swapagarwal / jarvis-on-messenger

:speech_balloon: A community-driven python bot that aims to be as simple as possible to serve humans with their everyday tasks

JARVIS on Messenger

Just A Rather Very Intelligent System, now on Messenger!

Messenger is now used by 1.2 billion people every month. With the launch of Send/Receive API, bots are about to take over.

Usage

JARVIS is at your service here.

Demo (Vimeo)

Why?

I created JARVIS with two goals in mind:

1. It should have a lot of useful features (both fun and commonly used).
2. Anyone can contribute to this project. (As this is module-based, anybody with a decent knowledge of Python can contribute.) One of the prime goals of this project is to lower the entry barrier into the world of open source.

Take a look at the contributing guidelines to see how easy it is to add your own code. I'll be waiting for your pull request! 😉

A massive Thank You to all contributors, and congratulations to people who made their first open-source contribution! 🎉

Modules

Feel free to add to this list by opening an Issue/Pull Request.

Name	Sample Query	Source (w/ Attribution)
anime	death note anime	Kitsu
book	anything you want book	Powered by Goodreads
bye	goodbye	---
coin	flip a coin	---
currency	usd to eur rate	Fixer.io
dice	roll a die	---
dictionary	define comfort	Words API
fact	tell me a fact	JARVIS
hello	Hi, Jarvis!	---
help	What can you do?	---
joke	tell me a joke	JARVIS
lyrics	paradise lyrics	Powered by musiXmatch
movie	iron man 2 movie plot
music	songs by linkin park	Spotify
news	latest news	Powered by NewsAPI
ping	ping google.com	Is it up?
quote	random quote	JARVIS
request	report a bug request a feature	---
thanks	Thank you!	---
time	time in seattle	TimeZoneDB API
url	shorten google.com expand http://goo.gl/7aqe	Google URL Shortener
video	videos of sia	YouTube
weather	weather in london	Info provided by OpenWeatherMap
wiki	wiki html	MediaWiki API
xkcd	show a random xkcd comic	xkcd

More sample queries can be found here.

Structure

├── modules/         # home for various features
├── modules/src/     # code goes here
├── modules/tests/   # tests go here
├── data/            # home for shared data
├── templates/       # for sending structured messages
├── CONTRIBUTING.md  # contributing guidelines
└── jarvis.py        # the main bot

Local Development / Testing

1. Clone this repo.
2. Linux:
   a) Debian (Ubuntu, Linux Mint, etc.): sudo apt-get install python-dev libffi-dev libssl-dev
   b) Arch Linux: sudo pacman -S python2 libffi openssl
   c) Fedora: sudo yum install python-devel libffi-devel openssl-devel
   Windows: These should already be pre-installed in your Python bundle.
   Mac/OS X:
   a) If you install Python using brew, the relevant headers are already installed for you. In other words, you don't need python-devel.
   b) brew install pkg-config libffi
export PKG_CONFIG_PATH=/usr/local/Cellar/libffi/3.0.13/lib/pkgconfig/ # May change with libffi version
   pip install cffi
   c) brew install libtins
3. pip install -r requirements.txt
4. python jarvis.py
5. Visit the following URLs to see results:
   http://localhost:5000/process/?q=<<YOUR_QUERY>> returns the intent of the query.
   http://localhost:5000/search/?q=<<YOUR_QUERY>> returns the search result of the query.

• The "process" endpoint returns what module the system classifies your query e.g. a dictionary query, a song search, etc. Visit the following URLs to understand the output format:
  http://localhost:5000/process/?q=tell%20me%20a%20joke
http://localhost:5000/process/?q=time%20in%20seattle
http://localhost:5000/process/?q=convert%2025%20usd%20to%20eur

You can mock the results for local testing by adding your queries here.

• The "search" endpoint returns the actual bot output, which you get when you interact with the bot using that query.

Note that for the search query to work, you have to set your own key (of the module that you want to test) in config.py

If you want a public endpoint, use the below button to deploy on Heroku and fill the relevant API keys that you would like to use:

TL;DR for Beginners

1. J.A.R.V.I.S. runs on Python 2
2. For the search query to work, you have to set your own key (of the module that you want to test) in config.py
3. The best place to ask anything: https://gitter.im/swapagarwal/JARVIS-on-Messenger
4. Some issues are reserved for you! Low-Hanging Fruit
5. If you're working on something, let everyone know by either creating an issue or commenting on an existing one so that work is not duplicated.
6. Prefer using an IDE (Use PyCharm if you don't have any preference yet)

History

I started out with a rule-based model, but it didn't scale well so now I've shifted to Natural Language Processing. Rest assured, I'll strive to keep it as simple as possible so that you, yes you, can contribute!

If you'd like to contribute to the old model, you are welcome to do so as well. I've created a new branch legacy for this purpose. I'll be accepting Pull Requests to this branch also. 😄

P.S. If you've come this far, you might as well contribute. Looking for a place to start? Take a look at some of the low-hanging fruits!

