Dataset Viewer
Auto-converted to Parquet Duplicate
country_fips
stringclasses
108 values
cameo_root
int64
1
20
cameo_root_ja
stringclasses
20 values
event_count
int64
1
173
goldstein_mean
float64
-10
10
observed_at
stringdate
2026-11-22 00:00:00
2026-11-22 00:00:00
US
1
ๅ…ฌใฎ็™บ่จ€
173
-0.094
2026-11-22
US
4
ๅ”่ญฐ
127
2.547
2026-11-22
SO
4
ๅ”่ญฐ
86
2.507
2026-11-22
US
5
ๅค–ไบคๅ”ๅŠ›
70
4.35
2026-11-22
US
2
ๅ‘ผใณใ‹ใ‘
67
2.979
2026-11-22
IS
4
ๅ”่ญฐ
64
3.161
2026-11-22
US
17
ๅผทๅˆถ
58
-5.493
2026-11-22
EG
4
ๅ”่ญฐ
51
4.547
2026-11-22
US
3
ๅ”ๅŠ›ใฎๆ„ๆ€
50
4.156
2026-11-22
US
11
้ž้›ฃ
49
-2
2026-11-22
IN
4
ๅ”่ญฐ
48
2.577
2026-11-22
KE
4
ๅ”่ญฐ
41
2.317
2026-11-22
US
19
ๆˆฆ้—˜
36
-9.986
2026-11-22
PK
4
ๅ”่ญฐ
35
3.94
2026-11-22
UK
4
ๅ”่ญฐ
35
2.474
2026-11-22
CA
1
ๅ…ฌใฎ็™บ่จ€
34
0.118
2026-11-22
SO
1
ๅ…ฌใฎ็™บ่จ€
34
-0.115
2026-11-22
IN
5
ๅค–ไบคๅ”ๅŠ›
30
3.867
2026-11-22
DJ
4
ๅ”่ญฐ
29
2.79
2026-11-22
IR
19
ๆˆฆ้—˜
27
-9.981
2026-11-22
CH
4
ๅ”่ญฐ
26
2.858
2026-11-22
US
8
่ญฒๆญฉ
25
6.48
2026-11-22
CA
5
ๅค–ไบคๅ”ๅŠ›
24
4.946
2026-11-22
IS
1
ๅ…ฌใฎ็™บ่จ€
23
0.122
2026-11-22
RP
4
ๅ”่ญฐ
23
1.978
2026-11-22
RS
4
ๅ”่ญฐ
23
3.087
2026-11-22
SO
5
ๅค–ไบคๅ”ๅŠ›
23
5.248
2026-11-22
ET
11
้ž้›ฃ
22
-2
2026-11-22
IR
4
ๅ”่ญฐ
22
4.805
2026-11-22
US
9
่ชฟๆŸป
22
-2
2026-11-22
US
12
ๆ‹’ๅฆ
22
-4
2026-11-22
IS
19
ๆˆฆ้—˜
21
-9.976
2026-11-22
SO
11
้ž้›ฃ
20
-2
2026-11-22
SA
4
ๅ”่ญฐ
19
2.8
2026-11-22
TU
4
ๅ”่ญฐ
19
2.974
2026-11-22
EG
5
ๅค–ไบคๅ”ๅŠ›
18
3.928
2026-11-22
UK
1
ๅ…ฌใฎ็™บ่จ€
18
-0.156
2026-11-22
NI
4
ๅ”่ญฐ
17
2.271
2026-11-22
IN
1
ๅ…ฌใฎ็™บ่จ€
16
0.4
2026-11-22
PO
4
ๅ”่ญฐ
16
3.588
2026-11-22
SO
3
ๅ”ๅŠ›ใฎๆ„ๆ€
16
4.594
2026-11-22
SU
4
ๅ”่ญฐ
16
1.881
2026-11-22
US
7
ๆดๅŠฉ
16
7.3
2026-11-22
ET
4
ๅ”่ญฐ
15
3.14
2026-11-22
UK
2
ๅ‘ผใณใ‹ใ‘
15
3.08
2026-11-22
CA
4
ๅ”่ญฐ
14
3.4
2026-11-22
IS
2
ๅ‘ผใณใ‹ใ‘
14
