The crawler should return an indexing file. If DocFetch is used, INDEX will be used during crawl (index) time. If DocFetch is not used, INDEX will be used during both crawl and clickthrough.
DocumentFormat Class | Plumtree.Remote.Crawler Namespace