Documentation Index
Fetch the complete documentation index at: /llms.txt
Use this file to discover all available pages before exploring further.
One-shot text generation with metrics and options
RunAnywhere.llm.generate is a suspending function that runs a single completion.
import com.runanywhere.sdk.public.RunAnywhere
import com.runanywhere.sdk.public.api.*
val result = RunAnywhere.llm.generate("Name three colours")
println(result.text)
val result = RunAnywhere.llm.generate(
"Explain quantum computing in simple terms",
LlmOptions(
model = "qwen3-0.6b",
maxOutputTokens = 200,
temperature = 0.7f
)
)
println(result.text)
println("${result.tokensPerSecond} tok/s")
suspend fun generate(prompt: String, options: LlmOptions? = null): GenerationResult
suspend fun generate(messages: List<ChatMessage>, options: LlmOptions? = null): GenerationResult
messages form is covered in Chat.
| Field | Type | Default |
|---|---|---|
model | String? | none; uses the loaded model |
maxOutputTokens | Int | DEFAULT_MAX_OUTPUT_TOKENS |
temperature | Float | DEFAULT_TEMPERATURE |
topP | Float | DEFAULT_TOP_P |
systemPrompt | String? | none |
stopSequences | List<String> | empty |
reasoning | ReasoningOptions? | none |
tools | List<ToolDefinition> | empty |
toolChoice | ToolChoice | Auto |
maxToolCalls | Int | DEFAULT_MAX_TOOL_CALLS (5) |
LlmOptions.DEFAULT_MAX_OUTPUT_TOKENS
LlmOptions.DEFAULT_TEMPERATURE
LlmOptions.DEFAULT_TOP_P
LlmOptions.DEFAULT_MAX_TOOL_CALLS
val result = RunAnywhere.llm.generateStructured(
prompt = "Extract the name and age: Dana is 34.",
schema = schema,
mode = StructuredOutputMode.VALIDATION_ONLY
)
| Mode | Behaviour |
|---|---|
VALIDATION_ONLY | Generate freely, then validate. The default. |
REPAIR | Validate, then retry once with a repair instruction if invalid. |
CONSTRAINED | Throws today. |
CONSTRAINED throws SDKException.unsupportedCapability. It needs engine-level constrained
decoding, which is not wired in yet. Use VALIDATION_ONLY or REPAIR.val result = RunAnywhere.llm.generate(
prompt,
LlmOptions(reasoning = ReasoningOptions(includeInOutput = true))
)
/no_think directives. You toggle the mode and render what
comes back.
SDKException when no model can be loaded or generation fails. See
Error handling.