The getDownloads query returns a files array enumerating every pre-built file with its scope, locus, resolution, format, and a ready-to-use url. Export files are research use only: filenames carry a RESEARCH_USE_ONLY_ prefix and a random suffix and cannot be guessed — pick the row you want and fetch its url (don't build paths by hand). The same list is on the Downloads page.

# R — pull bulk full-field for the current version (research use only)
library(httr); library(jsonlite)
key <- Sys.getenv("SEROTYPE_API_KEY")
q <- 'query { getDownloads { version files { url scope resolution format } } }'
res <- POST("https://serotype.org/api/graphql",
            add_headers(.headers = c("x-api-key" = key)),
            body = list(query = q), encode = "json")
files <- fromJSON(content(res, "text", encoding = "UTF-8"))$data$getDownloads$files
row <- subset(files, scope == "bulk" & resolution == "full_field" & format == "json")
doc <- jsonlite::fromJSON(gzcon(url(row$url)))
df <- doc$data                     # rows; doc$metadata says which version and build
cat(doc$metadata$build_id, "rows:", nrow(df), "\n")

# curl — discover the file URL, then download it
URL=$(curl -s https://serotype.org/api/graphql \
  -H "x-api-key: $SEROTYPE_API_KEY" -H 'content-type: application/json' \
  -d '{"query":"{ getDownloads { files { url scope resolution format } } }"}' \
  | jq -r '.data.getDownloads.files[] | select(.scope=="bulk" and .resolution=="full_field" and .format=="json") | .url')
curl -sL "$URL" | gunzip -c | jq '.metadata.build_id, (.data | length)'

See lesson 3 for a walk-through of all three patterns for bulk data (query narrowing, cap-error handling, manifest-driven discovery).