Version: 6.3.3


This is guide to terms you need to understand to use Cortex Fabric tools.


AccessCortex allows for third party authentication (e.g. Google) for accounts where that access has been configured in Admin Console.
AccountOrganization using Cortex for business solutions. Cortex users, security groups, and catalog elements (agents, skills, actions, connections, secrets, etc) are shared by users within an account (also see Tenant).
ActionsHow a skill carries out the model's intention using the data that comes from Input messages. There are two types of Actions in Cortex: jobs and daemons.
Admin ConsoleThe administrative user-interface for the Cortex system. On Admin Console a Cortex admin can manage users, group permissions, access for single sign-on, regions, actions logs, and catalog elements.
AgentDeployable applications that aggregate the logic, data, and models required to implement an augmented intelligence (AI) application. Agents are composed of skills, distinct units of functionality that provide data inputs and outputs, offered in the form of Services, Managed Content, or Connections (to external databases). Skills are wired together to control the flow of data through an agent. Agents can be triggered by other agents, timers, or external requests.
Agent lifecycleThe process of identifying, designing, building, training, testing, invoking, and using AI agents in the Cortex system.
AlgorithmA specification consisting of rules that uses an input to perform a calculation, process data, or automate reasoning tasks to provide an output.
AuthenticationSecure login to Cortex system environments and application. Cortex supports single sign-on (access to all Cortex systems and sites using a single login experience) and third party authentication (Google).
AuthorizationThe Groups, Regions, Environments assigned to each user that determine their system access.
Authorization ProfilesProfiles created when configuring the CLI that provide the ability to switch "use types" as a command option (--profile).
Building BlocksRefers to the building blocks of agents including: Skills (models and actions (jobs and daemons)), Services, and data flows (mapping).
CAMELCognitive Agent Modeling Execution Language. CognitiveScale's proprietary language for programming Cognitive Agents. The CAMEL specification provides the details of this language and how it is used in Cortex.
CampaignCampaigns are one of the key components of Cortex Fabric. Cortex Fabric Campaigns empower businesses to make data-driven decisions that result in the best possible outcomes. Campaigns provide a framework for defining the goals, KPIs, measurements, Data Sources, and missions to generate executable models for tracking business goals, KPIs through continuous learning and feedback.
ChannelsThe pathways and mapping that promote data flow through an agent from connection and service inputs to Skills (actions) to outputs/synapses to other skills. Also know as data flows or maps.
CLI (Cortex)Cortex's Command Line interface. Cortex component where users can programmatically configure agent components and agents and invoke agents.
CohortThe target of Campaign goals. Cohorts are defined using a query/filter applied to a Profile (e.g. Commercial patients).
Cohort GroupA a subset of the Cohort that you can define using a query/filter (e.g. to group Commercial patients with pre-existing conditions versus Medicare patients within the same cohort).
ConnectionsData sources that serve as inputs and outputs to agents. Cortex has two types of connections: managed content files, database connections.
ConsoleThe user-interface for the Cortex system. Console provides a visual interface for creating, managing, and monitoring Projects, Campaigns, Profiles, Skills, and Agents.
Cortex FabricThe Cognitive Scale augmented intelligence platform. Cortex 5 includes: CLI (command-line interface), CAMEL, Documentation portals, Console, OpenAPI, Operators etc.


