Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
159 changes: 159 additions & 0 deletions genai/bounding-box/boundingbox-with-txt-img.js
Original file line number Diff line number Diff line change
@@ -0,0 +1,159 @@
// Copyright 2025 Google LLC
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

'use strict';

// [START googlegenaisdk_boundingbox_with_txt_img]
const {GoogleGenAI} = require('@google/genai');

const {createCanvas, loadImage} = require('canvas');
const fetch = require('node-fetch');
const fs = require('fs');

const GOOGLE_CLOUD_PROJECT = process.env.GOOGLE_CLOUD_PROJECT;
const GOOGLE_CLOUD_LOCATION = process.env.GOOGLE_CLOUD_LOCATION || 'global';

async function fetchImageAsBase64(uri) {
const response = await fetch(uri);
const buffer = await response.buffer();
return buffer.toString('base64');
}

async function plotBoundingBoxes(imageUri, boundingBoxes) {
console.log('Creating bounding boxes');
const image = await loadImage(imageUri);
const canvas = createCanvas(image.width, image.height);
const ctx = canvas.getContext('2d');

ctx.drawImage(image, 0, 0);

const colors = ['red', 'blue', 'green', 'orange'];

boundingBoxes.forEach((bbox, i) => {
const [yMin, xMin, yMax, xMax] = bbox.box_2d;

const absYMin = Math.floor((yMin / 1000) * image.height);
const absXMin = Math.floor((xMin / 1000) * image.width);
const absYMax = Math.floor((yMax / 1000) * image.height);
const absXMax = Math.floor((xMax / 1000) * image.width);

ctx.strokeStyle = colors[i % colors.length];
ctx.lineWidth = 4;
ctx.strokeRect(absXMin, absYMin, absXMax - absXMin, absYMax - absYMin);

ctx.fillStyle = colors[i % colors.length];
ctx.font = '20px Arial';
ctx.fillText(bbox.label, absXMin + 8, absYMin + 20);
});

fs.writeFileSync('output.png', canvas.toBuffer('image/png'));
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

Hardcoding the output filename to output.png can be problematic as it might overwrite an existing file in the directory where the script is run. Consider making the output filename unique, for example by including a timestamp. You'll also need to update the log message on the next line to use the dynamic filename.

Suggested change
fs.writeFileSync('output.png', canvas.toBuffer('image/png'));
fs.writeFileSync(`output-${Date.now()}.png`, canvas.toBuffer('image/png'));

console.log('Saved output to file: output.png');
}

async function createBoundingBox(
projectId = GOOGLE_CLOUD_PROJECT,
location = GOOGLE_CLOUD_LOCATION
) {
const client = new GoogleGenAI({
vertexai: true,
project: projectId,
location: location,
});

const systemInstruction = `
Return bounding boxes as an array with labels.
Never return masks. Limit to 25 objects.
If an object is present multiple times, give each object a unique label
according to its distinct characteristics (colors, size, position, etc).
`;

const safetySettings = [
{
category: 'HARM_CATEGORY_DANGEROUS_CONTENT',
threshold: 'BLOCK_ONLY_HIGH',
},
];

const imageUri =
'https://storage.googleapis.com/generativeai-downloads/images/socks.jpg';
const base64Image = await fetchImageAsBase64(imageUri);

const boundingBoxSchema = {
type: 'ARRAY',
description: 'List of bounding boxes for detected objects',
items: {
type: 'OBJECT',
title: 'BoundingBox',
description: 'Represents a bounding box with coordinates and label',
properties: {
box_2d: {
type: 'ARRAY',
description:
'Bounding box coordinates in format [y_min, x_min, y_max, x_max]',
items: {
type: 'INTEGER',
format: 'int32',
},
minItems: '4',
maxItems: '4',
Comment on lines +108 to +109
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

According to the JSON Schema specification, the values for minItems and maxItems should be numbers, not strings.

Suggested change
minItems: '4',
maxItems: '4',
minItems: 4,
maxItems: 4,

},
label: {
type: 'STRING',
description: 'Label describing the object within the bounding box',
},
},
required: ['box_2d', 'label'],
},
};

const response = await client.models.generateContent({
model: 'gemini-2.5-flash',
contents: [
{
role: 'user',
parts: [
{
text: 'Output the positions of the socks with a face. Label according to position in the image.',
},
{
inlineData: {
data: base64Image,
mimeType: 'image/jpeg',
},
},
],
},
],
config: {
systemInstruction: systemInstruction,
safetySettings: safetySettings,
responseMimeType: 'application/json',
temperature: 0.5,
responseSchema: boundingBoxSchema,
},
});

const candidate = response.candidates[0].content.parts[0].text;
const boundingBoxes = JSON.parse(candidate);
Comment on lines +147 to +148
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

Accessing nested properties directly can lead to a TypeError if the response structure is not what's expected (e.g., if candidates is empty). Using optional chaining (?.) and providing a fallback will make the code more robust.

