Qdrant
App in the BluixApps catalog
What it is
Qdrant is a high-performance vector database written in Rust, designed for AI-powered search and recommendation at production scale. Open-source (Apache 2.0), single-binary deployment, gRPC + REST APIs, with hybrid (dense + sparse) search, payload filtering, and quantization for memory efficiency.
It's the backbone of RAG pipelines that need to scale beyond toy projects — million-vector collections, sub-100ms p99 latencies, horizontal sharding.
What it's for
- RAG retrieval at scale — embed your knowledge base, retrieve top-k passages for LLM context
- Semantic search — replace keyword search on docs, products, support tickets
- Recommendation systems — find similar items, users, content via vector similarity
- Multi-modal search — image, text, audio embeddings co-located in one collection
- Anomaly detection — outlier detection via vector distance thresholds
Who it's for
- AI engineers building production RAG and semantic search beyond proof-of-concept scale
- ML platform teams replacing Pinecone with self-hosted Qdrant for sovereignty + per-month cost predictability
- E-commerce engineering powering "find similar items" / personalized recommendations on millions of SKUs
- Search teams upgrading keyword-only to hybrid (dense + BM25) for relevance gains without re-indexing
- Researchers & academics working with multi-million vector datasets and needing reproducible local infra
Why teams pick Qdrant over alternatives
- Rust performance — sub-10ms query latency on million-vector collections
- Hybrid search — dense + sparse (BM25-style) combined natively
- Payload filtering — pre-filter by metadata before similarity, no Python re-scoring
- Quantization — INT8 + binary encoding cuts RAM 32× with minimal recall loss
- First-class clients — Python, JS, Rust, Go, Java, .NET, all type-safe
- Apache 2.0 — no commercial restrictions
- Snapshot + restore built into the binary
Integrations
- Client libraries — typed SDKs for Python, JS, Rust, Go, Java, .NET, PHP, Ruby
- LLM frameworks — LangChain, LlamaIndex, Haystack, Semantic Kernel ship Qdrant adapters
- Embedding providers — OpenAI, Cohere, Hugging Face, sentence-transformers, FastEmbed (built into Qdrant)
- Streaming ingestion — Apache Kafka / Pulsar via custom workers
- Backup — snapshot to local disk or S3-compatible object storage
- Observability — Prometheus metrics endpoint, distributed tracing via OpenTelemetry
- Protocols — gRPC (fast) + REST (universal); both auth-protected with API key
Notable users & community
- 20k+ GitHub stars
- Used by Disney, Visa, Bayer, X (Twitter), and many AI startups for production retrieval
- Strong Discord, monthly community calls, active engineering blog
- Common pairing with Flowise, AnythingLLM, n8n in self-hosted AI stacks
- Backed by Qdrant company (DE-based) — strong European OSS company with sustainable open-core model
Tips & operations
- Enable quantization —
quantization_config.scalar.type=int8cuts RAM 4×, binary cuts 32× with <2% recall loss - Create payload indexes before bulk insert —
create_payload_indexon filter fields speeds queries 10× post-insert - Run with replicas=2 even on a single VPS — protects against snapshot/data corruption without cross-node setup
- Snapshot weekly to S3 — built-in
/snapshotsendpoint + cron + S3 upload = cheap off-site backup - Use FastEmbed for built-in embedding — runs inside Qdrant; saves an external OpenAI Embeddings API round-trip
- Mind sharding above 10M vectors — single collection limits exist; design with
shard_numberfrom the start
What we ship in BluixApps
- Docker compose: Qdrant single-node (cluster mode available for Enterprise tier)
- Pinned
qdrant/qdrant:v1.13.0, weekly upstream tracking - API key auth enabled by default (random key shown in install report)
- Persistent storage volume at
/qdrant/storagefor collections + snapshots - gRPC + REST both exposed; HTTPS via Let's Encrypt on REST endpoint
- Pairs naturally with Flowise / AnythingLLM / n8n on same VPS for one-click RAG stack
- Backup hook captures storage volume + snapshot exports
Get this app — pick a BluixApps plan
Same catalog. Scaling tenant isolation, white-label and support tier.
| Tier | Tenants | Catalog | Support | White-label | Monthly | |
|---|---|---|---|---|---|---|
| Stacks | 1 | 19 curated stacks | Standard | — | $19/mo | DetailDeploy |
| Starter | 10 | Full catalog | Standard | +$15–25/mo | $49/mo | DetailDeploy |
| Pro | 25 | Full catalog | Priority bugfix | +$15–25/mo | $149/mo | DetailDeploy |
| Growth | 100 | Full catalog | Priority bugfix | +$15–25/mo | $349/mo | DetailDeploy |
| Scale | 500 | Full catalog | 7-day window | +$15–25/mo | $799/mo | DetailDeploy |
| Enterprise | Unlimited | Full catalog | Priority 7-day | Bundled | $1,499/mo | DetailDeploy |