References

• https://github.com/toddmotto/public-apis

jarvis-on-messenger's Issues

Introduction Message Error

When I enter a query such as Who are you? , like so http://localhost:5000/process/?q=Who are you? the output I get is [null, {}].

It would probably be better if the output would be the content of the help variable under modules/src/help.py

define superhero throws error; other define statements work fine

So I was testing Jarvis for any bugs and have stumbled upon this.

The thing is if you type "define " it works but if "define superhero" is issued, it shows the "Something didn't go well" or some message like that. (It's clearly a bug).

Toss a coin feature

Hey, I think it would be nice to have a simple feature where the users can ask JARVIS to toss a coin and it randomly replies either 'Head' or 'Tail'. What do you think of it?

Add images in coin and dice module

As pointed out in #114,
Instead of displaying text, we can show an image or a gif of a coin showing the result (heads/tails).
This can be used in the dice roll as well.
See the attachment template. Use the Image type.

Jarvis keyword

In any line containing the word Jarvis, the reply seems to be the same as a greeting. Even if I say bye Jarvis!
Screenshot:

Time: Format output

Current: 2016-06-09 13:24:43
New: Thu Jun 09 2016 13:24:43
Line#17

Movie: Add year

Use data['Year']

Support for AIML2 - training files

https://github.com/creatorrr/pyAIML

Add support for various message templates

• Text
• Image
• Button Template
• Generic Template
• Receipt Template

Music Lyrics API : Genius

Hey

I think the Genius API would be great for serving lyrics. https://docs.genius.com/

Best thing about Genius is it's annotations which actually specify the meaning and context of particular lines in a song. Great value add-on for JARVIS

What are your thoughts on this?

Currency: Remove duplicate output

Query: convert one usd to inr
The output 1 USD = 67.1430 INR is repeated two times!

Request module not working

Query: request
Expected: The text in the request.py file
Actual: No response

Integrating JARVIS with dmenu

I think I have a great idea to integrate JARVIS with the pretty popular program launcher dmenu.
It's possible to write a simple script to point to the URL to make the query, but the only issue is the output text from the URL isn't pretty.

So... what do you say?

Anime: Show rating only upto two places of decimal

Line#15
Current: Community Rating: 4.33105360168
Expected: Community Rating: 4.33

Time zone module

Support queries like "time in US"

Weather

Caching results to limit requests to APIs

A lot of API guidelines have one thing in common; they either limit the number of requests made per day, or throttle requests, especially in case of free APIs.

In such a case, eventually, users will be unable to avail these services even if they are repeating a query.

For API's where real-time results are not required (eg: movies, books, jokes, wiki) and where caching is not against the policy of using the API, we could maintain a server-side cache of the results.

For queries with randomized results (like jokes, quotes), we could enter a small probability that the joke is picked up from the cache, rather than a new request (of course, this is just an example.. If we use our own joke/quote list, this won't be required, but the principle applies to most other API's we'll use eventually)

Use cases:

1. I'm trying to decide which book to buy. I ask JARVIS for the rating of Throne of Glass. Then I ask it for Lady Midnight. But being my absent minded self, I forget the rating Throne of Glass had, and so I query it again. But alas, the query limit has been exceeded, and I can't see the rating anymore 😞 Must wait till tomorrow to decide.
2. I have an upcoming math test 😱 I ask JARVIS to tell me about the Pythagoras theorem, but eventually, because of exam stress 📚 I forget it. So I ask it again, but alas, Wikipedia has crashed because it's exam time everywhere, and now I'm going to fail 😭

Rule based to NLP. Are you sure?

You shifted from a rule based system to an online NLP based system. Is it a complete shift? This would mean that the system would not work offline. This would adversely affect in case of slow internet connection.

Add offline data sources

Below is the list of modules that can make use of an offline data source:

• fact
• joke
• quote

Add (Improve) the curated lists in data/* folder. Comment below if you'd like to take up one of these.

Python 3.5 - ImportError: No module named 'src'

After installing the dependencies, and running the python code

$ python -V; printf "\n\n"; python jarvis.py

I get the following error:

Python 3.5.1

Traceback (most recent call last):
File "jarvis.py", line 6, in
import modules
File "/tmp/JARVIS-on-Messenger/modules/init.py", line 2, in
from src import *
ImportError: No module named 'src'
`

Weather: Show temperature in multiple units

Current: 295.31 K
New: 22.16 °C / 71.888 °F
Replace current with new in Line#17

Anime rating

Hey, I was thinking of working on the Anime rating feature. There is this Api from hummingbird.me that seems to be reliable and easy to use.

Check the api structure:

API-v1

Let me know your thoughts

ImportError: No module named 'src'

I was trying to get jarvis running on my machine but i got an import error that it can't find src.

Did I do something wrong?

Wiki: Jarvis

If I type wiki jarvis, I should be getting wiki data from the Jarvis' page. But it doesn't.
PS: There exists a wiki page for Jarvis.
Screenshot:

Send large text in response

The Messenger Send API limits the text to 320 characters! Send them in chunks.

