Skip to main content
​
2026-11-06

​
Source groups

  • Source groups: search ready-made lists of publishers that the NewsCatcher team curates for common use cases, such as top financial news or central banks and regulators. Pass one or more slugs in the new source_groups parameter of Search, and list the available groups with GET /api/source-groups.
  • Fixed: COUNT("Apple", 2, "gt") in the *_entity_name parameters now returns the articles that mention the entity more often than the given count. A malformed COUNT call returns 422. See Search by entity.
  • Fixed: with by_parse_date=true, the Aggregation count endpoint counts articles per interval by parse date, the same date it filters on.
​
2026-10-22

​
Faster responses and fewer failed requests

  • Performance: major improvements of latency and response time across all endpoints.
  • Refunds: a request that fails with 503 because the search backend could not be reached is no longer counted against your quota. Such a request is also retried once automatically before the error is returned.
  • Languages: lang=zh and not_lang=zh are accepted and searched as Chinese (China), cn. See Enumerated parameters.
  • Fixed: an article tagged with both an IPTC tag and one of its child tags no longer makes the whole page come back empty.
​
2026-10-09
  • Search by link: the include_translation_fields parameter adds the English translation fields to each article. See Search by link.
  • Fixed: on plans with embeddings, Search by link returns nlp.qwen_embedding for articles from January 1, 2026, onward. It returned no embedding for these articles before.
​
2026-10-06

​
Clearer free trial limits

  • Free trial: a free trial key that has expired or used all its calls gets 403 with a message that says which limit was reached, instead of 401.
​
2026-10-01

​
Updated predefined sources

  • predefined_sources: the top publishers per country now come from a new, re-ranked list. A request such as predefined_sources=top 100 US can return a different set of publishers than before.
​
2026-09-24

​
Search similar retired

  • Breaking: /api/search_similar, deprecated earlier this year, is retired. Every request returns 410 Gone. Use Search instead.
  • Fixed: not_theme no longer drops articles without NLP data. It now removes only the articles that carry the excluded theme.
​
2026-09-08

​
Fixes

  • Fixed: when the search backend cannot be reached, the API returns 503. It returned 408 or an empty 200 before. 408 now means only that the query took too long. See Errors.
​
2026-09-03

​
Retry-After header

  • Rate limits: 429 responses include a Retry-After header that says how many seconds to wait. Requests that stay within your limit get fewer 429 responses. See Rate limits.
  • Fixed: a server that is too busy answers 503 right away instead of 408 after 30 seconds, and 408 messages now say what timed out.
  • Fixed: a q with an unbalanced quotation mark returns 422. It returned an empty 200 before.
​
2026-08-17

​
Fixes

  • Fixed: requests no longer fail with 408 under load when the query itself is fast.
​
2026-08-04

​
Free trial content preview

  • Free trial: articles include is_content_truncated: true when content is cut to a preview, and the preview ends with .... The highlight field is not returned on the free trial.
​
2026-07-28

​
Free trial plan

  • Free trial: a new plan to try the API. content and its translation are cut to a 300-character preview. All search_in fields are available. See Subscription plans.
​
2026-07-01

​
Sources and source rank combined

  • sources with from_rank and to_rank: you can combine a list of sources, or predefined_sources, with a rank range. You get the articles from the listed sources and from the sources in the rank range. The rank range was ignored when sources was set before.
  • Fixed: ranked_only is no longer ignored when sources is set.
​
2026-06-10

​
Fixes

  • Fixed: boolean parameters such as is_news_domain, is_headline and include_additional_info accept true and false in any letter case. Other values return 422.
​
2026-05-11

​
Fixes

  • Fixed: has_nlp=true returns every article that has NLP data, not only the articles with a summary in their original language.
​
2026-04-23

​
Fixes

  • Fixed: from_ and to_ values with a time zone, such as 2026-04-01T00:00:00+02:00, are converted to UTC.
​
2026-04-21

​
Source logos

  • Sources: the logo field is filled in for many more publishers. See Sources.
​
2026-04-15

​
New embeddings for 2026 articles

  • Embeddings: on plans with embeddings, results for date ranges from January 1, 2026, onward return nlp.qwen_embedding. Earlier date ranges still return nlp.new_embedding. See NLP enrichments.
  • Fixed: a quoted phrase inside NEAR matches the exact phrase. It allowed one extra word in between before.
​
2026-03-13

​
Fixes

  • Fixed: a POST request with invalid JSON returns 422 with a message about the JSON syntax. It returned 500 before.
  • Fixed: unexpected server errors return a JSON body with a message.
​
2026-02-18

​
Clustering for 2026 articles

  • Clustering: for date ranges from January 1, 2026, onward, clustering uses new embeddings, and the default clustering_threshold is 0.7. clustering_variable has no effect on these ranges. See Clustering.
  • New error: clustering_enabled=true with a date range that spans January 1, 2026, returns 422. Split the request into two ranges.
​
2026-01-15

​
Search by link text

  • all_links_text: a new filter that finds articles by the text of the links they contain. Matching articles include all_links_data with each link, its domain and its text.
  • Fixed: all_links and all_domain_links also match recent articles, and return these fields for them.
​
2025-12-26

​
Robots.txt compliance per source

  • Sources: with include_additional_info=true, each source includes robots_compliant, its robots.txt compliance as a percentage such as "100%".
​
2025-12-11

​
More languages

  • Languages: lang and not_lang accept about 120 ISO 639-1 codes, up from about 60. Codes such as am, hy, ka, kk, my and uz no longer return 422. See Enumerated parameters.
​
2025-12-10

​
Fixes

  • Fixed: theme and not_theme are case-insensitive and accept multi-word themes such as Financial Crime.
  • Fixed: exclude_duplicates=true no longer drops articles that were not checked for duplicates yet.
  • Fixed: with several languages in lang, every field in search_in is searched, not only the first one.
​
2025-11-17

​
Fixes

  • Fixed: NEAR with a multi-word phrase, such as NEAR("artificial intelligence", "regulation", 15), no longer returns 422. See Advanced querying.