3.071
2026-11-22
RS
3
ๅ”ๅŠ›ใฎๆ„ๆ€
14
4.086
2026-11-22
TU
5
ๅค–ไบคๅ”ๅŠ›
14
7.343
2026-11-22
UK
5
ๅค–ไบคๅ”ๅŠ›
14
4.971
2026-11-22
AS
4
ๅ”่ญฐ
13
2.315
2026-11-22
BG
4
ๅ”่ญฐ
13
2.385
2026-11-22
IR
1
ๅ…ฌใฎ็™บ่จ€
13
-0.708
2026-11-22
NI
1
ๅ…ฌใฎ็™บ่จ€
13
0
2026-11-22
UK
3
ๅ”ๅŠ›ใฎๆ„ๆ€
13
4
2026-11-22
UP
4
ๅ”่ญฐ
13
2.177
2026-11-22
US
13
ๅจๅš‡
13
-4.6
2026-11-22
CA
8
่ญฒๆญฉ
12
5.333
2026-11-22
IS
5
ๅค–ไบคๅ”ๅŠ›
12
4.683
2026-11-22
PE
4
ๅ”่ญฐ
12
2.3
2026-11-22
RS
19
ๆˆฆ้—˜
12
-9.917
2026-11-22
US
10
่ฆๆฑ‚
12
-5
2026-11-22
US
16
้–ขไฟ‚ๆ–ญ็ตถ
12
-4.583
2026-11-22
IS
8
่ญฒๆญฉ
11
8.364
2026-11-22
MX
4
ๅ”่ญฐ
11
4.382
2026-11-22
PK
3
ๅ”ๅŠ›ใฎๆ„ๆ€
11
4
2026-11-22
RS
11
้ž้›ฃ
11
-2
2026-11-22
CH
5
ๅค–ไบคๅ”ๅŠ›
10
3.44
2026-11-22
EG
1
ๅ…ฌใฎ็™บ่จ€
10
0.04
2026-11-22
KS
15
ๆญฆๅŠ›ใฎๆง‹ใˆ
10
-7.2
2026-11-22
SO
10
่ฆๆฑ‚
10
-5
2026-11-22
SO
12
ๆ‹’ๅฆ
10
-4.1
2026-11-22
UK
11
้ž้›ฃ
10
-2
2026-11-22
US
6
็‰ฉ่ณช็š„ๅ”ๅŠ›
10
6.12
2026-11-22
BY
4
ๅ”่ญฐ
9
2.1
2026-11-22
CH
1
ๅ…ฌใฎ็™บ่จ€
9
-0.044
2026-11-22
IN
2
ๅ‘ผใณใ‹ใ‘
9
3.111
2026-11-22
IN
3
ๅ”ๅŠ›ใฎๆ„ๆ€
9
4.133
2026-11-22
SA
5
ๅค–ไบคๅ”ๅŠ›
9
4.456
2026-11-22
BY
1
ๅ…ฌใฎ็™บ่จ€
8
0.1
2026-11-22
ER
4
ๅ”่ญฐ
8
2.35
2026-11-22
ET
1
ๅ…ฌใฎ็™บ่จ€
8
-0.325
2026-11-22
IR
2
ๅ‘ผใณใ‹ใ‘
8
2.5
2026-11-22
IR
11
้ž้›ฃ
8
-2
2026-11-22
IS
12
ๆ‹’ๅฆ
8
-4
2026-11-22
IT
4
ๅ”่ญฐ
8
2.35
2026-11-22
KE
1
ๅ…ฌใฎ็™บ่จ€
8
0
2026-11-22
LE
1
ๅ…ฌใฎ็™บ่จ€
8
0.05
2026-11-22
LE
2
ๅ‘ผใณใ‹ใ‘
8
2.588
2026-11-22
LE
4
ๅ”่ญฐ
8
2.013
2026-11-22
MX
8
่ญฒๆญฉ
8
7
2026-11-22
PL
4
ๅ”่ญฐ
8
2.013
2026-11-22
UK
8
่ญฒๆญฉ
8
7.25
2026-11-22
CF
4
ๅ”่ญฐ
7
3.014
2026-11-22
CH
19
ๆˆฆ้—˜
7
-9.929
2026-11-22
DJ
1
ๅ…ฌใฎ็™บ่จ€
7
0
2026-11-22
FR
4
ๅ”่ญฐ
7
2.886
2026-11-22
IN
11
้ž้›ฃ
7
-2
2026-11-22
IS
6
็‰ฉ่ณช็š„ๅ”ๅŠ›
7
6.571
2026-11-22
KE
3
ๅ”ๅŠ›ใฎๆ„ๆ€
7
4.171
2026-11-22
End of preview. Expand in Data Studio

