Class OpenCodeCliAdapter

constructor

• new OpenCodeCliAdapter(options?: BaseAdapterOptions): OpenCodeCliAdapter

  Parameters

  • Optionaloptions: BaseAdapterOptions

  Returns OpenCodeCliAdapter

Readonlyname

Readonlytransport

capabilities

• get capabilities(): CliCapabilityProfile

  Gets the capability profile for this CLI.

  Returns CliCapabilityProfile

getModelInfo

• getModelInfo(): CliModelInfo

  Gets OpenCode model information from canonical registry.

  Returns CliModelInfo

initialize

• initialize(): Promise<void>

  Initializes the adapter — probes available models. Warns if Anthropic provider is configured (#1429 — API key boundaries).

  Returns Promise<void>

listModels

• listModels(): Promise<readonly CliModelInfo[]>

  (#2540) Lists models the local OpenCode installation can route to. Wraps the existing probeAvailableModels() (cached for the process lifetime — see cachedModels at the top of this file) and reshapes the result into the CliModelInfo schema. Splits provider/model ids when present.

  Returns Promise<readonly CliModelInfo[]>

execute

• execute(
      task: CliTask,
      options?: CliExecutionOptions,
  ): Promise<Result<CliResponse, CliError>>

  Executes a task with error handling and retries.

  Timeout priority (highest to lowest):

  1. options.timeoutMs - explicit execution option
  2. task.timeoutMs - task-level setting
  3. getTimeoutForTaskAuto() - computed from task complexity and CLI

  Parameters

  • task: CliTask
  • Optionaloptions: CliExecutionOptions

  Returns Promise<Result<CliResponse, CliError>>

healthCheck

• healthCheck(): Promise<HealthStatus>

  Performs a health check.

  Returns Promise<HealthStatus>

getVersion

• getVersion(): Promise<string>

  Gets CLI version.

  Returns Promise<string>

getCapacity

• getCapacity(): Promise<CapacityStatus>

  Gets current capacity status based on tracked usage. Uses usage-based tracking since CLI subprocess execution doesn't expose HTTP rate limit headers.

  Returns Promise<CapacityStatus>

  See

  Issue #456 - Real API rate limit tracking

executeTask

• executeTask(
      task: CliTask,
      options: Required<CliExecutionOptions>,
  ): Promise<Result<CliResponse, CliError>>

  Executes a task via subprocess, with optional transient-error retry. When transientRetry.enabled is true, transient errors (timeout, rate_limit, connection, parse) are retried with exponential backoff (500ms, 1000ms). Parse errors get max 1 retry (#1533); others get 2.

  Parameters

  • task: CliTask
  • options: Required<CliExecutionOptions>

  Returns Promise<Result<CliResponse, CliError>>

dispose

• dispose(): Promise<void>

  Disposes the adapter (no-op for subprocess).

  Returns Promise<void>