thoth-qdrant 0.1.5

Native Qdrant implementation for Thoth Vector Database

Thoth Qdrant

A native Qdrant implementation for the Thoth Vector Database system, providing high-performance vector storage and similarity search capabilities without Haystack dependencies.

Features

• Native Qdrant Integration: Direct use of Qdrant client without Haystack
• Full API Compatibility: Same interface as thoth_vdb2 for seamless integration
• External Embeddings: Support for OpenAI, Cohere, Mistral, and HuggingFace
• Document Types: EvidenceDocument, SqlDocument, ColumnNameDocument
• Similarity Search: Native Qdrant search with document type filtering
• Batch Operations: Efficient bulk document insertion
• Caching: Intelligent embedding cache for performance

Installation

# Basic installation
pip install thoth-qdrant

# With OpenAI embeddings support
pip install thoth-qdrant[openai]

# With all embedding providers
pip install thoth-qdrant[all-providers]

Configuration

Environment Variables

# Embedding provider configuration
export EMBEDDING_PROVIDER=openai
export EMBEDDING_MODEL=text-embedding-3-small
export OPENAI_API_KEY=your-api-key

# Or use provider-specific keys
export OPENAI_API_KEY=sk-...
export COHERE_API_KEY=...
export MISTRAL_API_KEY=...

Qdrant Setup

Ensure Qdrant is running locally:

docker run -p 6333:6333 qdrant/qdrant

Usage

from thoth_qdrant import VectorStoreFactory
from thoth_qdrant.core.base import (
    ColumnNameDocument,
    SqlDocument,
    EvidenceDocument,
    ThothType,
)

# Create vector store
store = VectorStoreFactory.create(
    backend="qdrant",
    collection="my_collection",
    host="localhost",
    port=6333,
    embedding_provider="openai",
    embedding_model="text-embedding-3-small"
)

# Add documents
column_doc = ColumnNameDocument(
    table_name="users",
    column_name="email",
    original_column_name="email_address",
    column_description="User email for authentication",
    value_description="Valid email format"
)
doc_id = store.add_column_description(column_doc)

sql_doc = SqlDocument(
    question="How to find recent users?",
    sql="SELECT * FROM users WHERE created_at > NOW() - INTERVAL '30 days'",
    evidence="Filter by date using interval"
)
store.add_sql(sql_doc)

# Search similar documents
results = store.search_similar(
    query="user email authentication",
    doc_type=ThothType.COLUMN_NAME,
    top_k=5,
    score_threshold=0.7
)

# Bulk operations
documents = [column_doc, sql_doc]
doc_ids = store.bulk_add_documents(documents)

# Get document by ID
doc = store.get_document(doc_id)

# Delete document
store.delete_document(doc_id)

# Get all documents by type
all_columns = store.get_all_column_documents()
all_sql = store.get_all_sql_documents()

# Collection info
info = store.get_collection_info()
print(info)

API Reference

VectorStoreInterface Methods

• add_column_description(doc: ColumnNameDocument) -> str
• add_sql(doc: SqlDocument) -> str
• add_evidence(doc: EvidenceDocument) -> str
• search_similar(query: str, doc_type: ThothType, top_k: int = 5, score_threshold: float = 0.7) -> List[BaseThothDocument]
• get_document(doc_id: str) -> Optional[BaseThothDocument]
• delete_document(doc_id: str) -> None
• bulk_add_documents(documents: List[BaseThothDocument]) -> List[str]
• delete_collection(thoth_type: ThothType) -> None
• get_all_column_documents() -> List[ColumnNameDocument]
• get_all_sql_documents() -> List[SqlDocument]
• get_all_evidence_documents() -> List[EvidenceDocument]
• get_collection_info() -> Dict[str, Any]

Testing

# Run tests with local Qdrant
pytest tests/

# Run specific test
pytest tests/test_qdrant_adapter.py -v

# With coverage
pytest --cov=thoth_qdrant tests/

Development

# Install development dependencies
pip install -e .[dev,test]

# Format code
black thoth_qdrant tests
isort thoth_qdrant tests

# Type checking
mypy thoth_qdrant

# Linting
ruff thoth_qdrant

License

Apache License 2.0 - See LICENSE.md for details

Compatibility

This library is fully compatible with thoth_vdb2 API, allowing seamless migration from Haystack-based implementations to native Qdrant.