llama-index-packs-fusion-retriever

Fusion Retriever Packs

Hybrid Fusion Pack

This LlamaPack provides an example of our hybrid fusion retriever method.

This specific template fuses results from our vector retriever and bm25 retriever; of course, you can provide any template you want.

Check out the notebook here.

CLI Usage

You can download llamapacks directly using llamaindex-cli, which comes installed with the llama-index python package:

llamaindex-cli download-llamapack HybridFusionRetrieverPack --download-dir ./hybrid_fusion_pack

You can then inspect the files at ./hybrid_fusion_pack and use them as a template for your own project.

Code Usage

You can download the pack to a the ./hybrid_fusion_pack directory:

from llama_index.core.llama_pack import download_llama_pack

# download and install dependencies
HybridFusionRetrieverPack = download_llama_pack(
    "HybridFusionRetrieverPack", "./hybrid_fusion_pack"
)

From here, you can use the pack, or inspect and modify the pack in ./hybrid_fusion_pack.

Then, you can set up the pack like so:

# create the pack
hybrid_fusion_pack = HybridFusionRetrieverPack(
    nodes, chunk_size=256, vector_similarity_top_k=2, bm25_similarity_top_k=2
)

The run() function is a light wrapper around query_engine.query().

response = hybrid_fusion_pack.run("Tell me about a Music celebritiy.")

You can also use modules individually.

# use the fusion retriever
nodes = hybrid_fusion_pack.fusion_retriever.retrieve("query_str")

# use the vector retriever
nodes = hybrid_fusion_pack.vector_retriever.retrieve("query_str")
# use the bm25 retriever
nodes = hybrid_fusion_pack.bm25_retriever.retrieve("query_str")

# get the query engine
query_engine = hybrid_fusion_pack.query_engine

Query Rewriting Retriever Pack

This LlamaPack provides an example of query rewriting through our fusion retriever.

This specific template takes in a single retriever, and generates multiple queries against the retriever, and then fuses the results together.

Check out the notebook here.

CLI Usage

You can download llamapacks directly using llamaindex-cli, which comes installed with the llama-index python package:

llamaindex-cli download-llamapack QueryRewritingRetrieverPack --download-dir ./query_rewriting_pack

You can then inspect the files at ./query_rewriting_pack and use them as a template for your own project.

Code Usage

You can download the pack to a the ./query_rewriting_pack directory:

from llama_index.core.llama_pack import download_llama_pack

# download and install dependencies
QueryRewritingRetrieverPack = download_llama_pack(
    "QueryRewritingRetrieverPack", "./query_rewriting_pack"
)

From here, you can use the pack, or inspect and modify the pack in ./query_rewriting_pack.

Then, you can set up the pack like so:

# create the pack
query_rewriting_pack = QueryRewritingRetrieverPack(
    nodes,
    chunk_size=256,
    vector_similarity_top_k=2,
)

The run() function is a light wrapper around query_engine.query().

response = query_rewriting_pack.run("Tell me a bout a Music celebritiy.")

You can also use modules individually.

# use the fusion retriever
nodes = query_rewriting_pack.fusion_retriever.retrieve("query_str")

# use the vector retriever
nodes = query_rewriting_pack.vector_retriever.retrieve("query_str")

# get the query engine
query_engine = query_rewriting_pack.query_engine