I need to write something like circular buffer in TurboProlog 2.0 for calculating average. I don't know what predicates i need to write, and have no idea how link them together.
I'm not sure what functionality of "a circular buffer" needs to be realized for your application. In general a "buffer" would be reusable storage, often associated with I/O processes that handle asynchronous communications (hence the need for a buffer that allows one process to get ahead of the other). A "circular buffer" denotes a way of managing the available storage with pointers (both to the beginning and end of valid/unprocessed data) that wraparound through a (linear) contiguous region. This has an advantage over maintaining a FIFO queue with a fixed location for the beginning of valid data because no "shuffling" of unprocessed items is required.
In the general context of standard Prolog, where rewriting memory locations is not directly supported, that advantage doesn't make sense. Even in Turbo Prolog it has to be asked exactly what you want to accomplish, so that a skillful use of the extended/nonstandard features available can be made.
Here are some ideas:
Turbo Prolog supports lists that are in some ways more restrictive and perhaps in other ways more elaborate than the proper lists of standard Prolog. One of the restrictions is that in Turbo Prolog all items of a list must belong to the same "domain", a notion foreign to the weakly-typed character of standard Prolog. Also domains may be designated as "reference" domains in Turbo Prolog, a level of indirection that permits partially bound compound terms to be passed between subgoals. Without going into too much detail, one sense of "circular buffer" might be a "list" (formed from a reference domain) which wraps back around on itself (a cyclical reference). Such a term can be created in many other Prologs, the distinction being that it is not a proper list. Circular though such a term might be, it would not be much of a buffer (once created) because items in the list could not be rewritten.
Turbo Prolog supports dynamic assertion and retraction of facts, with metapredicates like asserta/1 and assertz/1 that allow serial positioning of new facts at the beginning or the end of those existing facts for the same predicate (and also if desired within a specified named "module" or factbase to use the Turbo Prolog terminology). If the simple management of items in a FIFO queue is your objective, then this is most likely the approach you want (at least for an initial implementation), with items encapsulated as facts.
Turbo Prolog also supports "external" factbases with some additional features, external in the sense of being stored (either in memory or on disk) in a way that allows persistence and expanded space beyond what is allocated for internal factbases. Given the modest fixed sizes usually associated with circular buffers (because they are meant for reuse and overflow is usually dealt with by blocking on the input process until the output process has a chance to catch up), external factbases don't seem to offer much of interest, though possibly the ability to persist "buffers" might be of interest for long-running processes.
Hopefully these suggestions will elicit some clarification on what really needs to be accomplished here.