When a user asks for help, Alexa should respond with a brief message giving a high level overview of available functionality.

When the application is stateful, it might be an idea to prefetch and cache some responses after Alexa has responded if we expect the user might ask for more details. This would improve percieved performance.

GetSpendingHistory is defined with two slots, but some of its sample utterances only define one slot. GetSpendingHistory needs to be seperated into two seperate Intents, with the appropriate number of slots corresponding in their utterances. They can still direct to a single handler on the back-end to avoid code duplication.

Add stateful features to the application. For example, when asking Monzo how much I've spent today, the application should ask if I want a more detailed break down after she's read the balance.

Currently only the basic Monzo categories are defined within the transaction_category slot. Any synonyms that a user might potentially use in that slot should also be included. We will also have to map these back into the appropriate category in the backend.

Slots can also not contain unpronouncable characters, such as underscores

Required for approval, need many in addition to the few that Amazon provides.

For example:

AMAZON.YesIntent yes
AMAZON.YesIntent yes please
AMAZON.YesIntent sure
AMAZON.YesIntent o. k.

Currently Alexa always responds with the exact same phrase for a given input, we should try and define a large dictionary of phrases with equivilent meaning that Alexa can then pick from at random. This would help Alexa to feel more like a real person, rather than just a robot.

At the moment, Alexa could freeze a card without the actual owner of the card knowing (e.g. if your children are feeling particularly spiteful). We should push a notification/feed item to Monzo to inform the user of the freeze.

When we respond with information about the user's bank balance or spending history, we should also put a card with that information (and a link to the monzo app? If there's documentation on the Monzo URI protocol) into the Alexa app.

Currently we send a request off to the Monzo API for account details every time an interaction takes place with the application, the account API has a standard latency of 125~250ms. This response could be cached for a short time to improve perceived performance when interacting with Alexa.

I'm having terrible difficulty authorising the Alexa skill.

I just keep getting a message saying that the skill can't be authorised and to try again later.

Any help would be appreciated.

Thanks!

Having trouble testing the skill - when I try a few of the sample utterances, I get the above error. Not sure if it's related, but in the lambda console when I try to run a simple test, I get the following error:

{
  "errorMessage": "Cannot find module '/var/task/index'",
  "errorType": "Error",
  "stackTrace": [
    "Module.require (module.js:353:17)",
    "require (internal/module.js:12:17)"
  ]
}

Any ideas what I could be doing wrong?

Edit: I'm guessing it's down to what I've labelled as the handler. I've tried index.handler (as above), Alexa, alexa, alexa-monzo, and monzo.

Target can be retrieved from the /targets?account_id=$account_id endpoint, and the response format at the time of writing is as follows:

{
    "currency": "GBP",
    "interval": {
        "end": "2016-10-31T23:59:59.999Z",
        "start": "2016-09-30T23:00:00Z"
    },
    "interval_type": "month",
    "targets": [
        {
            "amount": 36000,
            "name": "eating_out",
            "status": "OKAY",
            "type": "category"
        },
        {
            "amount": 70000,
            "name": "total",
            "status": "OKAY",
            "type": "total"
        }
    ]
}

Targets API is currently undocumented, so direct user to the app to set targets. In future we could try and parse and set a target for a time period.

Currently each intent has a few variations of sample utterance (23). To meet Amazon guidelines, there need to be around 1520 for each(!).

Small variations do apply however, for example

GetBalance   what my balance is
GetBalance   what my balance is please
GetBalance   what is my balance
GetBalance   what is my balance please
GetBalance   how much do I have
GetBalance   how much do I have today
GetBalance   how much money is on my account
GetBalance   how much money is on my card
...

Not sure if they're currently supported on the UK Alexa, but a flash briefing API skill for Monzo would be cool.

It could give you a summary of todays/yesterdays spending (depending on the time day?)

Currently Alexa sometimes gets confused between the two intents for UnfreezeCard/FreezeCard - so users may trigger one, while actually wanting the other. To minimize the annoyance of this, Alexa should ask if they want to toggle the state, rather than just bailing out.

• This would require setting up state on the Amazon.YesIntent and Amazon.NoIntent, so we know which handler to use immediately after a user has triggered FreezeCard/UnfreezeCard

We could use the LaunchRequest to prefetch some key information about the user, as it likely means they'll be using the application shortly.

Users should be given a high level overview of functionality they can trigger when activating the LaunchIntent for the first time (currently they only get this information in the reprompt).

You should be able to (un)freeze your card from the Alexa. A notification/feed item should be pushed to Monzo when this happens, in case the card was frozen by a nefarious family member

Provide the user with some example commands, and what they do.

The Monzo API only returns the last 100 transactions by default. When we're checking spending history we only get transactions from that default response, which may not be representative of someone's full spending history.

If there are >100 transactions in a time period we should iterate over the API method till we're sure we've got all transactions for the time period.

A fairly simple one: the /profile endpoint returns a response in the format:

{
  "address": {
    "administrative_area": "",
    "country": "GBR",
    "formatted_address": "",
    "locality": "London",
    "postal_code": "WS1",
    "street_address": [
      "221b Baker Street",
    ]
  },
  "date_of_birth": "YYYY-MM-DD",
  "email": "[email protected]",
  "name": "Alexander Example",
  "phone_number": "+44xxxxxx",
  "user_id": "user_xxxxxxxxxx",
  "user_number": 1
}

Alexa could respond simply with the user number.

My trivial and seemingly-innocuous PR never had tests run against it, and they broke (#15).

Not a huge deal, but Travis/Circle/anything running tests against PRs would have prevented it :)

Hey,

I keep getting "Unable to link your skill" errors any idea what I'm doing wrong?

In monzo I've set the redirect urls to "https://layla.amazon.co.uk/api/skill/link/M2ZMQNS2VXCBTQ" and confidential set to true

In the alexa skill I've got

• client ID: "oauthclient_00....."
• the 2 urls are correct
• the Client Authentication Scheme is set to "Credentials in request body"

If I open the email from monzo on my phone alexa app opens and there's no confirmation and the "Link account" button still shows. If I open it on my laptop I get the unable to link your skill error.

Any pointers would be much appreciated,
Cheers
Steve

All seems good when I try my skill in the developers console, but when I try it on an actual device I just get "sorry something went wrong".

Any tips of where to start debugging? Or do I just need to be patient and wait for it to update on all my devices? (If that's a thing?)