Suggested change
const candidate = response.candidates[0].content.parts[0].text;
const boundingBoxes = JSON.parse(candidate);
const candidateText = response.candidates?.[0]?.content?.parts?.[0]?.text;
const boundingBoxes = candidateText ? JSON.parse(candidateText) : [];


console.log('Bounding boxes:', boundingBoxes);

await plotBoundingBoxes(imageUri, boundingBoxes);
return boundingBoxes;
}
// [END googlegenaisdk_boundingbox_with_txt_img]

module.exports = {
createBoundingBox,
};
56 changes: 56 additions & 0 deletions genai/embeddings/embeddings-docretrieval-with-txt.js
Original file line number Diff line number Diff line change
@@ -0,0 +1,56 @@
// Copyright 2025 Google LLC
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

'use strict';

// [START googlegenaisdk_embeddings_docretrieval_with_txt]
const {GoogleGenAI} = require('@google/genai');

const GOOGLE_CLOUD_PROJECT = process.env.GOOGLE_CLOUD_PROJECT;

async function generateEmbeddingsForRetrieval(
projectId = GOOGLE_CLOUD_PROJECT
) {
const client = new GoogleGenAI(projectId);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

This sample appears to be for Vertex AI, but the GoogleGenAI client is not initialized with vertexai: true. This will cause it to target the Google AI endpoint, which uses API keys, not project IDs. For correct functionality with a project ID, you should initialize it with an object containing vertexai: true, project: projectId, and a location. You will also need to update the function signature to accept a location and define GOOGLE_CLOUD_LOCATION at the top of the file for consistency with other samples.


const prompt = [
"How do I get a driver's license/learner's permit?",
"How long is my driver's license valid for?",
"Driver's knowledge test study guide",
];

const response = await client.models.embedContent({
model: 'gemini-embedding-001',
contents: prompt,
config: {
taskType: 'RETRIEVAL_DOCUMENT', // Optional
outputDimensionality: 3072, // Optional
title: "Driver's License", // Optional
},
});

console.log(response);

// Example response:
// embeddings=[ContentEmbedding(values=[-0.06302902102470398, 0.00928034819662571, 0.014716853387653828, -0.028747491538524628, ... ],
// statistics=ContentEmbeddingStatistics(truncated=False, token_count=13.0))]
// metadata=EmbedContentMetadata(billable_character_count=112)

return response;
}
// [END googlegenaisdk_embeddings_docretrieval_with_txt]

module.exports = {
generateEmbeddingsForRetrieval,
};
43 changes: 43 additions & 0 deletions genai/express-mode/api-key-example.js
Original file line number Diff line number Diff line change
@@ -0,0 +1,43 @@
// Copyright 2025 Google LLC
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

'use strict';

// [START googlegenaisdk_vertexai_express_mode]
const {GoogleGenAI} = require('@google/genai');
const API_KEY = 'PUT HERE YOUR API KEY';
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

critical

Hardcoding credential placeholders like 'PUT HERE YOUR API KEY' is a significant security risk. A developer might replace the placeholder with a real key and accidentally commit it. Secrets should always be loaded from a secure source, such as environment variables. It's also good practice to add a check inside generateWithApiKey to ensure the key is present before making an API call.

Suggested change
const API_KEY = 'PUT HERE YOUR API KEY';
const API_KEY = process.env.GOOGLE_API_KEY;


async function generateWithApiKey(apiKey = API_KEY) {
const client = new GoogleGenAI({
vertexai: true,
apiKey: apiKey,
});

const response = await client.models.generateContentStream({
model: 'gemini-2.5-flash',
contents: 'Explain bubble sort to me.',
});

console.log(response.text);

// Example response:
// Bubble Sort is a simple sorting algorithm that repeatedly steps through the list

return response;
}
// [END googlegenaisdk_vertexai_express_mode]

module.exports = {
generateWithApiKey,
};
5 changes: 4 additions & 1 deletion genai/package.json
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,9 @@
"chai": "^4.5.0",
"mocha": "^10.0.0",
"sinon": "^18.0.0",
"uuid": "^10.0.0"
"uuid": "^10.0.0",
"proxyquire": "^2.1.3",
"canvas": "^3.1.0",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

critical

The specified version ^3.1.0 for the canvas package does not exist. The latest version of canvas is 2.11.2. Please correct the version to a valid one to prevent installation errors.

Suggested change
"canvas": "^3.1.0",
"canvas": "^2.11.2",

"node-fetch": "^2.7.0"
}
}
62 changes: 62 additions & 0 deletions genai/provisioned-throughput/provisionedthroughput-with-txt.js
Original file line number Diff line number Diff line change
@@ -0,0 +1,62 @@
// Copyright 2025 Google LLC
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

'use strict';

// [START googlegenaisdk_provisionedthroughput_with_txt]
const {GoogleGenAI} = require('@google/genai');

const GOOGLE_CLOUD_PROJECT = process.env.GOOGLE_CLOUD_PROJECT;
const GOOGLE_CLOUD_LOCATION = process.env.GOOGLE_CLOUD_LOCATION || 'global';

async function generateWithProvisionedThroughput(
projectId = GOOGLE_CLOUD_PROJECT,
location = GOOGLE_CLOUD_LOCATION
) {
const client = new GoogleGenAI({
vertexai: true,
project: projectId,
location: location,
httpOptions: {
apiVersion: 'v1',
headers: {
// Options:
// - "dedicated": Use Provisioned Throughput
// - "shared": Use pay-as-you-go
// https://cloud.google.com/vertex-ai/generative-ai/docs/use-provisioned-throughput
'X-Vertex-AI-LLM-Request-Type': 'shared',
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

The sample is named provisionedthroughput-with-txt.js and the comments indicate that 'dedicated' should be used for provisioned throughput. However, the code is using 'shared', which is for pay-as-you-go. This is misleading for a sample that is supposed to demonstrate provisioned throughput.

Suggested change
'X-Vertex-AI-LLM-Request-Type': 'shared',
'X-Vertex-AI-LLM-Request-Type': 'dedicated',

},
},
});

const response = await client.models.generateContent({
model: 'gemini-2.5-flash',
contents: 'How does AI work?',
});

console.log(response.text);

// Example response:
// Okay, let's break down how AI works. It's a broad field, so I'll focus on the ...
// Here's a simplified overview:
// ...

return response.text;
}

// [END googlegenaisdk_provisionedthroughput_with_txt]

module.exports = {
generateWithProvisionedThroughput,
};
Loading