Search, classify and moderate video, images, audio and documents on object storage
One data platform for every engine your application needs
Stage
Not given
Not given
Founded
Not given
Not given
Based in
US
IN
Pricing model
Freemium
Not given
Pricing
MVS Build: $25/month with up to 1M vectors and $0/month in indexing included. MVS Scale: $250/month with up to 25M vectors. Managed Build: $25/month with $10 included usage pool. Managed Scale: $250/month with $100 included usage pool. Usage rates: video $0.05/minute, audio $0.01/minute, images and
Not given
Key features
Hybrid dense, sparse and BM25 retrieval
Face and scene extraction from video
Speech transcription and speaker diarization
Multimodal embeddings from 50+ models
Object storage-native vector persistence
MCP server for agent integration
LangChain integration
Self-hosted deployment option
PostgreSQL-compatible SQL with ACID transactions, views, triggers, PL/pgSQL functions and window functions
Redis-compatible cache with all data types, Lua scripts, pub/sub, blocking commands and persistence
Kafka-compatible streaming with partitioned streams, consumer groups and transactions
S3-compatible object storage with versioning and multipart uploads
Full-text search with ranking and highlights
Cluster management with automatic failover and data replication
Row-level security enabled by default on all tables
Data migration from PostgreSQL, MySQL, SQLite, Redis, Kafka and S3 systems
Official SDKs in TypeScript, Python, C# and Java
pgvector and PostGIS support with HNSW indexes and spatial indexes
What makes it different
Works directly with files in S3, GCS, Azure or R2 without moving data
Sub-100ms query latency even with vectors on object storage
Multi-stage retrieval pipelines in single API call
Unified platform combining SQL, cache, streams, storage and search in a single binary with one identity and backup
Compatible with existing client libraries and protocols so no application changes required
Automatic failover and cluster scaling with every part replicated across servers
Tested against real systems with 36,000 SQL queries matching PostgreSQL results
Data sovereignty with complete data ownership and no proprietary lock-in
Mixpeek
Propus