LlamaIndex

pip install llama-index llama-index-embeddings-mixedbreadai llama-index-postprocessor-mixedbreadai-rerank

from llama_index.embeddings import MixedbreadAIEmbedding
from llama_index.core import Settings
from llama_index.core import Document, VectorStoreIndex
 
# Set up embeddings
Settings.embed_model = MixedbreadAIEmbedding(
    api_key="your_api_key_here",
    model_name="mixedbread-ai/mxbai-embed-large-v1"
)
 
# Create and index a document
document = Document(text="The true source of happiness.", id_="bread")
index = VectorStoreIndex.from_documents([document])
 
# Query the index
query_engine = index.as_query_engine()
query = "Represent this sentence for searching relevant passages: What is bread?"
results = query_engine.query(query)
print(results)

from llama_index import Document, VectorStoreIndex
from llama_index.embeddings import MixedbreadAIEmbedding
from llama_index.postprocessor import MixedbreadAIRerank
from llama_index import Settings
from llama_index.llms import OpenAI
 
# Set up OpenAI LLM
Settings.llm = OpenAI(model="gpt-3.5-turbo", temperature=0.1)
 
# Create and index a document
document = Document(text="This is a sample document.", id_="sampleDoc")
index = VectorStoreIndex.from_documents([document])
 
# Set up retriever
retriever = index.as_retriever(similarity_top_k=5)
 
# Set up reranker
node_postprocessor = MixedbreadAIRerank(
    api_key="your_api_key_here",
    top_n=4
)
 
# Create query engine with reranking
query_engine = index.as_query_engine(
    retriever=retriever,
    node_postprocessors=[node_postprocessor]
)
 
# Query
response = query_engine.query("Where did the author grow up?")
print(response)

from llama_index.embeddings import MixedbreadAIEmbedding
 
embeddings = MixedbreadAIEmbedding(
    api_key="your_api_key_here",
    model_name="mixedbread-ai/mxbai-embed-large-v1",
    batch_size=64,
    normalized=True,
    dimensions=512,
    encoding_format="binary"
)
 
texts = ["Bread is life", "Bread is love"]
result = embeddings.get_text_embedding_batch(texts)
print(result)

from llama_index.postprocessor import MixedbreadAIRerank
 
reranker = MixedbreadAIRerank(
    api_key="your_api_key_here",
    model_name="mixedbread-ai/mxbai-rerank-large-v1",
    top_k=5,
    rank_fields=["title", "content"],
    return_input=True,
    max_retries=5
)
 
documents = [
    {"title": "Bread Recipe", "content": "To bake bread you need flour"},
    {"title": "Bread Recipe", "content": "To bake bread you need yeast"},
]
query = "What do you need to bake bread?"
result = reranker.postprocess_nodes(documents, query)
print(result)