Data API

Historical, point-in-time series behind every reading on this site, for backtesting and integration.

Point-in-time. Each row is stored at the moment it was produced (t, UTC) and is not revised. The only exception is original-index points rebuilt after a collection outage from articles stored at the time. Those are flagged backfilled: true, so you can exclude them.

Access

	Limit
Open (no key)	60 requests per hour per IP
With `X-API-Key` header	3,000 requests per hour

Every response allows cross-origin requests. Up to 10,000 rows are returned per request. Page through longer ranges with start and end.

Endpoints

History

GET /api/v1/history/{series}?start=2026-10-01T00:00:00Z&end=2026-10-13T00:00:00Z&format=json

`series`	Rows	Extra parameter
`index`	Market Sentiment Index: `t, index, net, label, bullish, neutral, bearish, bull_prob, methodology, model_mode, backfilled`
`news-tone`	News tone stream: `t, news_net, two_se, items, methodology, model_mode, coverage_degraded`
`coin`	Per-coin index: `t, coin, index, net, label, items, stories`	`coin=bitcoin` (CoinGecko id)
`price`	Stored price snapshots: `t, price`	`coin=bitcoin`
`calls`	Every graded track-record signal: `t, source (index_call / crossover), direction, index, btc_24h, peak, hit, win`

start and end are ISO-8601 times. The default is the last 30 days. format=csv returns a CSV file instead of JSON.

Latest

Endpoint	Returns
`GET /api/v1/sentiment`	current news reading, evidence counts, leading narratives
`GET /api/v1/coins`	current per-coin index for all tracked coins (coins below the coverage gate are marked unavailable)
`GET /api/stats`	track record summary, crossovers, monthly results, index vs BTC chart series

Examples

curl "https://narratives.aribresearch.com/api/v1/history/index?start=2026-10-01T00:00:00Z&format=csv" -o index.csv

import pandas as pd
df = pd.read_csv("https://narratives.aribresearch.com/api/v1/history/coin?coin=ethereum&format=csv", parse_dates=["t"])

Notes

methodology identifies the version that produced each row. Rows from before versioning are labelled pre-1.0.
The market index is relative to its own trailing baseline (50 means normal). See Methodology.
Research data, not investment advice.