Query AWS Athena with Natural Language

Athena Reader allows you to execute SQL with AWS Athena. It uses SQLAlchemy and PyAthena under the hood. The source material demonstrates creating an Athena engine using AthenaReader.create_athena_engine() that can be used with LlamaIndex's SQLDatabase and NLSQLTableQueryEngine components.

Use Athena Reader when you need to execute SQL queries with AWS Athena. The source material demonstrates usage with LlamaIndex components for querying Athena databases. The documentation highly recommends using this loader with AWS EC2 IAM role.

Do not use this reader if you're not working with AWS Athena, as it requires Athena-specific configuration including AWS region, S3 staging directory, database name, and workgroup.

You provide AWS configuration parameters including region, S3 staging directory, database name, workgroup, and table names. The example demonstrates providing a query string ("Which blocknumber has the most transactions?") to a query engine. The reader returns query responses from your Athena database.

Athena Reader works with:

• AWS Athena: Executes SQL queries with Athena
• SQLAlchemy: Used as the underlying database abstraction layer
• PyAthena: Provides the Athena-specific database driver
• LlamaIndex Core: Demonstrated with SQLDatabase and NLSQLTableQueryEngine components
• OpenAI: The example shows usage with OpenAI LLMs (GPT-4 model)

Athena Reader is for developers who need to execute SQL queries with AWS Athena. Users should be familiar with AWS IAM permissions, S3 configuration, and the LlamaIndex components demonstrated in the example.

pip install llama-index-readers-athena

pip install llama-index-llms-openai

Example usage:

import os
import dotenv
from llama_index.core import SQLDatabase,ServiceContext
from llama_index.core.query_engine import NLSQLTableQueryEngine
from llama_index.llms.openai import OpenAI
from llama_index.readers.athena import AthenaReader

dotenv.load_dotenv()

AWS_REGION = os.environ['AWS_REGION']
S3_STAGING_DIR = os.environ['S3_STAGING_DIR']
DATABASE = os.environ['DATABASE']
WORKGROUP = os.environ['WORKGROUP']
TABLE = os.environ['TABLE']

llm = OpenAI(model="gpt-4",temperature=0, max_tokens=1024)

engine = AthenaReader.create_athena_engine(
    aws_region=AWS_REGION,
    s3_staging_dir=S3_STAGING_DIR,
    database=DATABASE,
    workgroup=WORKGROUP
)

service_context = ServiceContext.from_defaults(
  llm=llm
)

sql_database = SQLDatabase(engine, include_tables=[TABLE])

query_engine = NLSQLTableQueryEngine(
    sql_database=sql_database,
    tables=[TABLE],
    service_context=service_context
)
query_str = (
    "Which blocknumber has the most transactions?"
)
response = query_engine.query(query_str)