World Signals โ€” a daily cross-country snapshot of attention

Pulling this every day? Skip the diff work. What rose today across GitHub, npm, Hugging Face and GDELT, already compared with the previous day, in one JSON call: /rising/today on https://api.lilambd.com (endpoints: https://api.lilambd.com/llms.txt?ref=world-signals; $0.05 USDC per call via x402 on Base, no signup; add ?sample=1 for a free preview). The raw files below stay free.

One folder per day under data/YYYY-MM-DD/, and the same files copied to latest/. Built every morning (JST) by the EmpireOS world model. Nothing is generated by a model; every row is a measurement from a public source.

Use it in production

This repo is the free daily snapshot. If an app, agent, game or dashboard of yours pulls it every day, use the API instead: per-country and per-date filters, JSON responses, and plans with higher daily quotas on RapidAPI: World Signals. Paid plans are what keep the daily collection running. Need a table, country or field that is not here? Write to submissions@lilambd.com.

Pro version (World Signals Pro). The GDELT, GitHub, Hugging Face and npm/PyPI tables, deduplicated, keyed to ISO country codes, with previous value, day-over-day change, 7-observation average, z-score and an anomaly flag on every row, plus a per-country daily summary, country briefs and an anomaly digest. Parquet + CSV. Free 3-day sample; latest day $6, last 30 days $15, full history $79. Pay-per-call API, live now: https://api.lilambd.com/signals-pro (x402, $0.05 per call, USDC on Base, no signup; /signals-pro/tables is free). Search trends, podcast and Steam charts stay free-only here. Get the kit: free 3-day sample (Gumroad ยท Payhip), latest day $6 (Gumroad ยท Payhip), last 30 days $15 (Gumroad ยท Payhip), full history $79 (Gumroad ยท Payhip). Questions: submissions@lilambd.com

For AI agents: 14 more open-data APIs (airports, units, regions, power plants, markets, village and rural markets away from big cities, organisations, conflict, Japan municipalities and more), pay per call with x402 (from $0.005 USDC on Base, $0.05 for the computed daily answer at /rising/today and the village-market ledger, no signup), free sample with sample=1, also as an MCP server: https://api.lilambd.com/llms.txt?ref=world-signals

MCP server for agents: https://api.lilambd.com/mcp (tools are listed in https://api.lilambd.com/llms.txt?ref=world-signals).

Quick start (free):

from datasets import load_dataset
trends = load_dataset("Lilambd/world-signals", "search_trends", split="train")
# every collected day: add _history to the config name
trends_all = load_dataset("Lilambd/world-signals", "search_trends_history", split="train")
file what source
search_trends.csv rising searches, 30 countries, with approximate traffic and the headline that drove them Google Trends daily RSS
podcast_charts.csv top-100 podcasts, 30 countries Apple Podcasts marketing RSS
gdelt_events.csv world events of the last hour counted by country x CAMEO root code, with mean Goldstein scale GDELT 2.0 event export
github_new_repos.csv repositories created in the last 7 days with the most stars GitHub search API
hf_trending.csv trending models and datasets Hugging Face Hub API
steam_top.csv top 100 most played games, in Steam charts rank order (not sorted by peak_players; sort by it yourself if you need that) Steam charts API
package_downloads.csv npm packages by keyword (agent, mcp, llm, scraper, ...) and top PyPI packages with download counts npm registry, hugovk/top-pypi-packages
rule_changes.csv leading indicator: rules and laws moving in the last 7 days (proposed, consultation, adopted, made), with country, admin1, topic (tax, budget, trade, finance, energy, ...), effective and comment-close dates US Federal Register, Japan e-Gov public comments, legislation.gov.uk, EUR-Lex (acts and COM proposals)
public_money.csv leading indicator: public money moving (planned tenders, tenders, awards of $10M+ in the US) with buyer, amount, CPV/NAICS-style category, admin1 (US state, UK NUTS) and topic USAspending, EU TED, UK Find a Tender

Countries: JP, US, GB, DE, FR, ES, IT, BR, MX, IN, ID, KR, TW, VN, TH, PH, TR, SA, EG, NG, ZA, KE, PL, NL, SE, AU, CA, AR, CO, CL.

Same data as an API

https://api.lilambd.com โ€” filters, dates and an OpenAPI spec at https://api.lilambd.com/openapi.json. Free direct access is rate-limited. For daily use in a product, take a plan on RapidAPI.

GDELT deduplication (2026-11-03)

Until 2026-11-03, gdelt_events.csv contained exact duplicate rows (the same country x CAMEO root row repeated) in every day folder except 2026-09-22. On 2026-11-03 every data/*/gdelt_events.csv and latest/gdelt_events.csv was replaced with a deduplicated copy (exact duplicates dropped, nothing else changed). If you downloaded these files before that, re-download them or drop exact duplicates (df.drop_duplicates() in pandas, SELECT DISTINCT in DuckDB).

Gaps

Days with no folder were not collected (2026-10-15, 2026-10-16 and 2026-10-22 to 2026-10-31, while the collector was paused). Nothing is back-filled or estimated.

Partial days: on 2026-11-02 podcast_charts.csv covers only 16 of the 30 countries (1,594 rows; JP, US, ES, BR, MX, ID, KR, TW, TH, SA, NG, PL, CA and CO are missing). Most other days have 25 to 30 countries. On 2026-11-03 search_trends.csv has 590 rows, about twice the usual 300, because the feed returned up to 20 terms per country instead of 10. These are distinct terms, not duplicates (the same thing happened on 2026-10-09 and 2026-10-21).

