Class VercelAISDK

constructor

• new VercelAISDK(): VercelAISDK

  Returns VercelAISDK

ReadonlypromptsHelperImport

denormalizeUpdates

• denormalizeUpdates(
      updates: NormalizedPromptUpdates,
      _currentValues?: Record<string, PropValue>,
  ): Record<string, ModelPropValue | null>

  Convert NormalizedPromptUpdates (what the UI sends back) into the raw property-name-keyed updates that PromptFileType.updateProperty and friends operate on.

  Parameters

  • updates: NormalizedPromptUpdates

    Updates expressed in the normalized vocabulary.

  • Optional_currentValues: Record<string, PropValue>

  Returns Record<string, ModelPropValue | null>

  A Record keyed by the SDK's actual property names. Values may be null (to remove) or a PropValue.

executeConfig

• executeConfig(config: any, options?: ExecuteConfigOptions): Promise<void>

  Executes a prompt config object.

  Parameters

  • config: any

    The config object returned by the prompt function.

  • options: ExecuteConfigOptions = {}

    Optional execution options (trace id, prompt identity).

  Returns Promise<void>

getModelCatalog

• getModelCatalog(): Promise<
      {
          groups: {
              Anthropic: { customValueTemplates: { function: ModelPropValue } };
              Google: { customValueTemplates: { function: ModelPropValue } };
              OpenAI: { customValueTemplates: { function: ModelPropValue } };
          };
          models: ModelInfo[];
          modelValueTypes: {
              function: { description: string; label: string };
              string: { description: string; label: string };
          };
      },
  >

  Returns model catalog information: the set of known providers and a curated list of popular models for this SDK.

  Returns Promise<
      {
          groups: {
              Anthropic: { customValueTemplates: { function: ModelPropValue } };
              Google: { customValueTemplates: { function: ModelPropValue } };
              OpenAI: { customValueTemplates: { function: ModelPropValue } };
          };
          models: ModelInfo[];
          modelValueTypes: {
              function: { description: string; label: string };
              string: { description: string; label: string };
          };
      },
  >

getModelParameters

• getModelParameters(rootDir: string): PropDefinition[]

  Returns the list of model parameters that can be edited in the playground UI for projects rooted at rootDir. Typically extracted from the SDK's published TypeScript type definitions.

  Parameters

  • rootDir: string

    Absolute path to the project root.

  Returns PropDefinition[]

normalizePrompt

• normalizePrompt(prompt: ParsedPrompt): NormalizedPrompt

  Convert a low-level ParsedPrompt produced by a PromptFileType into a NormalizedPrompt that the UI can consume without knowing the SDK's specific property names or message shape.

  Parameters

  • prompt: ParsedPrompt

    The raw parsed prompt.

  Returns NormalizedPrompt

setupTraceIngestion

• setupTraceIngestion(): Promise<TraceIngestor | undefined>

  Called once, during server startup, before any prompt config is built, to perform whatever setup this SDK's tracing mechanism needs (e.g. registering a native telemetry integration, or standing up an OpenTelemetry pipeline).

  If this SDK's tracing is built on OpenTelemetry, its implementation of this method should call setupGlobalOTelPipeline, which ensures the global pipeline is only set up once, even if other SDK adapters also call it.

  Optional — adapters with no tracing support may omit it.

  Returns Promise<TraceIngestor | undefined>

  An ingestor to which TraceSinks should be added, or undefined if tracing cannot be set up.