AntScout

Search ant species

Resolve an ant name to the taxa present in the dataset. Matching is by substring, so one wrong letter finds nothing. When that happens the answer carries `did_you_mean`: names within a typo or two, with their fact and paper counts. Read them as candidates, never as synonyms — the corpus holds both "Lasius miger" (a misspelling of Lasius niger) and "Lasius neoniger" (a different animal), and they look alike from here. Matches as a substring against both canonical binomials and the raw dataset k

Get facts for a species

Every recorded fact about a species, grouped by the paper it came from. This is the main retrieval tool. Each source carries title, authors, year and `setting` -- field, lab, review, museum or mixed -- which is enough to cite a fact and weigh how it was produced. Well-studied species have thousands of facts; newest papers come first, and `total_facts` versus `returned_facts` tells you whether the response was truncated. Narrow with `topic` before paging through everything. `topic` never return

Search facts

Search every extracted fact across all species by keyword. Use this for cross-species questions ("which ants farm fungus?", "records of supercolony formation") or when you do not know which species to look at. Wrap a phrase in double quotes to require it verbatim. Each result carries a relevance `score` and results are sorted by it, best first. A query never returns nothing just because one term is missing. Facts are single sentences, and species names are stored as metadata rather than in the

Search papers

Find source papers by title, research focus, methods, year, language or setting. This is a catalogue search over paper METADATA, not over their contents. It answers "what has been published on this" and "where did this claim come from". It cannot answer a question about ants: the words have to appear in a title, focus or methods line, so a query describing biology matches nothing however much the corpus knows. Use `search_paper_text` for what papers say and `search_facts` for extracted claims.

Get a paper

Full record for one paper: metadata plus the species and facts from it. Use after `search_papers` or `search_facts` to read a source in full, for example to check the methods behind a surprising finding. Pass `species` to read one taxon's entry inside a large work — a field guide can cover 79 species, and without the filter they come back alphabetically.

Genus overview

List every species indexed under a genus and how well each is covered. Answers "what species of X are in the literature?" and, with `sort='facts_asc'`, "which are understudied?" — a species with almost no facts is a genuine gap in the literature, not a gap in this dataset.

Species profile

Where a species lives, how big it is and how well it is studied. A name that resolves to nothing comes back with `did_you_mean` rather than an empty profile, because an empty profile reads as "the literature knows nothing about this ant" when it usually means the name was mistyped. Aggregates the structured half of every paper's entry for this taxon: countries with a native / introduced / invasive status, the localities in the papers' own geography, measurements by caste, the split of evidence

Search open questions

Search what the literature leaves open about each species. Every species entry carries the questions its paper does not settle. They are full sentences, so a plain-language query matches them far more readily than it matches the terse facts — which makes this a good way in when you do not yet know the right keywords, and a direct answer to "what is still unknown about X". Each hit names the species and the paper to read.

Search distribution

Which ants the corpus records from a country or a named place. With `country` alone this lists the taxa recorded there, most-published first, each with the status the papers give it (native, introduced, invasive). With `place` it searches the locality text itself, which keeps each paper's own geography — "British Honduras", "Ceylon", "foot of the Sierra Nevada" — so a historical name finds the records that used it while the modern country code still finds them under `country`. Country codes ar

Search paper full text

Search the raw text of the papers, not the extracted summaries. The fact tools return what the extraction kept — a few dozen distilled claims per paper. This searches what the papers actually say: measurement tables, methods detail, discussion, caveats and everything else that never made it into a summary. Use it when `get_species_facts` is thin on a detail, when you need exact numbers, or to verify a fact against its source. Returns passages of about 1,400 characters, with `matched` naming wh

Read a paper's text

Read consecutive passages of one paper's full text. `search_paper_text` reports the hit as `passage_index`; this tool called it `start_passage`, so the obvious copy-paste silently read from passage 0 instead. Both spellings are accepted. Use after `search_paper_text` to read around a hit — a match at passage 40 usually needs its neighbours to make sense — or to walk a paper from the start. `next_start_passage` continues where the response stopped. `from_page` jumps straight to a page of the s

Full-text corpus statistics

Size and coverage of the raw-text corpus behind `search_paper_text`. Call this to know whether the full text is available at all, and how much of it links back to the summarised papers.

Get an AntWiki species page

Read AntWiki's curated account of one species. The fact tools return what individual papers claimed, each with its own date and scope. This returns the community's reconciled synthesis instead: identification, distribution, castes, biology and nomenclature, already weighed against each other. Use it for orientation on a species before diving into `get_species_facts`, and to sanity-check a paper-derived claim against current consensus. `Life History Traits` and `Flight Period` are structured re

Search AntWiki

Search AntWiki's species pages across every ant at once. Answers "which ants do X" against curated accounts rather than raw paper text — a cleaner signal than `search_facts` when the question is about general biology rather than a specific study's finding. Returns one hit per section, so you can follow it into `get_antwiki` for the whole page. `matched_with` reports how the query was satisfied: `all_terms` means every word was found, `any_term` means the query was relaxed to get any results at

AntWiki statistics

Size and coverage of the AntWiki index, and which sections exist. Call this to know whether AntWiki is available at all, and to see the real section names before passing `section` to the other AntWiki tools.

Search ant photographs

Find photographs of ants, ranked by quality and ready to cite. ~328k GBIF occurrence photos, each **described** by a vision model that was shown the community-verified identification and asked what it could see. The caption, `keywords`, `caste`, `setting` and `activity` are that description, in the model's own words -- a reading of the photograph, never an identification, and `caste` is the softest of them: a large worker and a soldier are one word apart. The species, country, licence and creat

Image coverage statistics

How many photographs exist, of what, how good, how big, under which licences. Call this before `search_images` when a search comes back thin: it says whether a taxon has photographs at all, prints the caste tags and the commonest keywords (the closest thing to a term list, since the rest of the description is free text), and gives the p10/median/p90 of `quality`, `identifiable` and `megapixels` -- which is what makes a `min_` threshold a decision rather than a guess.

Find AntWeb specimen images

Find AntWeb specimen photographs: pinned museum material, keyed by angle. ~269k photographs of ~64k pinned specimens of the California Academy of Sciences collection (via GBIF), each taken from AntWeb's standard angles: `head`, `profile`, `dorsal` and `label`, plus rare `ventral` and `full`. Unlike `search_images` these are museum specimens -- the identity comes from the pin label, not from what a camera happened to catch, and type specimens are in here: filter `types_only` or `type_status='hol

Dataset statistics

Corpus overview: totals, language mix, coverage by decade, top species. Call this to judge what the dataset can and cannot answer before relying on an absence of results as evidence.