Cache Embeddings to Save Time and Money

The waste

Re-embedding unchanged documents on every run burns money and time. Cache by a hash of the text.

import hashlib, json, os

def cached_embed(text, cache="emb.json"):
    store = json.load(open(cache)) if os.path.exists(cache) else {}
    key = hashlib.sha256(text.encode()).hexdigest()
    if key not in store:
        store[key] = embed(text)
        json.dump(store, open(cache, "w"))
    return store[key]

In production

Use a real store (Redis, your database) keyed by content hash, and re-embed only when the source text changes. Simple, and it pays for itself immediately on large corpora.

Cache Embeddings to Save Time and Money

What You'll Learn

Tutorial Content

The waste

In production

Your Progress

Tags