LlamaVector

Handles vector embedding generation, storage, and similarity search within the LlamaAI Ecosystem. Provides interfaces to various embedding models and vector databases.

Features

• Embedding Generation: Supports multiple embedding models (e.g., via sentence-transformers).
• Vector Storage Adapters: Interfaces for various vector databases (e.g., FAISS, ChromaDB, Pinecone, Qdrant, Weaviate).
• Similarity Search: Efficiently find vectors similar to a query vector.
• Data Models: Pydantic models for structured vector data.
• Indexing Utilities: Tools for building and managing vector indexes.
• (Optional) API: Can expose functionality via a FastAPI server.

Installation

# Core installation
pip install llamavector

# To install with specific vector database support (e.g., ChromaDB):
pip install llamavector[chromadb]

# To install with API support:
pip install llamavector[api]

Quick Start

# Example (TBD after code migration)
# from llamavector import VectorStore, EmbeddingModel

# model = EmbeddingModel(model_name='all-MiniLM-L6-v2')
# store = VectorStore(adapter='chromadb', collection_name='my_vectors')

# texts = ["This is the first document.", "This document is the second document."]
# embeddings = model.encode(texts)
# ids = ["doc1", "doc2"]

# store.add(ids=ids, embeddings=embeddings)

# query_embedding = model.encode(["A query about the second doc"])
# results = store.search(query_embeddings=query_embedding, k=1)
# print(results)

Contributing

Contributions are welcome! Please see CONTRIBUTING.md.

License

MIT License. See LICENSE file.