CR feedback

Signed-off-by: Hubert Zub <hubert.zub@databricks.com>

Author: hubertzub-db

integrations/appkit-agent/README.md

@@ -1,23 +1,23 @@
 # @databricks/appkit-agent
 
-Agent plugin for [Databricks AppKit](https://github.com/databricks/appkit). Provides two things:
+Agent plugin for [Databricks AppKit](https://github.com/databricks/appkit). You can define an agent using one of the following approaches:
 
-1. **`AgentInterface`** — a contract for writing custom agent implementations that speak the OpenAI Responses API format (streaming + non-streaming).
-2. **`StandardAgent`** — a ready-to-use LangGraph-based ReAct agent that implements `AgentInterface`, with streaming Responses API support, function tools, and Databricks-hosted tool integration (Genie, Vector Search, MCP servers).
+1. Declaratively define an agent by specifying `model`, `tools,` and `instructions`
+2. Implement a custom agent loop using **`AgentInterface`** — a contract for writing custom agent implementations that speak the OpenAI Responses API format (streaming + non-streaming).
 
 ## Installation
 
 ```bash
 npm install @databricks/appkit-agent
 ```
 
-The LangChain peer dependencies are required when using the built-in ReAct agent (not needed if you provide a custom `agentInstance`):
+The following peer dependencies are required when using the built-in agent (not needed if you provide a custom `agentInstance`):
 
 ```bash
 npm install @databricks/langchainjs @langchain/core @langchain/langgraph
 ```
 
-If you use hosted MCP tools (Genie, Vector Search, custom/external MCP servers):
+If you use hosted tools (Genie, Vector Search, custom/external MCP servers):
 
 ```bash
 npm install @langchain/mcp-adapters
@@ -63,7 +63,7 @@ agent({
   // Tools available to the agent (see Tools section below)
   tools: [myTool, genieTool],
 
-  // Or bring your own AgentInterface implementation (skips LangGraph setup)
+  // Or bring your own AgentInterface implementation
   agentInstance: myCustomAgent,
 });
 ```
@@ -109,7 +109,7 @@ Connect to Databricks-managed services without writing tool handlers:
 ```typescript
 // Genie Space — natural-language queries over your data
 const genie = {
-  type: "genie-space" as const,
+  type: "genie_space" as const,
   genie_space: { id: "01efg..." },
 };
 
@@ -214,7 +214,7 @@ class MyAgent implements AgentInterface {
 agent({ agentInstance: new MyAgent() });
 ```
 
-The `StandardAgent` class (exported from this package) is the built-in implementation that wraps a LangGraph `createReactAgent` and translates its stream events into Responses API format. When you pass `model` instead of `agentInstance`, the plugin uses `StandardAgent` under the hood.
+The `StandardAgent` class (exported from this package) is the built-in implementation used when you pass `model` instead of `agentInstance`. It translates the underlying agent's stream events into Responses API format.
 
 ## API Reference
 
@@ -223,7 +223,7 @@ The `StandardAgent` class (exported from this package) is the built-in implement
 | Export                | Kind           | Description                                              |
 | --------------------- | -------------- | -------------------------------------------------------- |
 | `agent`               | Plugin factory | Main entry point — call with config, pass to `createApp` |
-| `StandardAgent`       | Class          | LangGraph-backed `AgentInterface` implementation         |
+| `StandardAgent`       | Class          | Built-in `AgentInterface` implementation                 |
 | `createInvokeHandler` | Function       | Express handler factory for the `/api/agent` endpoint    |
 | `isFunctionTool`      | Function       | Type guard for `FunctionTool`                            |
 | `isHostedTool`        | Function       | Type guard for `HostedTool`                              |

integrations/appkit-agent/src/agent-plugin/hosted-tools.ts

@@ -11,7 +11,7 @@
 // ---------------------------------------------------------------------------
 
 export interface GenieTool {
-  type: "genie-space";
+  type: "genie_space";
   genie_space: { id: string };
 }
 
@@ -41,7 +41,7 @@ export type HostedTool =
 // ---------------------------------------------------------------------------
 
 const HOSTED_TOOL_TYPES = new Set([
-  "genie-space",
+  "genie_space",
   "vector_search_index",
   "custom_mcp_server",
   "external_mcp_server",
@@ -73,7 +73,7 @@ export async function resolveHostedTools(
 
   return tools.map((tool) => {
     switch (tool.type) {
-      case "genie-space":
+      case "genie_space":
         return DatabricksMCPServer.fromGenieSpace(tool.genie_space.id);
 
       case "vector_search_index": {

integrations/appkit-agent/src/tests/agent-plugin/function-tool.test.ts

@@ -51,7 +51,7 @@ describe("isFunctionTool", () => {
 
   test("returns false for hosted tool object", () => {
     expect(
-      isFunctionTool({ type: "genie-space", genie_space: { id: "123" } }),
+      isFunctionTool({ type: "genie_space", genie_space: { id: "123" } }),
     ).toBe(false);
   });
 

integrations/appkit-agent/src/tests/agent-plugin/hosted-tools.test.ts

@@ -12,7 +12,7 @@ import {
 } from "../../agent-plugin/hosted-tools";
 
 const genieTool: GenieTool = {
-  type: "genie-space",
+  type: "genie_space",
   genie_space: { id: "space-123" },
 };
 
@@ -94,7 +94,7 @@ describe("hosted tool types", () => {
   test("can be mixed in an array with discriminator", () => {
     const tools: HostedTool[] = [genieTool, vectorSearchTool];
     const types = tools.map((t) => t.type);
-    expect(types).toEqual(["genie-space", "vector_search_index"]);
+    expect(types).toEqual(["genie_space", "vector_search_index"]);
   });
 });
 
@@ -136,7 +136,7 @@ describe("resolveHostedTools", () => {
     vi.clearAllMocks();
   });
 
-  test("resolves genie-space tool via fromGenieSpace", async () => {
+  test("resolves genie_space tool via fromGenieSpace", async () => {
     const result = await resolveHostedTools([genieTool]);
 
     expect(mockFromGenieSpace).toHaveBeenCalledWith("space-123");