Two Android AAR modules:
ai.octoryn:octoryn-core: governed text generation, cancellable SSE, tool-call normalization, JSON Schema structured output, and evidence headers.ai.octoryn:octoryn-compose: Jetpack Compose chat state over the Core stream.
Install the signed Android AARs from Maven Central:
dependencies {
implementation("ai.octoryn:octoryn-core:0.2.0")
implementation("ai.octoryn:octoryn-compose:0.2.0")
}val client = OctorynClient(apiKey = token)
val result = client.generateText(TextRequest("openai/gpt-4.1-mini", prompt = "Hello"))Core returns tool calls but does not run an agent loop or claim an agent API on the Router.
Compose chat supports edit and regenerate. Supply OctorynChatTransport with
stable event IDs for lastEventId, canResume and cursor-aware resume().
The default OkHttp chat stream deliberately does not advertise cursor resume.