[AnsgarSchmidt]

Summary

solar-mcp 0.2.1 breaks against the current NOAA SWPC endpoints (verified 2026-08-28 against live services.swpc.noaa.gov). This PR fixes all six data tools:

• conditions() — products/summary/10cm-flux.json now returns a list of dicts ([{"flux": 122, "time_tag": "..."}], lowercase keys) instead of a single dict with Flux/TimeStamp. products/noaa-planetary-k-index.json now returns dict rows ({"time_tag", "Kp", ...}) instead of list rows. The parser now handles both old and new shapes (backward compatible).
• solar_wind() — products/solar-wind/{mag,plasma}-5-minute.json are gone (HTTP 404). Switched to json/rtsw/rtsw_mag_1m.json and json/rtsw/rtsw_wind_1m.json (source is DSCOVR or ACE), parsing bz_gsm/bt and proton_density/proton_speed. Bz falls back to bz_gse when bz_gsm carries missing data.
• _to_float() — new helper: NOAA's -9999 sentinel (and JSON null) become None instead of leaking bogus values into results.
• _get_json() — strips trailing NUL bytes (\x00) that NOAA's file servers occasionally append (observed live on json/goes/primary/xrays-6-hour.json; broke xray() with a JSONDecodeError).

Testing

• 55 unit/security tests pass — 7 new (SOLAR-L2-046..052) covering the 2026 formats plus legacy-format backward compatibility
• 15 live integration tests pass against real NOAA endpoints (pytest tests/test_live.py --live)

Tested with Python 3.12, FastMCP 3.4.7.

[bob-qg]

Thank you, @AnsgarSchmidt! Your fix shipped in solar-mcp 0.2.4 (https://pypi.org/project/solar-mcp/0.2.4/). Your commit is in main unchanged, merged through #9, and credited in the CHANGELOG.

It had conflicts with 0.2.3, which fixed SFI and Kp separately, so we resolved them in #9. We also added one change on top: the rtsw files list the newest rows first, with one row per spacecraft per minute, so solar_wind() now takes the newest reading from the active spacecraft rather than the last row, which was a day old.

73!