Daemons(Skill cortex/daemon runtime type) A skill runtime for web servers typically used for ML predictions and serving inquiries. When started, daemons run indefinitely until stopped.
Data FlowsThe pathways and mapping that promote data flow through an agent from connection and service inputs to Skills (actions) to outputs/synapses to other skills. Also known as channels or maps.
Data Quality IndicatorsSymbols that show the health of the data flows when going through a junction (where data is passed from one component to another) and provide the mapping (data compatibility) interface .
Data SourceThe connection(s) from which the data for a Profile or Agent is ingested further defined by the specified schema attributes.
DebugDisplays the full input and output messages from the test run as a series of tasks as input data is consumed and processed by the skills in the agent.
DeployTo set into a state that is ready for use. In Fabric Skills and Agents are deployed before they can be invoked.
Docker Imagea file, comprised of multiple layers, used to execute code in a Docker container. An image is essentially built from the instructions for a complete and executable version of an application
Docker RegistryA storage and content delivery system that holds named Docker images, available in different tagged versions. Users interact with a registry with Docker push and pull commands. Public - A Docker registry that is accessible by multiple users without authentication; Private - A Docker registry that is accessible only to users with authentication keys
Docker RepositoryA virtual location where you store one or more versions of a specific Docker image.
external-api(Skill cortex/external-api runtime type) Provides access to external APIs, where you specify the API URL, path, headers, and method in the skill definition.
GatewayAgent orchestration or lifecycle events based on an Activation ID (e.g. duration of activation, time of activation).
Goals (Campaign)A goal is a desired result of a Campaign. It reflects how the business objectives for the Campaign will be met using data-driven objective values that can be observed. Success of a goal is tracked by KPIs and measures that are associated with the goal. Each KPI targets a specific cohort or cohort group.
Goal StateThe mission success criteria for an individual member of the Cohort. (e.g. The patient gets a flu shot)
GroupsIn Cortex, groups provide a specified set of permissions for Cortex resources to the users who have been added to them. Groups are defined and assigned in Admin Console.
Input (Panel)The data coming from connections or services coming into an agent. Each input is connected to a skill in the data flow from the Inputs Panel. Skill outputs displayed in the Synapse Panel may also serve as inputs to other skills. Additionally, inputs are a property of skills. The inputs declared in a skill definition represent the structure of the input message that is required to communicate with the skill. This input message can come from a variety of sources, including inputs, service inputs, or outputs from other skills.
InquiryA request to a trained model that elicits an output response.
InstanceRefers to a specific snapshot of an agent deployed in a specific environment with a set of specified parameters. The Instance ID is a unique identifier for an instance of an agent.
InterventionThe specific activities composed of a cost estimate, pre-conditions, effects, and actions that are used to generate the models that are tested via simulations and ultimately deployed to production.
Intervention ActionActions are optional for running simulations. They are Skills, Agents, or scripts used to implement interventions.
Intervention CostThe cost of the action measured on a scale of 0-1 that evaluates factors like effort, financial expenditure, time required to implement, or other factors that must be considered in the decision making process.
Intervention EffectEffects represent the desired end-state for the intervention written as a logical or boolean expressions. For example: 5% of the cohort group makes an appointment after receiving 3 notifications.
Intervention Pre-conditionPre-conditions are the conditions that must be in place before an intervention can be undertaken written as a logical or boolean expression. For example a pre-condition of providing in app alerts is that the software must be developed that have this capability.
Jobs(Skill cortex/job runtime type) Used to process high volumes of data that would normally consume long-term memory if run in the foreground, as well as for running programs that require little to no user interaction and that need to be run on a regular basis. Can be run in groups and batched (run automatically).
JSONCode and protocols used in REST APIs.
KPIKey Performance Indicators defined for Campaign Goals and measured for interventions
Logging (Logs)Agent, service, and action execution events with timestamps, parameters, statuses, and identifiers that can be viewed, filtered, and downloaded from Admin Console or CLI


Managed ContentA type of Connection or data source from a storage service like AWS S3. File formats used by Cortex are csv and json. Database connection drivers are also stored in managed content.
MappingFor data to flow from Data Sources to skill and from skill to skill, the data types must match. Users create these matches by mapping origin data types to data types defined in the open source CAMEL spec. The mapping interface is displayed by the data quality indicators.
MasterIn an infrastructure cluster it is the node that manages the distribution of data and activity in the Workers for elastic search and performance optimization. The IP address of the Master node provides the access to the Rancher portal.
Messages (Input/Output)The input payloads for and output responses provided by services in an agent that are required by a skill. The input message or payload can be configured from the run button in the Developer Tray or from the CLI. Output messages are configured in the skill definition.
MissionsMissions are a key component of a Campaign that are meant to achieve a goal state. Missions use interventions that are optimized for cohorts to achieve the desired goal. Missions are tested and measured by running simulations using synthetic or simulated datasets prior to deployment to production.
ML (Machine Learning)A type of model built into a Skill that includes two action types: training and performing. The model applies what it learns from training to the performance action.
ModelThe concrete, runnable implementation of an ML algorithm that exists within a skill.
Model EventAn arbitrary piece of information related to a Model.
Model ServerA web server (in Docker) that responds to inquiry requests using the trained model.
NamespaceA logical grouping of resources in Kubernetes or a cloud provider account. Each resource name in a namespace is unique within that namespace.
OAuthA tool for enabling single sign-on in a system with multiple tools that may require independent authentication.
Output (Panel)The service(s) coming out of an agent which may be queried by an application to obtain the results set. Output services are listed in the Output Panel. Additionally, each skill has at least one output that is listed in the Synapse Panel. Skill outputs displayed in the Synapse Panel may serve as inputs to other skills. The skill outputs are properties of the skill.
ParametersThe elements of a schema or config file (json or yaml). Each parameter has a name and a data type. Parameter data types must match at data source to skill and skill to skill interfaces, but the names may differ.
PathAn address/location on a local drive, or a network URL where data, messages, or configuration can be retrieved to carry out an action (e.g. to invoke an agent you must supply the path to a JSON definition for the agent).
PayloadA message included in as a Service Input that is required to invoke an agent.
Pipelinea) A hierarchical construction of system environments whereby builds can be promoted from one to the next to provide quality assurance. In Cortex's pipeline, a python class provides a pipeline abstraction used to transform data. Pipeline steps are Python actions that accept a DataFrame as an argument and are expected to transform or enrich the DataFrame for a certain goal.
Plans (Planner)After you have created missions with interventions defined and you run the simulation, plans are displayed and evaluated by Domain experts. A plan is one or more interventions assigned to a group defined from the Cohort. The Planner is the section of the Campaign UI where users interact with Plans.
Plan RefinementThis is a step in the overall simulation process. Running a simulation on a Mission displays recommended plans (groups of interventions) for individuals in the cohort to reach the desired goal state. Upon generation of plans, SMEs then review the plans, provide feedback on them, and re-simulate to generate a newer, better model.
ProfileA collection of data attributes (profile schema) from one or more connections that define a specific entity. Profiles are used to identify cohorts for Campaigns.
PropertiesMetadata and other elements of agents, skills, datasets (connections and managed content), and service inputs/outputs that are used to define, reference, and invoke these entities. Properties of each entity above are defined in CAMEL and displayed in the Properties panel.


