ai-infra / API Reference

VectorStore

from ai_infra.embeddings import VectorStore

ai_infra.embeddings

Simple vector store for semantic search. Store documents and search by meaning using embeddings. Supports multiple backends: in-memory, Chroma, FAISS. Features: - Simple API: `add()`, `search()`, `delete()` - Multiple backends: memory, chroma, faiss - Auto-embedding: Just add text, embeddings handled automatically - Metadata filtering: Filter search results by metadata

python

from ai_infra import Embeddings, VectorStore

    # Create with embeddings
    embeddings = Embeddings()
    store = VectorStore(embeddings=embeddings)

    # Add documents
    store.add_texts(["Python is great", "JavaScript is popular"])

    # Search
    results = store.search("programming languages", k=2)
    for result in results:
        print(f"{result.score:.2f}: {result.document.text}")

Example - With metadata:

python

store.add_texts(
        texts=["Doc 1", "Doc 2"],
        metadatas=[{"source": "web"}, {"source": "book"}]
    )

    # Filter by metadata
    results = store.search("query", filter={"source": "web"})

Example - Persistent storage with Chroma:

python

store = VectorStore(
        embeddings=embeddings,
        backend="chroma",
        persist_directory="./my_db"
    )

Constructor

VectorStore(embeddings: Embeddings, backend: Literal['memory', 'chroma', 'faiss'] = 'memory', collection_name: str = 'default', persist_directory: str | None = None, kwargs: Any = {}) -> None

Parameter	Type	Default	Description
`embeddings`required	`Embeddings`	—	Embeddings instance for generating vectors.
`backend`	`Literal['memory', 'chroma', 'faiss']`	'memory'	Storage backend - "memory", "chroma", or "faiss".
`collection_name`	`str`	'default'	Name for the collection (chroma only).
`persist_directory`	`str\|None`	None	Directory for persistent storage.
`kwargs`	`Any`	{}	—

Methods

from ai_infra import Embeddings, VectorStore # Create with embeddings embeddings = Embeddings() store = VectorStore(embeddings=embeddings) # Add documents store.add_texts(["Python is great", "JavaScript is popular"]) # Search results = store.search("programming languages", k=2) for result in results: print(f"{result.score:.2f}: {result.document.text}")

store.add_texts( texts=["Doc 1", "Doc 2"], metadatas=[{"source": "web"}, {"source": "book"}] ) # Filter by metadata results = store.search("query", filter={"source": "web"})

Parameter

Type

Default

Description

embeddingsrequired

Embeddings

—

Embeddings instance for generating vectors.

backend

Literal['memory', 'chroma', 'faiss']

'memory'

Storage backend - "memory", "chroma", or "faiss".

collection_name

str

'default'

Name for the collection (chroma only).

persist_directory

str|None

None

Directory for persistent storage.

kwargs

Any

{}

—