A live Google Trends board is not a single astronomy series. At 2026-09-03T10:01:36Z, get_top_trends ranked lunar eclipse tonight 6. get_growth on Google Search for the week of 2026-08-29 returned score 100.0 for that exact phrase. YouTube for the same keyword scored 33.0. Google News scored 23.0, down from 100.0 on 2026-08-15, the week solar eclipse Search sat at 46.0. Wikipedia has no series for lunar eclipse tonight. It does have a monthly Lunar eclipse page at 168,134 views for 2026-08-01. A job that stores one row named "eclipse" mixes a tonight-shaped query, a shorter lunar token, a blood-moon token, and a solar control that already crashed. Split the keywords before the write.
What did the live boards show on 2026-09-03?
get_top_trends with type Google Trends, limit 25, as_of 2026-09-03T10:01:36Z:
| rank | keyword |
|---|---|
| 1 | tim curry |
| 2 | peter cullen |
| 3 | nepal flood |
| 6 | lunar eclipse tonight |
| 12 | tropical depression edouard |
| 16 | gta 6 release date |
The 7D rank_change sort on the same board, baseline 2026-08-27T02:01:32Z, returned 6 movers. mica miller climbed from prev_rank 142 to rank 5, rank_change +137. tim curry moved from 2 to 1, rank_change +1. lunar eclipse tonight was not in that mover list. Rank 6 with no 7D rank_change means the token was already high enough on 2026-08-27 that a rank delta did not fire, not that Search was quiet. Pair board rank with get_growth.
The 1D rank_change sort, baseline 2026-09-02T02:01:31Z, also omitted the eclipse token. Lead 1D movers were dyson toothbrush (+109, rank 40 vs 149) and pete hegseth pentagon leadership changes (+105, rank 47 vs 152). A daily climber job would have missed the eclipse. A weekly growth job would not.
Wikipedia Trending at the same timestamp had no lunar row in the top 15 (lead row Killing of the Clancy children, then .xxx, Toxic (2026 film)). Google News Top News, limit 15, also omitted eclipse tokens (lead rows were a Clancy jury deadlock, a ship deployment, and Iran fire). TikTok Trending Searches, limit 15, was iPhone 18 render rows and Nepal landslide tokens. Keyword TikTok and Reddit series for lunar eclipse tonight both returned 404 no_data. A monitor that only reads those live boards misses the Search 100.
Source fields for the Search board are on Google Trends. Wikipedia board fields are on Wikipedia Trends. The loop that stores board rank plus growth is on Demand sensing API.
What did get_growth return for lunar eclipse tonight?
get_growth on Google Search, keyword lunar eclipse tonight, recent_date 2026-08-29, recent_value 100.0.
| period | baseline date | baseline score | growth |
|---|---|---|---|
| 7D | 2026-08-22 | 0.0 | 999999 |
| 14D | 2026-08-15 | 11.0 | +809.09% |
| 30D | 2026-08-01 | 1.0 | +9,900% |
| 3M | 2026-05-30 | 1.0 | +9,900% |
| 12M | 2025-08-30 | 0.0 | 999999 |
| YTD | 2026-01-03 | 0.0 | 999999 |
The 7D, 12M, and YTD windows baselined at 0.0 and returned growth 999999. Print the raw 100.0 score. Do not store 999999 as a feature. The usable short window is 14D: 100.0 vs 11.0, +809.09 percent. get_time_series on the same keyword puts that 100 in a five-year weekly pull. Prior event weeks never cleared 100.
| week | score |
|---|---|
| 2021-11-20 | 26 |
| 2022-05-21 | 95 |
| 2022-11-12 | 30 |
| 2024-09-21 | 42 |
| 2025-03-15 | 70 |
| 2025-09-13 | 61 |
| 2026-03-07 | 18 |
| 2026-08-15 | 11 |
| 2026-08-22 | 0 |
| 2026-08-29 | 100 |
The 2026-08-29 week is the series peak at 100. The 2022-05-21 week at 95 is the next highest. Treat 100 vs 95 as the finding, not the 999999 label against a 0.0 baseline.
How do YouTube, News, and Shopping disagree on the same week?
They do not share a peak week, and they do not share a 0-100 scale.
YouTube lunar eclipse tonight scored 33.0 on 2026-08-29 vs 0.0 on 2026-08-22 (growth 999999, data_quality low, over 90 percent zeros). 14D was +200.0 percent vs 11.0 on 2026-08-15. 30D, 3M, and 12M also baselined at 0.0 with 999999. Video interest moved, but it did not hit 100 with Search. Store 33.0 next to Search 100.0. Do not average them.
