Prepare the interaction
The application initializes the framework and provides supported conversation definitions, resources, and configuration.
ARCHITECTURE
Andamento executes configurable conversation flows and coordinates speech and media. Your native application owns business decisions, data, and integrations—including optional Generative AI services.
SYSTEM OVERVIEW
See exactly where responsibility sits: your application makes decisions, configuration defines interaction behavior, and Andamento coordinates conversation execution with speech and media. The diagram shows relationships, not a required call sequence.
Business rules · callbacks · runtime data
Prompts · behavior · transitions
Flow execution · interaction state · transitions · coordination
Audio interruptions and other supported system events may affect the conversation lifecycle. Handling and recovery depend on the event and platform.
Affects runtime and host applicationThe application calls services, validates results, and supplies decisions or data through supported runtime mechanisms.
Integrated by the native applicationConceptual integration view—not an exact SDK class diagram or an assurance of automatic interruption recovery. Refer to the developer documentation for supported platform-specific behavior.
Andamento coordinates the conversation; the application stays responsible for the outcome.
| Component | Primary responsibility |
|---|---|
| Native application | Business decisions, data, validation, UI, and integrations with external services. |
| Conversation definitions | Configured prompts, interaction steps, transitions, and supported responses. |
| SpeechTiles Andamento | Execution of configured flows, turn coordination, interaction events, and application callbacks. |
| Speech and media services | Recognition, synthesis, playback, and recording as provided and configured for the application. |
| Optional generative AI | Interpretation or generation when selected and managed by the host application; not a required Andamento component. |
| Operating system | Audio and lifecycle conditions, including interruptions and routing events that can affect runtime behavior. |
INTEGRATION LIFECYCLE
The application initializes the framework and provides supported conversation definitions, resources, and configuration.
The host registers callbacks and supplies or updates runtime data as required by the interaction.
Andamento coordinates configured execution and surfaces supported events. The application processes results and directs business decisions.
SYSTEM INTERRUPTION HANDLING
Phone calls, audio-session interruptions, application lifecycle changes, and audio-route events can affect a live voice interaction. These are operating-system events—not user barge-in or an application-requested flow transition.
Andamento's runtime and underlying media components provide platform-specific mechanisms for responding to supported events, helping applications coordinate audio operations and conversation execution as operating conditions change. The host application can participate in determining how the conversation responds to supported system interruptions.
Review documented interruption behavior →Determine which interruptions are handled by the framework and which are exposed to the application.
Consider active audio operations and conversation state when the system changes availability.
Recovery behavior varies with the event, platform, and documented runtime conditions.
This section identifies an architectural concern; it does not claim universal automatic recovery or identical behavior across platforms.
GENERATIVE AI + ANDAMENTO
Generative AI and Andamento complement each other. An application might ask an AI service to interpret an open-ended response or generate an explanation, then use that result within a configured conversational interaction.
The host application chooses the AI service, validates its output, and provides appropriate results to the conversation through supported framework mechanisms.
Explore the Andamento product →Illustrative host-managed integration pattern. AI-service connectivity is supplied by the application; no built-in LLM connector is claimed.