API Terminology Mapping

This document maps the conceptual terms used in this documentation to the corresponding API endpoint paths and SDK resource names. Developers implementing against the Amigo API should reference this mapping alongside the Developer Guide.

Core Concepts → API Resources

Platform Concept
API Path Segment
SDK Resource
Notes

Context Graph

service_hierarchical_state_machine

organization.serviceHierarchicalStateMachines

The API uses the internal system name. Context Graphs are versioned.

Actions (custom programs)

tool

tools

"Actions" in platform docs = "Tools" in the API. Includes versioning, env vars, invocation.

Dynamic Behaviors

dynamic_behavior_set

dynamicBehaviorSets

Versioned behavior sets with trigger-based activation.

Agent Core

agent

organization.agents

Agent definitions with persona, background, directives. Versioned.

Evaluations / Metrics

metric + simulation

metrics + simulation endpoints

Metrics for measurement, Simulation for personas/scenarios/unit tests.

Agent Forge

CLI tool (forge)

N/A (uses existing API endpoints)

Configuration management tool, not a separate API.

Memory Architecture → API Endpoints

Memory Layer
Description
API Endpoint
Notes

L0 - Raw Transcripts

Complete conversation records

GET /conversation/{id}/messages/

Full message history with all types

L1 - Extracted Memories

Net-new information extracted from conversations

GET /user/{user_id}/memory

Individual memory entries with content, context, importance

L2 - Episodic User Models

Synthesized understanding (internal)

No direct endpoint

Internal processing layer, not directly accessible via API

L3 - Global User Model

Dimensional framework, contextual conditioning

GET /user/{user_id}/user_model

Structured model with dimensions and insights

State Types → API Schemas

Docs State Type
OpenAPI Schema Name
Notes

Decision State

DecisionState

Routing and escalation decisions

Action State

ActionState

User-facing responses

Reflection State

ReflectionState

Strategic reasoning

Recall State

RecallState

Memory integration

Annotation State

AnnotationState

Tagging and metadata

Side-effect State

ToolCallState

External action execution. "Side-effect" in docs = "ToolCall" in API

Event Types (Streaming)

Conceptual Event
NDJSON Event Type
WebSocket Event

Conversation created

conversation-created

server.conversation-created

Agent message chunk

new-message

server.new-message

Interaction complete

interaction-complete

server.interaction-complete

Dynamic behavior triggered

current-agent-action (subtype: dynamic-behavior-triggered)

server.current-agent-action

Tool execution started

current-agent-action (subtype: tool-call-started)

server.current-agent-action

Processing delay filler

current-agent-action (subtype: action-too-long)

server.current-agent-action

Error

error

N/A (WebSocket uses close frames)

Version Set Presets → Channel Profiles

Docs Channel Profile
Version Set Preset
Constraints

Voice (standard)

voice

Max 5 states, action/annotation/tool-call only

Voice (premium)

voice_premium

Same constraints, premium LLM

Voice (ultra-low latency)

voice_ultra_low

Single-state only, no memory retrieval

Text

text

Unlimited states, all types

Text (premium)

text_premium

Same, premium LLM

Async

async

Unlimited states, dynamic behaviors

Async (premium)

async_premium

Same, premium LLM

Last updated

Was this helpful?