Home > Tasks > Runtime > CodingTask

CodingTask

A high-performance execution engine bridging LLM code generation with live runtimes. Orchestrates iterative feedback loops, automatic error correction, and multi-format execution transcripts.

Side-Effect: Variable Auto-Fix Supported Multi-Runtime
⚙️ CodingTask Config

val task = CodingTask(
    dataStorage = storage,
    session = session,
    user = currentUser,
    ui = socketManager,
    codeRuntime = KotlinRuntime(),
    symbols = mapOf("api" to apiClient),
    model = openAIModel,
    mainTask = parentTask,
    autoFix = true
).start(CodeRequest(listOf(
    "Fetch user profile for ID 123" to Role.user
)))
→
👁️ SessionTask UI
Code
val user = api.getUser("123")
println("User: ${user.name}")
▶ Run Click to execute in KotlinRuntime
View Transcript (MD/HTML/PDF)

Workspace Explorer

Browse generated transcripts and execution artifacts in the CodingTask test workspace.

Constructor Parameters

Parameter Type Description
codeRuntime* CodeRuntime The execution environment (e.g., Kotlin, Python, Bash).
symbols* Map<String, Any> Context variables injected into the runtime environment.
model* ChatInterface The LLM used for code generation and error correction.
mainTask* SessionTask The parent task node in the UI tree.
autoFix Boolean If true, executes generated code immediately without user intervention.
retryable Boolean Wraps the process in a Retryable block for UI-driven re-runs.
temperature Double LLM sampling temperature. Defaults to 0.1 for precision.
details String? Additional context or system instructions for the code agent.
describer TypeDescriber Handles reflection-based documentation of the symbols provided to the LLM.

Key Methods

Method Return Description
start(codeRequest, task) Unit Entry point. Initiates the subtask and starts the generation flow.
execute(task, response, request) String Runs the code and returns a formatted string of Result and Output.