The OpenCog CogServer is a network scheme & python command-line server for the OpenCog AtomSpace (hyper-)graph database. It allows network users to run arbitrary python and scheme code on the server, and access the AtomSpace over the network. The cogserver also provides a pseudo-JSON interface, as well as a high-speed data transfer format that is used for building network-distributed AtomSpaces. The CogServer is a part of the OpenCog project.
The CogServer provides a network command-line console and a WebSocket server. The network console server provides a fast, efficient telnet interface, giving access to Scheme (guile), Python and JSON command-lines. These can be used by multiple users at the same time, all obtaining access to the same AtomSpace. This is also shared by the WebSocket interface, so that all users see the same data, irrespective of the network connection.
This capability is useful in several different ways:
-
General maintenance on long-running OpenCog or AtomSpace processes (e.g. robot control, large batch-job processing or long-running data-mining servers.) This includes running ad-hoc commands, monitoring status, and poking around and performing general maintenance on long-running servers.
-
Network command line. Ordinary Python does not allow multiple users to access it at the same time. With the CogServer, multiple Python users can use it simultaneously. As to scheme/guile, whereas there is an ice-9 REPL server, the CogServer is an order of magnitude faster, lower latency/higher throughput, and infinitely more stable; its free of lockups, hangs and crashes. It's fast.
-
WebSocket API. All interfaces are accessible through websockets. The only difference is that prompts are not sent. For example, the python API is available at
ws:https://localhost:18080/py
. At this time, encryption is not supported, sowss:https://
URL's will not work. See the websocket example for more. -
JSON-style interface. This is useful for creating JavaScript-powered visualizers and user interfaces. Suitable for people who are more comfortable working with JSON. This API is available at
ws:https://localhost:18080/json
. -
Bulk data transfer. The base "s-expression" encoding of Atoms and (Truth)Values is UTF-8 text string format. It's human-readable, easy and efficient. It does not require fiddling with complex binary formats or protocols or the use of protocol libraries or API's. (We're looking at you, HTTP, REST, ZeroMQ, ProtoBuff and friends. You are all very sophisticated, yes, but are hard to use. And sometimes painfully slow.)
-
Network-distributed processing. The StorageNode API provides a uniform data transfer API to local disk, 3rd-party databases and network. The CogServer implements this API, thus allowing multiple AtomSpaces distributed on the network to share data.
-
Proxy Agents. The proxy infrastructure has been replaced by a more general, more flexible, more powerful and configurable proxying system. See the main AtomSpace git repo, in the opencog/persist/proxy directory.
-
The
stats
command provides atop
-like command for viewing who is connected to the Cogserver, and what they are doing. Typehelp stats
for more info.
For more info, please consult the CogServer wiki page.
This is version 3.2.0. The code is stable, it's been used in production settings for a decade. There are no known bugs. There are some planned features; see below.
There are three ways to start the cogserver: from a bash shell prompt (as a stand-alone process), from the guile command line, or from the python command line.
-
From bash, just start the process:
$ build/opencog/cogserver/server/cogserver
-
From guile:
(use-modules (opencog cogserver)) (start-cogserver)
-
From python:
import opencog.cogserver
and then??? start_cogserver() ???
(where's the documentation for this?) (At this time, there is no currently active python maintainer. Help wanted.)
Once started, one can obtain a shell by saying rlwrap telnet localhost 17001
, and then py
, scm
or json
to obtain python, scheme or json
shells. This can be done as many times as desired; all shells share the
same AtomSpace, and the system is fully multi-threaded/thread-safe.
The status of all network connections is displayed by stats
. For more
info, type help py
, help scm
, help json
and help stats
.
The rlwrap
utility simply adds arrow-key support, so that up-arrow
provides a command history, and left-right arrow allows in-place editing.
Note that telnet
does not provide any password protection! It is
fully networked, so you can telnet from other hosts. The default port
number 17001
can be changed; see the documentation.
The socket protocol used is 'trivial'. Thus, besides telnet
, one can
also use netcat
, or access the socket directly, with ordinary socket
connect, open and read/write calls.
WebSocket programmers will find it convenient to use their own favorite tools to access the json API. A very simple example can be found in the WebSocket Example directory.
The CogServer is built exactly the same way that all other OpenCog components are built:
clone https://github.com/opencog/cogserver
cd cogserver
mkdir build
cd build
cmake ..
make -j
For additional information on dependencies and general hand-holding with the build, see the building Opencog wiki.
To build and run the CogServer, you need to install the AtomSpace first.
OpenCog AtomSpace database https://github.com/opencog/atomspace It uses exactly the same build procedure as this package. Be sure to
sudo make install
at the end.
The WebSockets server needs the OpenSSL devel environment to be installed.
OpenSSL On Debian/Ubuntu,
sudo apt install libssl-dev
To build and run the unit tests, just say
make test
from the ./build
directory.
See also these README's:
There are two major open ToDo items for the CogServer. These are:
-
Distributed computing. How to build a distributed computing fabric for AtomSpace data? The read-thru and write-thru proxies provide basic building blocks for distributed computing: they can forward I/O traffic to other servers. See the proxy/README for details. There are no automation tools for configuring these into a large complex network. There is a project for this in the AtomSpace Agents git repo. It is currently abandoned (due to lack of interest).
-
Security. Right now, anyone who has network access can attach to the CogServer, and do anything. There are three different ways to do 'anything':
- The top-level CogServer shell allows arbitrary loadable modules to be loaded.
- The python and scheme shells allow the execution of arbitrary python and scheme code, including system calls. This includes file writes.
- The sexpr and json shells are "controlled", in that they do NOT
allow access to system calls. However, they do allow arbitrary
changes to the made to the AtomSpace, including insertion and
deletion of Atoms. The
GroundedPredicateNode
allows for the arbitrary execution of arbitrary scheme and python code.
It's not clear how to introduce security into this model, other than to only open network connections to trusted, authorized users. Presumably, there are several off-the-shelf solutions for controlling network access, but no one has picked a good one. Obviously, exporting the CogServer socket via SSH is a good way of controlling access to who can use it. But SSH proxying is very low-level, and not admin-friendly. Nor particularly user-friendly, either: the user can't just login into some website and ask for access.