RegionA Cortex stack division. By default all Cortex accounts have at least one region.
Reference ConnectionsStreamed read-only input to a skill
Reinforcement LearningAn area of machine learning concerned with how software agents ought to take actions in an environment in order to maximize the notion of cumulative reward. Reinforcement learning is one of three basic machine learning paradigms, alongside supervised learning and unsupervised learning.
RoutesThe pathways and mapping that promote data flow through an agent from connection and service inputs to Skills (actions) to outputs/synapses to other skills etc. Mapping is enabled by CAMEL data type definitions. Also known as data flows and maps.
RuntimeRefers to the events that happen when actions in an agent are triggered or invoked. Includes API calls to and from data connections, the running of jobs and daemons, and the responses generated by those actions.
S3A managed content cloud storage location provided by AWS. S3 is one of the tested and recommended data source connections that can be configured in Cortex.
SchemaA list of data field (column) names and types that is used to organize data in a database or a managed content file.
SecretsA skill's properties can be marked as secure in the skill definition. To set a skill's secure properties, the user selects predefined variables from the Secret dropdown in the skill's Properties panel. A variable has two parts: the key, which is the name of the variable, and a value, which is a secure value that can be defined in the Cortex Vault in Admin Console or in CLI.
ServiceA service's input/output pair that provides the payload/response required to invoke a skill. Additionally, API Services are used to create and run agents via the CLI and the Admin Console backend (i.e. actions service, connections service)
Single Sign-onThe ability to sign into a system of multiple packages and tools with a single authentication action and token. In Cortex JWT tokens and cookies are used to accomplish single sign-on.
SimulationsSimulations are part of a Mission aimed at generating the right plans for the cohort subjects. The activities within a simulation include generation of a model that predicts the right plans, creation of relevant plans for review by a Subject Matter Expert (SME). Note, AI developers cannot edit or delete a Mission while a simulation is running.
SkillsThe computational components of an agent. A skill executes an atomic unit of work and can be triggered by one or more inputs to produce one or more outputs. Skills use Models and Actions to provide core functionalities like:
  • Ingesting, enriching, and storing data from a stream; training and testing an ML algorithm to generate an ML model.
  • Loading data from an external source like a file and transforming the raw data to a Dataset for further processing.
  • Extracting features from datasets; invoking services.
SnapshotSimilar to a label in a version control system, a snapshot points to a specific version of an agent. Explicit snapshots are created by users and have a user defined name, and implicit snapshots are created by the system and are nameless.
Supervised LearningThe process of teaching a model by feeding it input data as well as correct output data. This input/output pair is usually referred to as "labeled data."
Synapse (Panel)The output of a skill when added to Agent composer. From the synapse panel these outputs may be added to the data flows as inputs to other skills.


TaskA single activation of the services defined and queued by a job in Cortex. Each time a job runs a task record is created and displayed in the logs.
Trained ModelA serialized piece of code that results from training via a Model Implementation.
TriggersTriggers are the cron schedules that invoke Skill or Agent jobs.
Types (data)Refers to data types from dataset and mapping schemas. Types that provide the data mapping so in Cortex are defined in the CAMEL spec.
Unsupervised LearningThe use of artificial intelligence (AI) algorithms to identify patterns in data sets containing data points that are neither classified nor labeled.
VaultCortex Vault displays and creates key-value pairs (aka secrets). The keys defined by admins can be selected as secrets for properties within in a skill.
VersionsAn integer that is incremented each time a user modifies a versioned resource. Versioned resources include, for example, agent definitions and skill definitions. Versioning is essential for maintaining the integrity of agent snapshots. The version of each building block in the agent as well as the agent itself provide the definition of the snapshot.
YAMLA human friendly data serialization standard for all programming languages. It is the basis for the CAMEL language used in Cortex.