Text generation is the core LlamaBridge workflow.

Basic one-shot generation#

Use generate(...) when each request is independent.

val ok = LlamaBridge.initGenerateModel(modelPath)
check(ok)

val answer = LlamaBridge.generate("Explain Llamatik in one sentence.")
println(answer)

This is the right choice for:

• one-off prompts
• utility screens
• simple summarization tasks
• commands that do not need chat history

Context-aware generation#

Use generateWithContext(...) when you want to separate:

• system instruction
• external context
• end-user message

val system = "You are a helpful assistant."
val context = "Project: Llamatik is a Kotlin-first llama.cpp integration."
val user = "Write a short README intro."

val text = LlamaBridge.generateWithContext(system, context, user)

This is particularly useful for RAG-like workflows and structured prompting.

Streaming instead of waiting#

If the response may be long, prefer streaming so the UI can render tokens progressively. See the dedicated Streaming guide for details.

Cancelling generation#

LlamaBridge.nativeCancelGenerate()

Call this when:

• the user presses Stop
• the screen is closed
• you no longer need the result

Resource lifecycle#

Do not call shutdown() after every generation. In most apps, you load the model once, reuse it across multiple prompts, and release it only when the feature is torn down or the app is terminating.