HTM In The Cloud (HITC)
This application provides a REST API for use with Nupic. Currently, it is meant to be deployed on Heroku.
Get a local instance working for testing purposes by cloning this repository
Once you've tested it out locally, you can deploy to heroku:
- Install the heroku toolbelt
- In the root directory of this project, run
heroku create --stack cedarThis will give the git repo a remote repo for heroku
git push heroku master heroku scale web=1 heroku ps # to see if it's running heroku logs -t
Your API should be ready to go!
All methods return a JSON response, if a method required a GUID (globally unique identifier), and there is no model associated with it, a 404 error will be returned.
While still in development the REST API is as follows:
Creates a model,
Takes a dict of arguments in with the following fields:
guid: an optional name for the model, by default a UUID4 will be generated. The guid must be unique.
predictedField: the field to be predicted
modelParams: the nupic model parameters to be used. If none are provided, default parameters are provided based on those from the hot gym anomaly detection.
A JSON object containing the GUID of the model and parameters used.
Gets a list of all models and their associated information.
A JSON list of JSON objects containing:
- The GUID of the model
- The parameters used for the mdoel
- The predicted field
- The number of input records processed
- The last input record processed
- The temporal field if it exists
If the GUID is not found, a 404 error will be returned
Details of the model (same format as
Deletes the model
404 if no such model exists. The body will be a JSON object with:
- success: true if model was deleted, false otherwise
- guid: the guid of the model to deleted
Run data through a model
Takes a list of objects or a single object. If using a datetime field, then they must be pushed in chronological order. Each input object must have those fields given in the parameters. Timestamp must be an integer or float that is a unix timestamp.
Returns the output of the nupic model:
- anomaly score
Resets a model