add semantic search
This commit is contained in:
39
scripts/generate-embeddings.js
Normal file
39
scripts/generate-embeddings.js
Normal file
@@ -0,0 +1,39 @@
|
||||
// eslint-disable-next-line @typescript-eslint/no-unused-vars
|
||||
import * as tf from '@tensorflow/tfjs-node';
|
||||
import use from '@tensorflow-models/universal-sentence-encoder';
|
||||
import fs from 'fs-extra';
|
||||
import glob from 'glob';
|
||||
import path from 'path';
|
||||
import { marked } from 'marked';
|
||||
|
||||
async function extractTextFromMarkdown(filePath) {
|
||||
const markdown = await fs.readFile(filePath, 'utf8');
|
||||
return marked(markdown).replace(/<[^>]*>/g, ''); // Strip HTML tags generated by marked
|
||||
}
|
||||
|
||||
async function generateEmbeddingsForDirectory(directoryPath) {
|
||||
// Get all markdown files in directory
|
||||
const files = glob.sync(`${directoryPath}/*.md`);
|
||||
|
||||
// Extract texts from markdown files
|
||||
// eslint-disable-next-line @typescript-eslint/no-unused-vars
|
||||
const poems = await Promise.all(files.map(async (file, _index) => ({
|
||||
id: path.basename(file, '.md'), // Use filename as ID
|
||||
text: await extractTextFromMarkdown(file)
|
||||
})));
|
||||
|
||||
// Load the Universal Sentence Encoder model
|
||||
const model = await use.load();
|
||||
const embeddings = await Promise.all(poems.map(poem => model.embed([poem.text])));
|
||||
|
||||
// Map embeddings back to poem objects
|
||||
const poemEmbeddings = poems.map((poem, index) => ({
|
||||
id: poem.id,
|
||||
vector: embeddings[index].arraySync()[0] // Extract the vector
|
||||
}));
|
||||
|
||||
// Save embeddings to JSON file
|
||||
fs.writeJson('embeddings.json', poemEmbeddings);
|
||||
}
|
||||
|
||||
generateEmbeddingsForDirectory('src/posts/poetry'); // Update path accordingly
|
Reference in New Issue
Block a user