FLARE Query Engine

Adapted from the paper "Active Retrieval Augmented Generation"

Currently implements FLARE Instruct, which tells the LLM to generate retrieval instructions.

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

python

%pip install llama-index-llms-openai

python

!pip install llama-index

python

from llama_index.llms.openai import OpenAI
from llama_index.core import Settings

Settings.llm = OpenAI(model="gpt-4", temperature=0)
Settings.chunk_size = 512

Download Data

python

!mkdir -p 'data/paul_graham/'
!wget 'https://raw.githubusercontent.com/run-llama/llama_index/main/docs/examples/data/paul_graham/paul_graham_essay.txt' -O 'data/paul_graham/paul_graham_essay.txt'

Load Data

python

from llama_index.core import SimpleDirectoryReader, VectorStoreIndex

documents = SimpleDirectoryReader("./data/paul_graham").load_data()
index = VectorStoreIndex.from_documents(
    documents,
)

python

index_query_engine = index.as_query_engine(similarity_top_k=2)

python

from llama_index.core.query_engine import FLAREInstructQueryEngine

flare_query_engine = FLAREInstructQueryEngine(
    query_engine=index_query_engine,
    max_iterations=7,
    verbose=True,
)

python

response = flare_query_engine.query(
    "Can you tell me about the author's trajectory in the startup world?"
)

python

print(response)

python

response = flare_query_engine.query(
    "Can you tell me about what the author did during his time at YC?"
)

python

print(response)

python

response = flare_query_engine.query(
    "Tell me about the author's life from childhood to adulthood"
)

python

print(response)

python

response = index_query_engine.query(
    "Can you tell me about the author's trajectory in the startup world?"
)

python

print(str(response))

python

response = index_query_engine.query(
    "Tell me about the author's life from childhood to adulthood"
)

python

print(str(response))