Stores

Stores form the foundation of the AgenticRAG system, providing a consistent way to manage various data types. Each store is specialized for a particular data format and typically connects to a storage backend.

Available Storage Backends

The library provides two built-in storage backends: - SQLBackend: Uses relational databases for structured data storage - ChromaBackend: Uses vector databases for embedding-based retrieval

You can implement custom backends by inheriting from the BaseBackend abstract class and implementing methods for CRUD.

from agenticrag.stores import BaseBackend
from agenticrag.types import BaseData

class MyCustomBackend(BaseBackend[SchemaType]):
    def __init__(self, schema: Type[SchemaType]...):
        # Implement Your logic

    def add(self, data):
        # Implement adding data
        pass

    def get(self, id):
        # Implement data retrieval
        pass

    # Implement other required methods

To make store for specific type of data, you can inherit any of the backend as required and implement with specific schema, model or other parameters as required.

Core Store Types

The library includes four primary store types:

MetaStore

Maintains metadata about all datasets, essential for decision-making by the RAGAgent.

from agenticrag.stores import MetaStore
from agenticrag.types import MetaData, DataFormat

# Initialize metastore
meta_store = MetaStore(connection_url="sqlite:///agenticrag.db")

# Add dataset metadata
meta_store.add(
    MetaData(
        name="company_docs",
        format=DataFormat.TEXT,
        description="Internal company documentation",
        tags=["internal", "documentation"]
    )
)

# Query metadata
text_datasets = meta_store.filter(format=DataFormat.TEXT)

TextStore

Manages text data with support for embeddings and semantic search.

from agenticrag.stores import TextStore
from agenticrag.types import TextData

# Initialize with persistent vector DB
text_store = TextStore(persistent_dir="./vector_db")

# Add text data
text_store.add(
    TextData(
        id="doc_123",
        name="architecture_overview",
        text="The system follows a microservice architecture...",
        source="internal_wiki"
    )
)

# Search similar content
similar_texts = text_store.search_similar(
    "system design patterns", 
    document_name="architecture_overview", 
    top_k=10
)

TableStore

Specialized for tabular data like CSVs and dataframes.

from agenticrag.stores import TableStore
from agenticrag.types import TableData

# Initialize 
table_store = TableStore(connection_url="sqlite:///agenticrag.db")

# Add table reference
table_store.add(
    TableData(
        id=1,
        name="sales_data",
        path="data/sales_2023.csv",
        structure_summary="""
        Sales data with columns:
        - date (datetime): Transaction date
        - product_id (int): Product identifier
        - quantity (int): Units sold
        - revenue (float): Total revenue
        - region (str): Sales region
        """,
    )
)

ExternalDBStore

Interfaces with external databases via connection strings.

from agenticrag.stores import ExternalDBStore

# Initialize
external_db_store = ExternalDBStore(connection_url="sqlite:///agenticrag.db")

# Add database connection
external_db_store.add_connection(
    name="production_db",
    connection_url="postgresql://user:pass@localhost:5432/production",
    description="Production database with customer and order tables",
    tables=["customers", "orders", "products"]
)