Class: SupervisorApiAdapter

Adapter that calls the Databricks AI Gateway Responses API (/ai-gateway/mlflow/v1/responses).

Streams SSE events in the OpenAI Responses API wire format and maps them to the AppKit AgentEvent protocol. Tool execution is handled server-side, so the adapter ignores the agents-plugin tool index.

Authentication is handled via the Databricks SDK credential chain — the same mechanism used by DatabricksAdapter.fromModelServing. The transport is injected via SupervisorApiAdapterCtorOptions.streamBody; the fromSupervisorApi factory wires it through the SDK's apiClient.request({ raw: true }).

Set DEBUG=appkit:agents:supervisor-api to log the outbound request shape (model, instructions length, input shape, tool count) and to be notified when the recovery path engages (no incremental deltas, text pulled from response.completed.output[]). The no-delta warning includes a per-turn event-type histogram and the SA-reported status/error/ incomplete_details, so it's already actionable without DEBUG.

Tools are not configured on the adapter. Declare them via createAgent({ tools: () => ({ key: supervisorTools.genieSpace({...}) }) }) (or markdown frontmatter referencing an ambient supervisorTools.* entry); the agents plugin / standalone runAgent aggregates hosted-supervisor entries and routes them to the adapter via AgentInput.extensions[SUPERVISOR_EXTENSION_KEY]. Advanced callers invoking adapter.run(...) directly populate that key themselves.

Example

import { createApp, createAgent } from "@databricks/appkit";
import {
  agents,
  DatabricksAdapter,
  supervisorTools,
} from "@databricks/appkit/beta";

await createApp({
  plugins: [
    agents({
      agents: {
        assistant: createAgent({
          instructions: "You are a helpful assistant.",
          model: DatabricksAdapter.fromSupervisorApi({
            model: "databricks-claude-sonnet-4",
          }),
          tools: () => ({
            nyc: supervisorTools.genieSpace({
              id: "01ABCDEF12345678",
              description: "NYC taxi trip records and zones",
            }),
          }),
        }),
      },
    }),
  ],
});

Implements

• AgentAdapter

Constructors

Constructor

new SupervisorApiAdapter(options: SupervisorApiAdapterCtorOptions): SupervisorApiAdapter;

Parameters

Returns

SupervisorApiAdapter

Properties

acceptsExtensions

readonly acceptsExtensions: readonly ["databricks.supervisor"];

Capability negotiation: the adapter reads its hosted-tool payload from AgentInput.extensions under SUPERVISOR_EXTENSION_KEY. The agents plugin uses this list to warn at registration when the tool index produces extensions the adapter wouldn't consume.

Implementation of

AgentAdapter.acceptsExtensions

consumesInputTools

readonly consumesInputTools: false = false;

Capability negotiation: the adapter does not consume input.tools. Tool execution is owned by the Databricks AI Gateway server-side, so any function tools or local sub-agents declared on this agent would be silently dropped — the agents plugin warns at registration when that combination is detected.

Implementation of

AgentAdapter.consumesInputTools

Methods

run()

run(input: AgentInput, context: AgentRunContext): AsyncGenerator<AgentEvent, void, unknown>;

Parameters

Returns

AsyncGenerator<AgentEvent, void, unknown>

Implementation of

AgentAdapter.run