Google News for lunar eclipse tonight scored 23.0 vs 4.0 on 2026-08-22, 7D +475.0 percent. 14D was -77.0 percent vs 100.0 on 2026-08-15. 30D was +475.0 percent vs 4.0. 3M and 12M baselined at 0.0 with 999999 and a low-quality warning. News peaked two weeks before Search. That 2026-08-15 News 100.0 is the same week solar eclipse Search scored 46.0. A job that reads only News on 2026-08-29 would report a crash. A job that reads only Search would report a peak.
Google Shopping for lunar eclipse tonight scored 22.0 on 2026-08-29. 7D, 14D, 30D, and 3M all baselined at 0.0 with growth 999999 and the same low-quality warning. Print 22.0. Shopping intent exists. It is not a volume figure.
Shorter token lunar eclipse on YouTube scored 39.0 vs 3.0 on 2026-08-22, 7D +1,200 percent. 14D was +550.0 percent vs 6.0. 30D was +1,850 percent vs 2.0. 12M was +1,850 percent vs 2.0. The shorter token is the cleaner YouTube series. It still did not hit 100. News for lunar eclipse scored 42.0 vs 6.0 on 7D (+600.0 percent) and 42.0 vs 100.0 on 14D (-58.0 percent). Same two-week News lead as the tonight-shaped query.
What do lunar eclipse, blood moon, and solar eclipse do on Search?
Call each exact phrase. Do not substring-match eclipse.
| keyword | week | score | 7D baseline | 7D growth | 14D baseline | 14D growth |
|---|---|---|---|---|---|---|
| lunar eclipse tonight | 2026-08-29 | 100.0 | 0.0 | 999999 | 11.0 | +809.09% |
| lunar eclipse | 2026-08-29 | 66.0 | 9.0 | +633.33% | 16.0 | +312.5% |
| blood moon | 2026-08-29 | 17.0 | 4.0 | +325.0% | 3.0 | +466.67% |
| solar eclipse | 2026-08-29 | 2.0 | 2.0 | 0.0% | 46.0 | -95.65% |
lunar eclipse Search 12M was +2,100 percent vs 3.0 on 2025-08-30. 30D was +3,200 percent vs 2.0. 3M was +3,200 percent vs 2.0. YTD was +6,500 percent vs 1.0. get_time_series on lunar eclipse puts the current 66 below the series peak of 100 on 2025-09-13. Other event weeks: 2022-05-21 scored 98, 2025-03-15 scored 68, 2026-03-07 scored 54, 2026-08-15 scored 16, 2026-08-22 scored 9. The tonight-shaped query is at its own 100. The shorter token is at 66. Those are different 0-100 scales. Do not rank 100 against 66.
blood moon is a related folk token, not a substitute. 17.0 vs 4.0 on 7D is a move. It is not the live-board phrase. solar eclipse is the control. 2.0 vs 2.0 on 7D is flat. 14D vs 46.0 on 2026-08-15 is -95.65 percent. 12M vs 0.0 returned 999999. A calendar job that only watches eclipse as a substring would have mixed a solar crash into a lunar peak.
What did Wikipedia return, and what should a calendar job store?
Wikipedia keyword lunar eclipse tonight returned 404 no_data. Keyword Lunar eclipse resolved and is monthly. August 2026 scored 43.4 with 168,134 page views vs July 1.4 / 20,894, 30D growth +3,000 percent, volume growth +704.7 percent. 3M vs May 0.9 / 19,025 was +4,722.22 percent. 12M vs 2025-08-01 (4.5 / 31,790) was +864.44 percent. 7D and 14D returned collapsed_range. Do not join those August page views onto Search week 2026-08-29. The Wikipedia grain is a month. Search grain is a week.
TikTok keyword series 404 and Reddit keyword series 404 are findings. Absence of a series is not a zero. Store data_unavailable and keep the Search row.
A weekly job that POSTs get_growth for lunar eclipse tonight, lunar eclipse, blood moon, and solar eclipse on Google Search, plus YouTube and News for the tonight-shaped query, is 4 Search + 2 YouTube + 2 News = 8 calls. Add Wikipedia Lunar eclipse and the total is 9. A Monday cron stays inside the 100-request free tier. Daily polling of that set is about 270 calls, which exceeds free. Only HTTP 200 responses count. Free tier 100 per month. Starter 5,000. Pro 25,000. Business 100,000.
body in the response is a JSON string and must be parsed a second time. The Python POST that materializes those rows into a warehouse is on dbt Trends API. The same urllib helper without dbt is on Python Trends API.