Files
gh-jezweb-claude-skills-ski…/templates/embeddings-fetch.ts
2025-11-30 08:24:54 +08:00

158 lines
3.9 KiB
TypeScript

/**
* Gemini Embeddings with Fetch (Cloudflare Workers)
*
* Demonstrates embedding generation using fetch API instead of SDK.
* Perfect for Cloudflare Workers and edge environments.
*
* Setup:
* 1. Add GEMINI_API_KEY to wrangler.jsonc secrets
* 2. npx wrangler secret put GEMINI_API_KEY
* 3. Deploy: npx wrangler deploy
*
* Usage:
* GET /?text=your+text+here
*/
interface Env {
GEMINI_API_KEY: string;
}
interface EmbeddingRequest {
content: {
parts: Array<{ text: string }>;
};
taskType?: string;
outputDimensionality?: number;
}
interface EmbeddingResponse {
embedding: {
values: number[];
};
}
export default {
async fetch(request: Request, env: Env): Promise<Response> {
// CORS headers for browser access
const corsHeaders = {
'Access-Control-Allow-Origin': '*',
'Access-Control-Allow-Methods': 'GET, POST, OPTIONS',
'Access-Control-Allow-Headers': 'Content-Type'
};
// Handle CORS preflight
if (request.method === 'OPTIONS') {
return new Response(null, { headers: corsHeaders });
}
try {
// Get text from query param or request body
const url = new URL(request.url);
let text: string;
if (request.method === 'POST') {
const body = await request.json<{ text: string }>();
text = body.text;
} else {
text = url.searchParams.get('text') || 'What is the meaning of life?';
}
console.log(`Generating embedding for: "${text}"`);
// Prepare request
const embeddingRequest: EmbeddingRequest = {
content: {
parts: [{ text }]
},
taskType: 'SEMANTIC_SIMILARITY',
outputDimensionality: 768
};
// Call Gemini API
const response = await fetch(
'https://generativelanguage.googleapis.com/v1beta/models/gemini-embedding-001:embedContent',
{
method: 'POST',
headers: {
'x-goog-api-key': env.GEMINI_API_KEY,
'Content-Type': 'application/json'
},
body: JSON.stringify(embeddingRequest)
}
);
if (!response.ok) {
const error = await response.text();
throw new Error(`Gemini API error: ${response.status} - ${error}`);
}
const data = await response.json<EmbeddingResponse>();
const embedding = data.embedding.values;
// Calculate vector magnitude
const magnitude = Math.sqrt(
embedding.reduce((sum, v) => sum + v * v, 0)
);
// Return formatted response
return new Response(JSON.stringify({
success: true,
text,
embedding: {
dimensions: embedding.length,
magnitude: magnitude.toFixed(4),
firstValues: embedding.slice(0, 10).map(v => parseFloat(v.toFixed(4))),
fullVector: embedding
}
}, null, 2), {
headers: {
'Content-Type': 'application/json',
...corsHeaders
}
});
} catch (error: any) {
console.error('Error:', error.message);
return new Response(JSON.stringify({
success: false,
error: error.message,
hint: error.message.includes('401')
? 'Check GEMINI_API_KEY secret is set'
: error.message.includes('429')
? 'Rate limit exceeded (Free tier: 100 RPM)'
: 'Check error message for details'
}, null, 2), {
status: 500,
headers: {
'Content-Type': 'application/json',
...corsHeaders
}
});
}
}
};
/**
* Example wrangler.jsonc configuration:
*
* {
* "name": "gemini-embeddings-worker",
* "main": "src/index.ts",
* "compatibility_date": "2025-10-25",
* "vars": {
* "ENVIRONMENT": "production"
* }
* }
*
* Set secret:
* npx wrangler secret put GEMINI_API_KEY
*
* Test locally:
* npx wrangler dev
* curl "http://localhost:8787/?text=Hello+world"
*
* Deploy:
* npx wrangler deploy
*/