At least, the following filters must be implemented (in this order):
- afs_folder_watch, in charge of watching the folder and sending each new document to its successor filter.
- afs_doc_index, in charge of indexing data.
- afs_search_build, in charge of building reply databases.
- afs_search_defrag, in charge of defragmenting reply databases. Because of the process used for live indexing, a lot of small reply databases can be quickly generated. In order not to decrease reply performance, it is necessary to defragment built reply databases.
- afs_search_deploy, in charge of deploying reply databases to the Update Manager (UM then update all reply servers of the datacenter with the most up-to-date databases).