fix: @invertase update Vertex AI reranker app with proper model and setup and docs

js/testapps/vertexai-reranker/README.md

@@ -1,6 +1,6 @@
 # Sample Vertex AI Plugin Reranker with Fake Document Content
 
-This sample app demonstrates the use of the Vertex AI plugin for reranking a set of documents based on a query using fake document content. This guide will walk you through setting up and running the sample.
+This sample app demonstrates the use of the Vertex AI plugin for reranking a set of documents based on a query using fake document content. Follow this guide to set up and run the sample.
 
 ## Prerequisites
 
@@ -9,6 +9,7 @@ Before running this sample, ensure you have the following:
 1. **Node.js** installed.
 2. **PNPM** (Node Package Manager) installed.
 3. A **Vertex AI** project with appropriate permissions for reranking models.
+4. The **Discovery Engine API** enabled on your Google Cloud project. [Enable it here](https://console.cloud.google.com/apis/api/discoveryengine.googleapis.com/metrics).
 
 ## Getting Started
 
@@ -26,30 +27,48 @@ Create a \`.env\` file in the root directory and set the following variables. Yo
 
 \`\`\`plaintext
 PROJECT_ID=your_project_id_here
-LOCATION=your_location_here
+LOCATION=global
 \`\`\`
 
 These variables are required to configure the Vertex AI project and location for reranking.
 
-### Step 3: Run the Sample
+### Step 3: Build the Project
 
-Start the Genkit server:
+Run the following command to build the project:
 
 \`\`\`bash
-genkit start
+pnpm build
 \`\`\`
 
-This will launch the server that hosts the reranking flow.
+### Step 4: Run the Sample
+
+To start the development environment and test the flow, run:
+
+1. In the first terminal:
+
+   \`\`\`bash
+   pnpm genkit:dev
+   \`\`\`
+
+   This will start the Genkit development server.
+
+2. In another terminal, use the UI or invoke the flow:
+
+   \`\`\`bash
+   genkit flow:run rerankFlow --input '{"query": "quantum mechanics"}'
+   \`\`\`
+
+   Alternatively, access the Genkit UI to interact with the flow.
 
 ## Sample Explanation
 
 ### Overview
 
-This sample demonstrates how to use the Vertex AI plugin to rerank a predefined list of fake document content based on a query input. It utilizes a semantic reranker model from Vertex AI.
+This sample demonstrates how to use the Vertex AI plugin to rerank a predefined list of fake document content based on a query input. It leverages the \`semantic-ranker-512@latest\` model from Vertex AI.
 
 ### Key Components
 
-- **Fake Document Content**: A hardcoded array of strings representing document content.
+- **Fake Document Content**: A hardcoded array of strings simulating document content.
 - **Rerank Flow**: A flow that reranks the fake documents based on the provided query.
 - **Genkit Configuration**: Configures Genkit with the Vertex AI plugin, setting up the project and reranking model.
 
@@ -58,7 +77,7 @@ This sample demonstrates how to use the Vertex AI plugin to rerank a predefined
 The \`rerankFlow\` function takes a query as input, reranks the predefined document content using the Vertex AI semantic reranker, and returns the documents sorted by relevance score.
 
 \`\`\`typescript
-export const rerankFlow = defineFlow(
+export const rerankFlow = ai.defineFlow(
 {
 name: 'rerankFlow',
 inputSchema: z.object({ query: z.string() }),
@@ -73,9 +92,9 @@ async ({ query }) => {
 const documents = FAKE_DOCUMENT_CONTENT.map((text) =>
 Document.fromText(text)
 );
-const reranker = 'vertexai/reranker';
+const reranker = 'vertexai/semantic-ranker-512@latest';
 
-    const rerankedDocuments = await rerank({
+    const rerankedDocuments = await ai.rerank({
       reranker,
       query: Document.fromText(query),
       documents,
@@ -92,12 +111,34 @@ const reranker = 'vertexai/reranker';
 
 ### Running the Server
 
-The server is started using the \`startFlowsServer\` function, which sets up the Genkit server to handle flow requests.
+The server is started using the \`startFlowServer\` function, which sets up the Genkit server to handle flow requests.
 
 \`\`\`typescript
-startFlowsServer();
+ai.startFlowServer({
+flows: [rerankFlow],
+});
 \`\`\`
 
+### Example Input and Output
+
+- **Input**:
+
+  \`\`\`json
+  {
+  "query": "quantum mechanics"
+  }
+  \`\`\`
+
+- **Output**:
+
+  \`\`\`json
+  [
+  { "text": "quantum mechanics", "score": 0.95 },
+  { "text": "schrodinger's cat", "score": 0.85 },
+  { "text": "e=mc^2", "score": 0.80 }
+  ]
+  \`\`\`
+
 ## License
 
 This project is licensed under the Apache License, Version 2.0. See the [LICENSE](LICENSE) file for details.
@@ -106,4 +147,4 @@ This project is licensed under the Apache License, Version 2.0. See the [LICENSE
 
 This sample provides a basic demonstration of using the Vertex AI plugin with Genkit for reranking documents based on a query. It can be extended and adapted to suit more complex use cases and integrations with other data sources and services.
 
