Bug: append_datasource hardcodes multipart form field name to 'csv', breaking Parquet and NDJSON file uploads

[zodiac12k]

Description

When using append_datasource to upload a local .parquet or .ndjson file, the Tinybird API rejects the request with the following error:

NDJSON/Parquet multipart requests require name field set to 'ndjson'/'parquet'. Example: curl -F "ndjson=@events.ndjson" ...

Root Cause

In tinybird_sdk/api/api.py (around line 275 in v0.4.0), the multipart form field name is strictly hardcoded to "csv", regardless of the actual file format being uploaded:

# tinybird_sdk/api/api.py
            with open(file_path_str, "rb") as fp:
                file_content = fp.read()
            content_type, multipart = create_multipart_body(
                files=[("csv", file_path_str, file_content, None)],  # 🐛 BUG: hardcoded to "csv"
            )

Even though detect_data_format(source_ref) correctly sets query["format"] = "parquet", the Tinybird API expects the multipart field name to be "parquet" as well, causing a schema validation failure on the server side.

Steps to Reproduce

from tinybird_sdk import create_tinybird_api

api = create_tinybird_api({"token": "your_token", "base_url": "https://api.tinybird.co"})

# Attempting to upload a valid parquet file
api.append_datasource(
    "krx_daily_stocks",
    {"file": "/path/to/data.parquet"}
)
# Raises TinybirdApiError: NDJSON/Parquet multipart requests require name field set to 'ndjson'/'parquet'.

Proposed Fix

The field name in create_multipart_body should be dynamically set based on the detected_format (e.g., "csv", "parquet", "ndjson").

            # Proposed Fix
            form_field_name = detected_format if detected_format else "csv"
            content_type, multipart = create_multipart_body(
                files=[(form_field_name, file_path_str, file_content, None)],
            )

Environment

• SDK Version: 0.4.0
• Python Version: 3.13