---

title: Workflows description: Choose a source-reading journey or prepare a real text, voice, character, or integration session with explicit acceptance boundaries.

Workflows

The showcase explains AIRI. A full upstream application runs character sessions after configuration. These are different workflows.

The runtime procedures here describe source-supported paths. They are not a record of requests or device checks performed during wiki authoring.

Explore without configuring a provider

1. Open the showcase.
2. Use architecture to choose a surface and its shared modules.
3. Open the source references for the capability that interests you.
4. Read capability-boundaries before choosing a runtime experiment.

The result is an understanding of the system. It is not a generated model response, a microphone session, or an integration connection.

Prepare a text conversation

Prerequisites: a full upstream application from setup, a reachable provider, and a model available through that provider.

1. Open the provider settings at /settings/providers in the full application.
2. Configure the intended provider endpoint and its required credentials.
3. Open /settings/modules/consciousness.
4. Select the provider before selecting its model.
5. Send a short, non-sensitive message through the conversation interface.
6. Inspect the actual response or provider error.

The provider change clears the prior model selection. A model identifier belongs to its provider’s catalog, not to all providers.

Acceptance requires an actual response from the selected provider and model. A configured indicator alone does not prove endpoint access or sufficient account quota.

See providers-and-models for direct endpoints, local servers, and account-backed services.

Add voice one direction at a time

Prerequisites: a usable audio device, browser or OS permission, and compatible transcription or speech providers.

1. Open /settings/modules/hearing.
2. Select a transcription provider and its required model.
3. Enable the microphone through the full application.
4. Inspect the transcript before enabling automatic sending.
5. Open /settings/modules/speech.
6. Select a speech provider, model, and voice as required.
7. Exercise speech output and inspect actual playback.

Hearing automatic sending defaults to disabled. Speech starts with speech-noop. Neither setting guarantees a spoken conversation without further configuration.

flowchart LR
    Mic[Microphone input] --> Hearing[Hearing provider]
    Hearing --> Text[Inspect transcript]
    Text --> Send[Send a text turn]
    Send --> Model[Generation provider]
    Model --> Reply[Inspect response]
    Reply --> Speech[Speech provider]
    Speech --> Playback[Inspect audible output]

Each step has a separate failure boundary. A successful text reply does not prove transcription or playback.

Import a character asset

Prerequisites: a compatible asset, permission to use it, and a working renderer on the selected surface.

1. Open /settings/models in the full application.
2. Import a compatible display-model file.
3. Select the imported model.
4. Inspect the rendered character on the stage.
5. Reload the application and inspect the selected model again.

An imported file entry does not prove rendering. The MMD import path can retain a model without a thumbnail after preview generation fails.

Keep the original asset outside browser storage. Read character-and-audio for format and persistence boundaries.

Connect an external integration

1. Choose one integration from services-and-integrations.
2. Read its source guide and permission requirements.
3. Configure its separate process in a trusted environment.
4. Inspect a real connection and one intended operation.
5. Stop the process after the experiment.

A game or messaging settings page is not acceptance evidence. The external service must connect and perform the intended operation.

Source context

The source snapshot is recorded in provenance.

• apps/stage-web/vite.config.ts, lines 150–159: shared route generation.
• packages/stage-pages/src/pages/settings/providers/index.vue: provider settings route.
• packages/stage-ui/src/stores/modules/consciousness.ts, lines 69–119: model reset and selection requirements.
• packages/stage-ui/src/stores/modules/hearing.ts, lines 332–398: transcription configuration and automatic sending.
• packages/stage-ui/src/stores/modules/speech.ts, lines 80–88: independent speech selection.
• packages/stage-ui/src/stores/display-models.ts, lines 117–166: model import and preview behavior.

Navigation: index · setup · operations