bvanberg / herzog Goto Github PK

Get presigned url from the API
Push image to S3 using the presigned URL

Create a android application that will allow a user to snap a photo of whatever their heart desires and submit the photo to a REST API which will ultimately push the photo and optional description text onto a Kafka stream.

Add an API endpoint that will return a presigned url for clients to upload data to.

Now that we have the basic API application in place it's time to define the API completely.

1. Define all endpoints needed for all actions known at this point (post photos, get photos, supply meta-data along with photos, etc.)
2. Define all parameters (meta data) needed and how those will be handled by the API.
3. Implement all the missing pieces.
4. Document the API endpoints (README)

Get the API running on an instance in AWS. If we're in the free tier this could run 24x7.

elastic beanstalk should work with java ami.

Support scanning bar codes in the client app. Phone only of course.

When running the web service api locally, you will need to provide your AWS accessKey and secretKey via jvm args. This is only required when running the application locally. In an AWS environment, the application will get it's AWS permissions from the IAM role and these arguments will not be needed.

What's indicated in the README does not work, they've just been added there as placeholders.

Get Kafka running on an AWS instance.

Post JSON payload to the API with the photo URL (S3) along with any metadata.

Add a new end point that will accept a JSON payload containing image URL (S3) along with any metadata. As part of this give some thought to the schema required for the metadata.

Given the 2 phase nature of the client it would be nice to provide a simple client to handle this.

Clients can submit a photo and metadata in a single call. The client would handle getting a presigned url, uploading to S3 and submitting the final metadata.

Post an item to the API.
Framework exists such that pluggable processors are quick and easy. The idea being that we can try out ideas pretty quick.

Stage an instance of Kafka to run on AWS.

Implement matching algorithm based on bar code.

Bar code should give us 100% match which should be simple to implement. This task should more focus on the infrastructure required to make adding new matching algorithms easy.

Currently there's a todo comment in the IdentificationResource to remove the hard-coded extension (.jpg) as the suffix for the filename which the image will be uploaded to S3 with. We need a better way of doing this, whether it's adding another query parameter to the resource endpoint which represents the filename to be used, or something a bit more clever.