PGVector integration

[BjornMoren]

It would be nice to be able to send an embedding vector directly to PostgreSQL without having to serialize it into a string first.

// Example with Transformers.js library and PGVector extension
const transformers = await import('@xenova/transformers');
const similarityEmbedder = await transformers.pipeline('feature-extraction', 'Xenova/paraphrase-MiniLM-L6-v2');
const result = await similarityEmbedder(text, { pooling: 'mean', normalize: true });

const embedding = '[' + [...result.data].join(',') + ']';   // Please get rid of this step so we can send the Float32Array directly 

await client.query(
   `INSERT INTO post_embedding (post_id, embedding) 
   VALUES ($1, $2)`,
   [postID, embedding]);