Why this exists

Agents and analysts keep re-scraping the same public endpoints every day. This dataset does it once, keeps the history, and gives it in one flat shape.

Quick start: datasets, pandas, DuckDB

from datasets import load_dataset
gdelt = load_dataset("Lilambd/world-signals", "gdelt_events", split="train")
steam = load_dataset("Lilambd/world-signals", "steam_top", split="train")
import pandas as pd  # hf:// paths need `pip install huggingface_hub`
trends = pd.read_csv("hf://datasets/Lilambd/world-signals/latest/search_trends.csv")
print(trends["country"].value_counts().head(10))

# gdelt_events.csv was deduplicated on 2026-11-03; drop_duplicates() also covers older downloaded copies
gdelt = pd.read_csv("hf://datasets/Lilambd/world-signals/latest/gdelt_events.csv", dtype={"cameo_root": str}).drop_duplicates()
print(gdelt.groupby("country_fips")["event_count"].sum().sort_values(ascending=False).head(10))

# any past day: data/YYYY-MM-DD/<table>.csv
steam = pd.read_csv("hf://datasets/Lilambd/world-signals/data/2026-11-03/steam_top.csv")
print(steam[["rank", "name", "peak_players"]].head(10))
-- DuckDB 0.10+ reads hf:// paths directly
-- top games by peak concurrent players today
SELECT rank, name, peak_players
FROM 'hf://datasets/Lilambd/world-signals/latest/steam_top.csv'
ORDER BY peak_players DESC LIMIT 10;

-- GDELT events per country, exact duplicate rows removed
SELECT country_fips, sum(event_count) AS events
FROM (SELECT DISTINCT * FROM 'hf://datasets/Lilambd/world-signals/latest/gdelt_events.csv')
GROUP BY country_fips ORDER BY events DESC LIMIT 10;

-- history: new GitHub repos with the most stars, across every collected day
SELECT observed_at, repo, stars, language
FROM read_csv('hf://datasets/Lilambd/world-signals/data/*/github_new_repos.csv', union_by_name = true)
ORDER BY stars DESC LIMIT 20;

A runnable notebook is in notebooks/quickstart.ipynb.

Use cases

  • Track what rises in search, podcasts, games, packages and Hugging Face across 30 countries, day by day.
  • Build features for attention or trend forecasting (data/YYYY-MM-DD/ keeps the history).
  • Compare country-level news intensity from GDELT event counts and mean Goldstein scale.
  • Seed agents or dashboards with a free daily snapshot before moving to the API.

Update cadence

Daily: built every morning (JST) into data/YYYY-MM-DD/ and copied to latest/; days with no folder were not collected (see Gaps).

Citation

@misc{lilambd2026_world_signals,
  author       = {Lilambd},
  title        = {World Signals: a daily cross-country snapshot of attention},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/Lilambd/world-signals}},
  url          = {https://huggingface.co/datasets/Lilambd/world-signals}
}

Please also credit the original sources listed in this card.

Attribution

GDELT data is free for any use with attribution to the GDELT Project. Google Trends, Apple, GitHub, Hugging Face, Steam and npm data are reproduced as published by their public endpoints; check each provider's terms before redistribution. PyPI counts are from hugovk/top-pypi-packages. rule_changes and public_money: US Federal Register and USAspending (US public domain), e-Gov (Japan government terms, CC BY 4.0 compatible), legislation.gov.uk and Find a Tender (Open Government Licence v3.0), EUR-Lex and TED (Commission Decision 2011/833/EU); each row carries its licence and source url. Supplier names appear only for US awards of $10M or more.

Maintained by Lilambd. Questions: submissions@lilambd.com


Missing something? Tell us in one line what you'd gladly pay for, if someone made it โ€” one-line form: https://docs.google.com/forms/d/1tzMQtHsEbMg7JyMk7YGqslHijfNTciByi8wrwjQpUPY/viewform (or email submissions@lilambd.com, subject "Data wish"). If enough people ask, we build it, and you get it free.

Using this? Say hi on the Community tab โ€” one line on what you built helps us decide what to collect next.

Used this? Tell us what you used it for, what was missing, and what you used before: https://docs.google.com/forms/d/1tzMQtHsEbMg7JyMk7YGqslHijfNTciByi8wrwjQpUPY/viewform (one minute, no email asked).

Need a specific file or dataset made? Ask here (no promise, quick form): https://docs.google.com/forms/d/e/1FAIpQLSf6V7YALfwE42SaF1re_cLzAoP96TpRg61jE3Nf7UTWEtWRxA/viewform

Downloads last month
4,480

Spaces using Lilambd/world-signals 2

Collection including Lilambd/world-signals