A chatbot integrated with Facebook messenger, api.ai, and codeforces.com that allows users to accomplish most of their browsing activities on Codeforces faster, easier and using natural language queries.
The chatbot is integrated with Facebook messenger through which users and system administrators can interact with the system. The server component is implemented using NodeJS as a web server hosted on Heroku.
For a natural language query preprocessor, the system uses api.ai.
The sever component simply uses the local file system to store data as JSON files. There are three JSON files in the system, one contains data of all problems on Codeforces, one for data of all contests, and the last one is used to initialise an empty object used to cache information about upcoming or running contests. New files may be added as needed if any new features were introduced that needs other entities to operate. These files are automatically loaded on server startup and can be updated and reloaded in the runtime. The data files currently included in the repository contains example data.
There are three core server routines implemented within the server module, these server routines are responsible for loading data files, other server routines and intent handlers. Other than these three server routines an arbitrary number of server routines can exist as separate modules. All server routines are loaded automatically on server startup and can be reloaded manually any time later by invoking the routine responsible for loading server routines. The system includes two separate server routines, one for updating problems data file, and the other for updating the contests data file. Each server routine can be invoked by a predefined command. For each incoming message, the system checks if it starts with the admin password, and consider the rest of the message as the command, and invokes the corresponding routine.
As with server routines, intent handlers are implemented as separate modules and are loaded in the same manner. There are three main features, thus three intent handlers. In addition, there are three basic intent handlers for the welcome intent, the fallback intent and the help intent. Ideally, each intent handler should map to an intent in api.ai agent, however, there is no direct coupling between them, so it can be controlled by implementing the appropriate logic. The three main features are:
- Get problem: This intent is triggered with any phrase starting this the words “problem name”, ”problem id”, or “problem number“. Following these words the user is supposed to specify the intent parameter, which may be the problem name or the problem id. If the user’s message started with just the word “problem”, api.ai will consider all of the following as the problem name. Get problem intent handler simply searches for the problem with the specified id or name, and replies to the user with a link to it.
- Get random problem: The random problem intent is triggered by any phrase like “get me any problem” or “do you have any problem”. The user may specify the problem difficulty and/or its type, if the value of a parameter in this intent isn’t specified by the user, it will have the value “random”. The intent handler will then proceed to return the user links to three problems matching his search criteria, if any.
- Get upcoming contests: This intent is triggered by any phrase asking for upcoming contests. The intent handler replies to the user with information about upcoming contests.
- Before running the server you should replace
CLIENT_ACCESS_TOKENin server.js with the correct access token of your api.ai agent. This access token can be found in your agent's settings in API Keys section. - You should also replace
FB_PAGE_IDandVERIFY_TOKENin server.js with the correct values. FB_PAGE_ID is the id of the facebook page to which the app is subscribed. VERIFY_TOKEN is the choosen token used to verify your webhook. ADMIN_PASSWORDis used to access adminstrative features, so it can be of any value.