Buzzwords API

Basics

Requests

GET and HEAD only. Any other method returns 405.
Responses are JSON (application/json), cached for 60 seconds.
No authentication. No cookies are read or set.
Browsers: CORS is open to the Buzzwords family of sites. Server-side clients are not restricted.

Limits and errors

Rate	10 requests/s per IP (burst 20); legacy `/v2/compat` 3/s (burst 10)
`400`	A parameter is missing or malformed
`404`	Unknown path or word
`429`	Too many requests; retry after a short pause
`502`	The data store did not answer; retry later

Windows

Most calls take a window. All times are UTC.

Form	Meaning	Example
`R24`	The rolling last 24 hours, ending at the latest full hour. Recalculated every hour.	`R24`
`D:YYYYMMDD`	One UTC calendar day. Today's day grows until midnight UTC.	`D:20260927`
`H:YYYYMMDDHH`	One UTC hour.	`H:2026092814`

The colon can be sent raw or as %3A. Use /v2/windows to discover which windows exist.

Scores

Field	Meaning
`count`, `buzzword_mentions`	How many times the word was mentioned in the window's headlines
`pct_of_headlines`	Share of the channel's headlines that mention it
`buzzword_score`	Mentions ÷ total words analysed
`buzzword_headlines_positivity_value`	Sentiment of the headlines that mention the word, −10 (very negative) to +10 (very positive)
`channel_headlines_groupid_positivity_value`	Sentiment of all the channel's headlines, same scale
`buzzword_positivity_impact`	The word's sentiment relative to its channel's
`mentions_throttle`	Minimum mentions for a word to rank in that channel (All: 30; most channels: 4; tickers: 1)

Endpoints

GET /v2/top

The ranked buzzwords of one channel in one window.

Parameter	Required	Values
`window`	yes	`R24`, `D:…`, `H:…`
`channel`	no	channel slug from `/v2/channels`; default `all`
`n`	no	1 to 100; default 10

GET /v2/word/{word}

Where and how much a single word ranks: one row per channel and window. Words are case-sensitive, as counted (Trump, AI, Stock).

Parameter	Required	Values
`window`	no	limit to one window; default = most recent first
`n`	no	1 to 100

GET /v2/channels

The channel catalog: slug, name, group, parent, sla_minutes and status. ACTIVE channels are calculated every hour. NO_SOURCE channels have no active feed. RETIRED channels keep their history. Filter on status client-side.

GET /v2/channel/{slug}/sentiment

The mood of a channel's headlines: value, headline count and words analysed.

Parameter	Required	Values
`window`	no	one window; otherwise the latest windows of `kind`
`kind`	no	`D` (default) or `H`
`n`	no	how many windows, newest first

GET /v2/sentiment

Every channel's mood for one window in a single call: one row per channel, the same fields as /v2/channel/{slug}/sentiment. Use it instead of one call per channel.

Parameter	Required	Values
`window`	yes	`R24`, `D:…`, `H:…`

GET /v2/windows

Which windows have been calculated for a channel, newest first.

Parameter	Required	Values
`channel`	no	slug; default `all`
`kind`	no	`D` or `H`
`n`	no	how many

GET /v2/health

Service freshness: whether feeds, calculations and the API are within their time limits (overall_effective = PASS when all are).

Headlines

The headlines behind the numbers come from the Drupal site, not /v2:

GEThttps://buzzwords.technology/api/headlines3?_format=json

Parameter	Values
`channel`	channel name (e.g. `All`, `Business`)
`buzzword`	a whole word in the title (`Stock` does not match "Stocks"); case-insensitive
`timestamp[min]`, `timestamp[max]`	`YYYY-MM-DD`; the max day itself is excluded
`items_per_page`	5, 8, 10, 20, 25, 50, 100 or 1000 (other values return an empty list)
`page`	0-based

Always send a date range: with timestamp[min]/timestamp[max] a buzzword query answers in well under a second. A query that would take longer than 8 seconds is stopped and answers quickly with HTTP 503 and an empty list []: treat that as "try a narrower range", not as "no headlines".

Each item has title, timestamp (epoch seconds), postedby_fid_name (the feed), field_feed_tag (its channels) and feed_link. Titles arrive HTML-encoded; decode before display and escape again when you render.

Indicators

Per-company and per-sector news signals, hourly: how many headlines named the entity, how unusual that is against its own 30-day baseline, and how those headlines read.

News volume and tone, not price. Nothing here is investment advice.

Parameter	Values
`symbol`	comma list: `AAPL MSFT NVDA META AMD AMZN NFLX BABA TSLA F`, sectors such as `SECTOR::tech_megacap`, and `MARKET::news_mood`; default all
`since`	epoch seconds; default the last 24 hours
`kind`	`H` (default) or `D`
`n`	windows per symbol, default 48, max 500
`format`	`csv` for CSV instead of JSON

Field	Meaning
`mentions`	distinct headlines whose title names the entity
`mentions_z`	(mentions − baseline mean) ÷ max(sd, 1); `null` until the baseline has 7 windows (`z_reason` says why)
`sentiment_avg`, `sentiment_delta`	patent sentiment of those headlines, and its change from the previous window
`confidence`	0 to 1: volume × source diversity

Changelog

Date	Change
2026-10-06	Headlines: `buzzword` now matches whole words (was: substring); `items_per_page` also accepts 8 and 20; a query slower than 8 s returns a fast JSON `503 []` instead of hanging.
2026-09-28	`/v2/sentiment` added (every channel for one window); this reference published.

Legacy (2019–2020 clients)

/v2/compat/getbuzzwords?channel=&day=&limit= and /v2/compat/bzchannels?day= return the old {"Items":[…]} shape for apps built against the 2020 API. day accepts R24, latest or YYYY-MM-DD. New clients should use /v2/top and /v2/channel/{slug}/sentiment, which carry the same fields.