to be careful about with pd.read_json(): it defaults to orient="columns" which works for flat tables but breaks on nested JSON. The endpoint would need to either flatten nested structures or reject them with a clear error message, since the rest of the transformation pipeline assumes a flat DataFrame.

For .xlsx, openpyxl is the dependency pandas uses under the hood, it would need to be added to requirements.txt if not already present.