Messages in continuation

The bot should remember what it said the last time, especially if the bot itself is asking a question.

Travel

We could use API's like Skyscanner to find out travel information to a certain country. Maybe include the weather and a few accommodation links, maybe with the TripAdvisor API https://www.tripadvisorsupport.com/hc/en-us/articles/200615457

Could do something like :
Travel to Bali

Response:
Bali prices : Emirates $2000...etc
Weather: ...
Hotel links: [] [] []

Add a curated list of jokes

The API call in the joke module can be prevented by adding a curated list of jokes in the source file itself.
Pros:

• Fewer API calls
• Curated jokes (only include the ones that tickle your funny bone)
• No offensive jokes (sometimes the API returns such type of jokes)

Con:

• The curation effort (only one time!)

Wiki: Add more at wikipedia link

Link to the corresponding wikipedia page at the end of the summary.

Show specific error messages

Right now, whenever an error occurs, a generic error message (something didn't work as expected) is displayed to the user, who can't determine what went wrong!

Restaurant rating

Given the name of a restaurant/cafe/pub, find out locations and a (user) rating.

Nice to have:

1. Timings
2. Average bill for 2
3. Suggest location nearest to my current location
4. Specify cuisine

Ambitious:

1. Suggest 5 nearest restaurants based on my preference (eg: Veg, pizza, burger, italian, etc)

Sample queries:

1. How is the food at Starbucks?
2. Does Starbucks have good food?
3. Is Starbucks a good place to eat?
4. I'm hungry. Where's the nearest Starbucks?
5. I'm craving for some good food. How's Starbucks?
6. Do you know if Starbucks has good coffee?
7. Does Starbucks have cheap food?
8. What is the food rating for Starbucks.
9. I want to eat at Starbucks. Will it be open?

Rolling a dice

Added a pull request for a dice roll: #108

Random fact module needs work

The random fact module needs some work. I have noticed two issues:

1. The are exponential values not being superscripted and so we have values like 6.022e+23 in the output.
2. There are few characters like the single quotation mark which is sometimes shown as \U2019.

This is a medium priority bug and I feel it can be assigned the Low hanging fruit label too.

Calculator feature

Hey, it would be pretty cool if JARVIS could return some basic calculation results . What do you think?

I also found an api for the same 😛
Calcatraz

Making general conversations

I think it'll be good if Jarvis can make conversations with us. Say I'm bored and neither of my close friends are online, so I'll text Jarvis like "Hey", " Hi" or "Hello" and so begins the conversation.

Movie: Add IMDb link

Link to the corresponding IMDb page at the end of the response.

Response for 'bye'

Currently Jarvis responds with-
"I'm sorry; I'm not sure I understand what you're trying to say sir.
Try typing "help" or "request" for a "bye"
It should have a better system to deal with keywords such "bye" ,"thanks you" etc

Personalize the greeting message

In the help module, replace hi there with the first name of the user.

Songs

If I enter a query for which the bot is unable to find a match, it doesn't reply anything (when I entered chrs brown instead of chris brown). I had to wait for some time to realize that 😝
Also, no responses for wiz khalifa? 😱

Xkcd: Show random comic

Right now, it shows the latest.

Add some philosophic answers

It will be nice to add some answers to philosophic.
e.g is there a life after death?
do you believe in god?
do you have a soul?
The idea is taken from the popular virtual assistants like Google allo, and Cortana etc.

Replace links with buttons

• book
• movie
• request
• wiki

Better dictionary API

The current API is not so good; replace it with a better one!

Can JARVIS post images or Gifs?

An example is, in the coin toss feature, instead of displaying text it shows an image or a gif of a coin showing the result (heads/tails). This can be used in the dice roll as well.

HTML entities in output

Query: define a superhero
Response: ... in popular children's and fantasy literature ...
Expected: ... in popular children's and fantasy literature ...

Random fact feature

Hey, There is this interesting api i found, which provides random facts related to numbers and dates. We could use this for now.
Numbersapi

Another one,
FactAPI
(returns data in xml)
Let me know your opinions 😃

Show rating scale

Show the maximum rating, like 3.73 / 5 instead of just 3.73.

• anime
• book
• movie

Add error handling

Below is the list of modules that need a output['error_msg'] in case of an exception:

• anime
• book
• currency
• dictionary Line#20
• movie
• music
• time
• wiki

Time queries not working properly for all locations

As you can see in the screenshot below, for some locations the query in the form time in ____ works perfectly fine, but the query _____ time doesn't.

Pick Random Feature

It would be nice if you could decide group decisions based on what you tell jarvis, who would pick one essentially at random. An example query could be

pick chinese mexican thai

to which Jarvis would reply a random entry

chinese. Have fun.

Not sure if you create the issue or PR first (new to contributing) but that's my idea if someone could help me out with next steps (i.e do I have to have code already done, or is what I have here good enough to get started)