PyRAG

A library that makes it easy to get started with RAG applications with SingleStore.

Installation

To install the package run:

pip install pyrag

Usage Example

1. Import the PyRAG class from the pyrag package

from pyrag import PyRAG

2. Create a PyRAG instance

pyrag = PyRAG(
    connection_url='DB_CONNECTION_URL',
    embedding_model_name='text-embedding-3-small',
    openai_api_key='OPENAI_API_KEY',
)

Note, you can also use HugginFace models. In this case you need to provide a huggingfacehub_api_token.

3. Provide a knowledge source in the form of a URL or S3 bucket

Supported file types: csv, json, txt, pdf

URL Knowledge Source

pyrag.files.url.sync_file(url="URL", table_name='TABLE_NAME')

S3 Bucket Knowledge Source

pyrag.files.s3(
    access_key_id='AWS_ACCESS_KEY_ID',
    secret_access_key='AWS_SECRET_ACCESS_KEY',
    bucket_name='AWS_BUCKET_NAME',
).sync_files(
    # This parameter is optional. If you want to use all files from the s3 bucket, remove this parameter.
    allowed_files=['file_name_1', 'file_name_2', 'file_name_3'],
    # This parameter is optional. Use it if you want to rename the table. By default, the table serializes the file name.
    table_names={'file_name_1': 'file_name_1'}

4. Create a chat instance and chat session

chat = pyrag.chat.create(
    id=1,
    model_name='gpt-3.5-turbo',
    knowledge_sources=[{'table': 'file_name_1'}],
    store=True,
    store_messages_history=True
)

chat_session = chat.create_session(id=1)

5. Use the chat instance

response = chat_session.send('PROMPT')
print(response)

Find more usage examples here