-For more information, please refer to the official [Firebase Genkit documentation](https://firebase.google.com/docs/genkit).
+For more information, refer to the official [Firebase Genkit documentation](https://firebase.google.com/docs/genkit).

js/testapps/vertexai-reranker/package.json

@@ -8,7 +8,9 @@
     "compile": "tsc",
     "build": "pnpm build:clean && pnpm compile",
     "build:clean": "rimraf ./lib",
-    "build:watch": "tsc --watch"
+    "build:watch": "tsc --watch",
+    "genkit:dev": "cross-env GENKIT_ENV=dev pnpm dev",
+    "dev": "tsx --watch src/index.ts"
   },
   "keywords": [],
   "author": "",
@@ -24,6 +26,7 @@
     "google-auth-library": "^9.11.0"
   },
   "devDependencies": {
+    "cross-env": "^7.0.3",
     "rimraf": "^6.0.1",
     "typescript": "^5.5.2"
   }

js/testapps/vertexai-reranker/src/index.ts

@@ -14,15 +14,15 @@
  * limitations under the License.
  */
 
-//  Sample app for using the proposed Vertex AI plugin retriever and indexer with a local file (just as a demo).
+// Sample app for demonstrating the Vertex AI plugin retriever and reranker capabilities using a local file for demo purposes.
 
 import { Document, genkit, z } from 'genkit';
-// important imports for this sample:
+// Import necessary Vertex AI plugins and configuration:
 import { vertexAI } from '@genkit-ai/vertexai';
 import { vertexAIRerankers } from '@genkit-ai/vertexai/rerankers';
 import { LOCATION, PROJECT_ID } from './config';
 
-// Configure Genkit with Vertex AI plugin
+// Initialize Genkit with Vertex AI and Reranker plugins
 const ai = genkit({
   plugins: [
     vertexAI({
@@ -37,13 +37,17 @@ const ai = genkit({
       location: LOCATION,
       rerankOptions: [
         {
-          model: 'vertexai/reranker',
+          model: 'semantic-ranker-512@latest', // Semantic ranker model to be used
         },
       ],
     }),
   ],
 });
 
+/**
+ * Mock data to simulate document retrieval.
+ * Each item represents a document with simplified content.
+ */
 const FAKE_DOCUMENT_CONTENT = [
   'pythagorean theorem',
   'e=mc^2',
@@ -62,36 +66,77 @@ const FAKE_DOCUMENT_CONTENT = [
   'movies',
 ];
 
+/**
+ * @flow rerankFlow
+ * Defines a flow to rerank a set of documents based on a given query.
+ *
+ * @flowDescription
+ * This flow takes a query string, retrieves predefined documents, and ranks them
+ * based on their relevance to the query using Vertex AI's reranker model.
+ *
+ * @inputSchema
+ * The flow expects an object containing:
+ * - query: A string representing the user's search query.
+ *
+ * @outputSchema
+ * Returns an array of objects containing:
+ * - text: The content of the ranked document.
+ * - score: The relevance score assigned by the reranker model.
+ *
+ * @example
+ * Input:
+ * {
+ *   "query": "quantum mechanics"
+ * }
+ *
+ * Output:
+ * [
+ *   { "text": "quantum mechanics", "score": 0.95 },
+ *   { "text": "schrodinger's cat", "score": 0.85 },
+ *   { "text": "e=mc^2", "score": 0.80 }
+ * ]
+ */
 export const rerankFlow = ai.defineFlow(
   {
     name: 'rerankFlow',
-    inputSchema: z.object({ query: z.string() }),
+    inputSchema: z.object({ query: z.string() }), // Input must be an object with a 'query' string
     outputSchema: z.array(
       z.object({
-        text: z.string(),
-        score: z.number(),
+        text: z.string(), // Each result includes a document's text
+        score: z.number(), // And its relevance score
       })
     ),
   },
   async ({ query }) => {
+    console.log('Received query:', query);
+
+    // Convert fake document content into Document objects
     const documents = FAKE_DOCUMENT_CONTENT.map((text) =>
       Document.fromText(text)
     );
-    const reranker = 'vertexai/reranker';
 
+    // Specify the reranker to be used
+    const reranker = 'vertexai/semantic-ranker-512@latest';
+
+    // Call the reranker with the query and documents
     const rerankedDocuments = await ai.rerank({
       reranker,
       query: Document.fromText(query),
       documents,
     });
 
+    // Return the reranked documents with text and score
     return rerankedDocuments.map((doc) => ({
       text: doc.text,
       score: doc.metadata.score,
     }));
   }
 );
 
+/**
+ * Starts the Flow Server for testing and UI interaction.
+ * This allows the `rerankFlow` to be invoked from the Genkit UI or programmatically.
+ */
 ai.startFlowServer({
-  flows: [rerankFlow],
+  flows: [rerankFlow], // Registers the defined flow for the server
 });

js/testapps/vertexai-reranker/tsconfig.json

@@ -2,7 +2,8 @@
   "compileOnSave": true,
   "include": ["src"],
   "compilerOptions": {
-    "module": "commonjs",
+    "module": "Node16",
+    "moduleResolution": "node16",
     "noImplicitReturns": true,
     "outDir": "lib",
     "sourceMap": true,