Data Sources

A data source is your connection to the knowledge base that your RAG system retrieves from. Vecta supports two categories: vector databases and file stores.

Vector Databases

Connect an existing vector database where your chunks are already stored and embedded. Vecta reads the chunks and metadata through a connector.

Supported databases:

Database	SDK Connector	Server Managed
ChromaDB (local)	`ChromaLocalConnector`	✅ (upload via API)
ChromaDB (cloud)	`ChromaCloudConnector`	❌ (reads directly)
Pinecone	`PineconeConnector`	❌ (reads directly)
pgvector	`PgVectorConnector`	❌ (reads directly)
Weaviate	`WeaviateConnector`	❌ (reads directly)
Azure Cosmos DB	`AzureCosmosConnector`	❌ (reads directly)
Databricks	`DatabricksConnector`	❌ (reads directly)
LangChain	`LangChainVectorStoreConnector`	✅ (upload via API)
LlamaIndex	`LlamaIndexConnector`	✅ (upload via API)

Every vector-database connector requires a VectorDBSchema that tells Vecta how to extract id, content, source_path, and page_nums from each record. See Accessor Syntax.

File Stores

Upload files directly and Vecta will ingest, chunk, and embed them automatically using markitdown. Supported formats include PDF, DOCX, PPTX, XLSX, TXT, HTML, and more.

Store	SDK Connector	Notes
Local files	`FileStoreConnector`	Ingested with markitdown
Upload via API	`VectaAPIClient.upload_local_files()`	Server handles embedding

File store connectors do not require a VectorDBSchema — Vecta handles the schema internally because it controls how the chunks are created and stored.

Connecting via the Platform

In the Data Sources dashboard you can:

Upload files — Drag and drop PDF, DOCX, and other files
Connect a vector database — Provide connection credentials and configure the schema through the UI
Import from Hugging Face — Pull standard datasets (see Hugging Face)

Connecting via the SDK

API Client (hosted)

from vecta import VectaAPIClient

client = VectaAPIClient()

# Upload local files
ds = client.upload_local_files(file_paths=["report.pdf", "manual.docx"])

# Connect Pinecone
ds = client.connect_pinecone(
    api_key="pk-...",
    index_name="my-index",
)

# Connect Chroma Cloud
ds = client.connect_chroma_cloud(
    tenant="my-tenant",
    database_name="my-db",
    api_key="ck-...",
    collection_name="documents",
)

Local Client

from vecta import VectaClient, ChromaLocalConnector, VectorDBSchema

schema = VectorDBSchema(
    id_accessor="id",
    content_accessor="document",
    metadata_accessor="metadata",
    source_path_accessor="metadata.source_path",
    page_nums_accessor="metadata.page_nums",
)

connector = ChromaLocalConnector(
    client=chroma_client,
    collection_name="my_docs",
    schema=schema,
)

vecta = VectaClient(data_source_connector=connector)
vecta.load_knowledge_base()

Next Steps

File Store Connectors — Details on file ingestion
Vector DB Connectors — Connector reference for every database
Accessor Syntax — How to write schema paths

Overview