Under Development

Attributes

1. Adding the 'q' parameter to the existing REST API, which returns a list of resolved attributes for the query expression.

Variants

1. Added a new parameter 'vt' to specify variants. Users can now retrieve all canonicals for given set of variants. e.g. vt=category:value
2. Added a new parameter 'cn'  to specify canonicals. Users can now retrieve all variants for a given set of canonicals. e.g. cn=category:value 

Build 2440

Resources

1. Addition to the existing REST API that allows users to pause, clear and resume all resources in queue.  Also a GET on /resources will return the top-level ingestion status.
2. Param 'rref' changed to 'rr' 

Spaces

1. Added sorting to /spaces list

Cluster

1. New REST API for stopping, starting and restarting the saffron cluster.  Also a GET on /cluster returns top-level cluster status and s/w version number.
2. New REST API for retrieving and clearing logs

Classifications

1. Allows users to specify inputs and outputs as parameters in the url instead of specifying analogy categories. This gives the user more flexibility since the JSON response now returns the classified attribute with a normalized metric for each matrix.
2. The existing classifications API has been changed to reflect a hetero-associative query, rather than an indirect connections query via an analogy.  As a result 2 tings have changed - (a) the old 'classifications' query using 'analogy' parameters ac (analogy category) and mxa (max analogies) have been relocated to the 'connections' query.  Rather than using the concept of 'analogies', the concept is that the query is performing a connection query by way of the nearest- neighbor.  Consequently parameter 'ac' has been changed to 'nc' (neighbor category) and 'mxa' has been changed to 'mxn' (max nieghbors). (b) the new classifications query now returns the best matrix label as the output.  The 'q' param reflects the rows that are read and the 'ca' param is required now for /classifications which reflect the columns.  For more info on classifications, see the documentation.

Connections

1. 2 new parameters have been added to this api called neighbor category (nc) and max neighbors (mxn). This allows users to make a connections call to a set of inputs using its neighbors.

Attributes

1. Parameters 'c' or 'cp' are no longer required.  If not specified, cp=* is assumed.

Metrics

1. fc (factor connections) is a new parameter in the metric settings.  Factor connections is on by default but can be turned off if the rank ordering should be based on just the sum of the weights.  Normally results rank ordering considers the number of inputs (rows) connected to any particular output (col) before assessing the weights of those connections.  This option gives the caller the ability to omit connection scores and just use weight scores.

JSON Response Updates

1. Normalized metric (nm) was replaced by normalized score (ns).  The metric is always normalized, it is the score in the advanced metric that may or may not be normalized.
2. Resource JSON has been modified to more closely match XML format.
3. Modified (Episodes, Information, Resources) to standard JSON attribute format: "a":{"c":"category","v":"value"}
4. Improved the way scopes are handled on REST calls. Order matters, so multiples are supplied as comma-separated list sc=0,1,2

AQL

1. Attribute properties: signal and priority have been added.  Attribute properties are specified following the term in square brackets, e.g. person:john[p:1] (sets person:john with an elevated priority of 1).
2. Hierarchical AND-GROUPS, allowing for finer grain control of related (important groups).  For example, "((a b) c)" elevates and connection scores of a and b if connected to an output.

ObservationFilters

1. RegexObservationFilter - allows specification of fine grain filtering rules during ingestion

Information

1. New query to return in information metric of an attribute across a user specified collection of categories.