generated: '2026-08-23'
method: derived
source: openapi/koreai-abl-runtime-*-openapi.json + https://docs.kore.ai/agent-platform/concepts
note: >-
  Derived from the path templates and parameters of the three published ABL Runtime specs,
  cross-read against the Agent Platform "Building blocks" concept page and the MCP tool
  catalogue. The specs are thin on schemas — components.schemas holds only ErrorResponse —
  so the relationships below come from path structure (which is real and unambiguous:
  /api/v1/project/{projectSlug}/{env}/agent/{agentSlug}/chat states the containment
  hierarchy) rather than from $ref graphs. Field-level attributes are deliberately NOT
  invented.
identifier_style:
  style: slug-and-id-mixed
  note: >-
    Runtime routes address resources by SLUG (projectSlug, agentSlug, workflowSlug) plus an
    environment segment, while management routes and MCP tools address them by ID
    (projectId, deploymentId, workflowId, sessionId, executionId, callbackId). A consumer
    has to hold both forms. Project IDs appear in examples with a proj_ prefix and callbacks
    with a cb_ prefix; API keys use abl_, ak_, pk_ and sdk_ prefixes (see
    authentication/koreai-authentication.yml).
entities:
- name: Workspace
  description: Top-level tenant container; a user switches between workspaces.
  evidence: platform_workspaces MCP tool (list, switch, inspect); docs.kore.ai/agent-platform/administration/workspace-and-team
  relationships:
  - {type: has_many, target: Project, via: workspace membership}
- name: Project
  description: The unit of work — holds agents, workflows, tools, knowledge, configuration and deployments.
  identifiers: [projectSlug, projectId]
  evidence: 'path segment /api/v1/project/{projectSlug}/ and /api/projects/{projectId}/'
  relationships:
  - {type: has_many, target: Agent, via: projectSlug}
  - {type: has_many, target: Workflow, via: projectSlug}
  - {type: has_many, target: Tool, via: projectId}
  - {type: has_many, target: Version, via: projectId}
  - {type: has_many, target: Deployment, via: projectId}
  - {type: has_many, target: Session, via: projectId}
  - {type: has_many, target: AgentTable, via: projectId}
- name: Environment
  description: A named environment segment inside a project's runtime route (the {env} path parameter).
  evidence: 'path segment /api/v1/project/{projectSlug}/{env}/'
  relationships:
  - {type: belongs_to, target: Project, via: path containment}
- name: Agent
  description: A deployed conversational or task agent, authored in ABL.
  identifiers: [agentSlug]
  evidence: 'runtime.agent.chat, runtime.agent.complete, runtime.agent.stream on /agent/{agentSlug}/'
  relationships:
  - {type: belongs_to, target: Project, via: projectSlug}
  - {type: has_many, target: Session, via: agent execution}
- name: Workflow
  description: A deterministic multi-step flow that can be invoked directly or exposed to an agent as a tool.
  identifiers: [workflowSlug, workflowId]
  evidence: runtime.workflow.invoke, runtime.workflow.execute, runtime.workflow.executeVersion
  relationships:
  - {type: belongs_to, target: Project, via: projectSlug}
  - {type: has_many, target: WorkflowExecution, via: workflowId}
  - {type: has_many, target: WorkflowVersion, via: version}
- name: WorkflowExecution
  description: One asynchronous run of a workflow, pollable for status.
  identifiers: [executionId]
  evidence: 'runtime.workflow.getExecution on /api/v1/workflows/{workflowId}/executions/{executionId}'
  relationships:
  - {type: belongs_to, target: Workflow, via: workflowId}
- name: WorkflowVersion
  description: A pinned, executable version of a workflow.
  identifiers: [version]
  evidence: 'runtime.workflow.executeVersion on /api/v1/workflows/{workflowId}/versions/{version}/execute'
  relationships:
  - {type: belongs_to, target: Workflow, via: workflowId}
- name: Session
  description: A durable conversation/execution session carrying persisted traces.
  identifiers: [sessionId]
  evidence: runtime.legacyChat.createSession; debug_session_history; /api/projects/{projectId}/sessions
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
  - {type: has_many, target: Trace, via: sessionId}
- name: Trace
  description: An execution event inside a session (llm_call, tool_call and others), organised into a span tree.
  evidence: debug_traces, debug_get_span_tree, debug_session_history types filter
  relationships:
  - {type: belongs_to, target: Session, via: sessionId}
- name: Tool
  description: A capability an agent can call — code tool, HTTP tool, SOAP/WSDL-derived tool, MCP tool, or an exposed workflow.
  evidence: platform_tools MCP tool; docs.kore.ai/agent-platform/tools-overview
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
  - {type: belongs_to, target: AuthProfile, via: opaque profile ID}
- name: AuthProfile
  description: Credential holder bound to integrations and MCP servers by opaque reference; secrets never enter tool arguments.
  evidence: platform_auth_profiles MCP tool
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
- name: Integration
  description: A configured connector connection (Confluence, Google Drive, ServiceNow, SFTP, SharePoint, custom HTTP).
  evidence: platform_integrations MCP tool; docs.kore.ai/agent-platform/connectors/
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
  - {type: belongs_to, target: AuthProfile, via: opaque profile ID}
- name: Version
  description: An immutable published version of a project, with qualifications, audit trail and diff.
  evidence: platform_versions MCP tool (list, get, publish, qualifications, audit, diff)
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
  - {type: has_many, target: Deployment, via: promotion}
- name: Deployment
  description: A typed deployment of a version to an environment; supports promote, rollback, restore and retire.
  identifiers: [deploymentId]
  evidence: 'POST /api/projects/{projectId}/deployments/{deploymentId}/retire and /rollback; platform_deployments MCP tool'
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
  - {type: belongs_to, target: Version, via: promotion}
- name: KnowledgeBase
  description: Indexed content an agent retrieves against, populated by connectors.
  evidence: docs.kore.ai/agent-platform/knowledge
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
- name: AgentTable
  description: Structured tabular data definitions and project-scoped rows agents read and write.
  evidence: agent_tables MCP tool
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
- name: EvalSet
  description: A set of eval scenarios, personas and evaluators used to score agent behaviour.
  evidence: platform_eval_sets, platform_eval_scenarios, platform_eval_personas, platform_eval_evaluators
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
  - {type: has_many, target: EvalRun, via: eval set}
- name: EvalRun
  description: One execution of an eval set, with preflight, cases and heatmaps.
  evidence: platform_eval_runs MCP tool
  relationships:
  - {type: belongs_to, target: EvalSet, via: eval set}
- name: Channel
  description: A delivery surface an agent is published to (Web SDK, Teams/Copilot, Slack, Genesys Open Messaging, voice, webhook).
  evidence: platform_sdk_channels MCP tool; docs.kore.ai/agent-platform/channels
  relationships:
  - {type: belongs_to, target: Project, via: projectId}
- name: Callback
  description: A resume handle for a suspended agent awaiting an external human task or tool result.
  identifiers: [callbackId]
  evidence: 'POST /api/v1/callbacks/{callbackId} with x-callback-signature'
  relationships:
  - {type: belongs_to, target: Session, via: suspension}
render: null
render_note: No subway/ diagram exists in this repo yet.