Cloudflare Workers AI Embeddings

Setup

Install library via pip

%pip install llama-index-embeddings-cloudflare-workersai
# %pip install -e ~/llama_index/llama-index-integrations/embeddings/llama-index-embeddings-cloudflare-workersai

To acess Cloudflare Workers AI, both Cloudflare account ID and API token are required. To get your account ID and API token, please follow the instructions on this document.

# Initilise with account ID and API token

# import os

# my_account_id = "example_id"
# my_api_token = "example_token"
# os.environ["CLOUDFLARE_AUTH_TOKEN"] = "my_api_token"

import getpass

my_account_id = getpass.getpass("Enter your Cloudflare account ID:\n\n")
my_api_token = getpass.getpass("Enter your Cloudflare API token:\n\n")

Text embeddings example

from llama_index.embeddings.cloudflare_workersai import CloudflareEmbedding

my_embed = CloudflareEmbedding(
    account_id=my_account_id,
    auth_token=my_api_token,
    model="@cf/baai/bge-small-en-v1.5",
)

embeddings = my_embed.get_text_embedding("Why sky is blue")

print(len(embeddings))
print(embeddings[:5])

384
[-0.04786296561360359, -0.030788540840148926, -0.07126234471797943, -0.04107927531003952, 0.02904760278761387]

Embed in batches

As for batch size, Cloudflare’s limit is a maximum of 100, as seen on 2024-03-31.

embeddings = my_embed.get_text_embedding_batch(
    ["Why sky is blue", "Why roses are red"]
)
print(len(embeddings))
print(len(embeddings[0]))
print(embeddings[0][:5])
print(embeddings[1][:5])

2
384
[-0.04786296561360359, -0.030788540840148926, -0.07126234471797943, -0.04107927531003952, 0.02904760278761387]
[-0.08951402455568314, -0.015274363569915295, 0.04728245735168457, 0.05478525161743164, 0.05978189781308174]