SinglebaseCloud

Provides a managed vector database service that stores high-dimensional embeddings and performs approximate nearest neighbor (ANN) search for semantic similarity queries. The system handles embedding generation, indexing with HNSW or similar algorithms, and retrieval-augmented generation (RAG) pipelines without requiring separate infrastructure management. Integrates with LLM providers to automatically embed documents and queries for semantic matching.

Provides a managed document database (similar to MongoDB or Firestore) that stores semi-structured JSON documents with flexible schemas, supporting nested objects, arrays, and dynamic field addition. Includes indexing on arbitrary fields, querying with filter operators, and transactions for multi-document consistency. Designed to coexist with the vector database for storing document metadata, user data, and application state without requiring a separate database service.

Provides built-in authentication infrastructure supporting multiple identity providers (OAuth2, SAML, email/password, social login) with session management, JWT token generation, and role-based access control (RBAC). Integrates directly with the document and vector databases to enforce row-level and field-level access policies, preventing unauthorized data access at the database layer rather than application layer.

Provides real-time change streams and WebSocket-based subscriptions that notify clients when documents or vectors are created, updated, or deleted. Clients can subscribe to specific collections, queries, or document IDs and receive live updates without polling. Useful for collaborative applications, live dashboards, and reactive UIs that need to reflect backend changes instantly.

Allows developers to write and deploy serverless functions (similar to AWS Lambda or Vercel Functions) that have direct, pre-authenticated access to Singlebase databases, vectors, and auth context. Functions receive request context including authenticated user information and can query/mutate data without additional authentication steps. Supports scheduled execution (cron jobs) and event-driven triggers (on document changes, user actions).

Provides a system for generating, rotating, and revoking API keys that enable service-to-service communication and third-party integrations. Keys can be scoped to specific collections, operations (read/write), and rate limits. Integrates with the auth layer to allow API key authentication alongside user authentication, enabling both client applications and backend services to access Singlebase APIs securely.

Automatically generates embeddings for text fields in documents using integrated LLM providers (OpenAI, Anthropic, etc.) and stores them in the vector database. When documents are created or updated, the system detects text changes and regenerates embeddings without manual intervention. Supports batch embedding operations for backfilling existing documents and configurable embedding models to balance cost and quality.

Provides full-text search capabilities that index document text fields and support keyword queries with boolean operators, phrase matching, and field-specific searches. Integrates with the document database to enable hybrid search combining full-text relevance with semantic vector similarity and structured filters. Supports configurable analyzers (tokenization, stemming) and custom stop words for language-specific search optimization.

Provides automated backup mechanisms that periodically snapshot the entire database state, enabling point-in-time recovery to restore data to any previous state. Supports exporting data in standard formats (JSON, CSV) for analysis or migration to other systems. Backups are encrypted and stored redundantly to prevent data loss from accidental deletion or corruption.

Provides APIs for bulk insert, update, and delete operations that process large datasets efficiently without individual round-trips. Supports importing data from CSV, JSON, or other formats with schema validation and error handling. Batch operations are transactional (all-or-nothing) and optimized for throughput, making them suitable for data migrations, periodic syncs, and large-scale updates.