Qdrant
High-performance open-source vector search engine written in Rust, designed for AI and ML applications. Provides a free cloud tier and self-hosted option with advanced filtering, payload storage, and approximate nearest neighbor search.
Use Cases
Free Tier
Free cloud tier: 1GB RAM, 4GB disk, 1 cluster with 1 node. Also free self-hosted (Apache 2.0) with no limits. Cloud free tier requires no credit card.
How to Maximize the Free Tier
Sign up at cloud.qdrant.io and create a free cluster (choose 'Free' plan) — you get 1GB RAM and 4GB disk with no credit card required. Alternatively, run 'docker run -p 6333:6333 qdrant/qdrant' for self-hosted local development. Use the Python or Node client to create a collection, set vector dimensions, and upsert embeddings. The free tier is sufficient for prototyping and small-scale apps with under 100K vectors.
Getting Started
1) For cloud: sign up at cloud.qdrant.io, create a free cluster, copy the API key. 2) For self-hosted: run 'docker run -p 6333:6333 qdrant/qdrant'. 3) Use the Python/Node client to create a collection, set vector size, and upsert vectors. 4) Query with a search vector and optional payload filters.
Pros
- Performance: Rust-based implementation delivers extremely low latency and high throughput for vector search, even with millions of vectors
- Filtering: Supports complex payload filtering (nested objects, ranges, geo) combined with vector similarity in a single query — rare among vector DBs
- Free Tier: Generous free cloud cluster (1GB RAM) lets you test and prototype without any cost or credit card
Cons
- Niche Focus: Primarily a vector search engine — not suited for traditional full-text search or log analytics without additional tooling
- Ecosystem: Smaller community and fewer integrations compared to Elasticsearch or OpenSearch, though client libraries exist for all major languages
- Cloud Limits: Free tier limited to 1GB RAM and 4GB disk, which may be insufficient for large-scale production workloads