Bedrock Embeddings

If you’re opening this Notebook on colab, you will probably need to install LlamaIndex 🦙.

%pip install llama-index-embeddings-bedrock

import os

from llama_index.embeddings.bedrock import BedrockEmbedding

embed_model = BedrockEmbedding(
    aws_access_key_id=os.getenv("AWS_ACCESS_KEY_ID"),
    aws_secret_access_key=os.getenv("AWS_SECRET_ACCESS_KEY"),
    aws_session_token=os.getenv("AWS_SESSION_TOKEN"),
    region_name="<aws-region>",
    profile_name="<aws-profile>",
)

embedding = embed_model.get_text_embedding("hello world")

List supported models

To check list of supported models of Amazon Bedrock on LlamaIndex, call BedrockEmbedding.list_supported_models() as follows.

from llama_index.embeddings.bedrock import BedrockEmbedding
import json

supported_models = BedrockEmbedding.list_supported_models()
print(json.dumps(supported_models, indent=2))

Provider: Amazon

Amazon Bedrock Titan embeddings.

from llama_index.embeddings.bedrock import BedrockEmbedding

model = BedrockEmbedding(model_name="amazon.titan-embed-g1-text-02")
embeddings = model.get_text_embedding("hello world")
print(embeddings)

Provider: Cohere

cohere.embed-english-v3

model = BedrockEmbedding(model_name="cohere.embed-english-v3")
coherePayload = ["This is a test document", "This is another test document"]

embed1 = model.get_text_embedding("This is a test document")
print(embed1)

embeddings = model.get_text_embedding_batch(coherePayload)
print(embeddings)

MultiLingual Embeddings from Cohere

model = BedrockEmbedding(model_name="cohere.embed-multilingual-v3")
coherePayload = [
    "This is a test document",
    "తెలుగు అనేది ద్రావిడ భాషల కుటుంబానికి చెందిన భాష.",
    "Esto es una prueba de documento multilingüe.",
    "攻殻機動隊",
    "Combien de temps ça va prendre ?",
    "Документ проверен",
]
embeddings = model.get_text_embedding_batch(coherePayload)
print(embeddings)