/ V0.16.1

The Vector Database Built for Production AI.

Billion-vector search on the storage you already operate. Encrypted in use, not just at rest.

Start free Request a demo

TRUSTED BY

cyborgdb · client.ts

> client.query( "patient SSN 123-45-6789" → "RlcjognV5Ikx2QaP" )

QUERY

> embed( [0.127, -0.384, 0.552, ...] )

VECTOR

> encrypt( dims=1536, norm=L2 )

SEALED

> store( Dx: Hypertension )

INDEXED

> match( top_k=10, ε=0.02 )

> retrieve( match: Dx: Hypertension )

RESULT

> decrypt( auth: tenant_a9f2 )

REVEAL

LATENCY 8ms

THROUGHPUT 80.4k/s

KEY BYOK / HYOK

Encrypted at search time, not just at rest.

At-rest encryption only protects against disk theft. CyborgDB keeps the index ciphertext through transit and search, too. Your KMS holds the only key that can unseal anything.

At rest

The index is ciphertext.

Embeddings, IDs, and metadata are stored as ciphertext with per-record IVs and AEAD authentication. Compromise the disk, read nothing.

Records AES-256-GCM

Index nodes ciphertext-only

In transit

Encrypted before it hits the network.

Payloads are AEAD-encrypted by the client before transport. TLS is layered on top, but it isn't what's keeping the data safe.

Payload AES-256-GCM

Transport TLS

In use

Search runs over ciphertext.

Forward-secure ANN search runs on encrypted index nodes. Your server never holds plaintext embeddings or full index keys.

Search tokens scoped

Index traversal ciphertext

Keys Client-side custody. BYOK or HYOK. Nothing in the server-side pipeline can decrypt without your key.

Why teams rely on CyborgDB.

SOC 2 Type II CLEARED

HIPAA READY

GDPR READY

ISO 27001 IN PROGRESS

Pass the security review on the first pass.

End-to-end encryption and per-tenant key custody answer the questions an enterprise security team is going to ask. Audits stop being a quarter-long detour.

YOUR STACK

postgres redis s3

proxy

Run on the storage you already operate.

Proxy mode sits in front of Postgres, Redis, or S3-compatible storage. Your data plane stays put. No new database for your team to staff.

Mathematical boundaries between tenants.

Every tenant gets their own key. One bug in your filter logic doesn't cross a cryptographic boundary, because the bytes are unreadable to begin with.

Encrypted. Faster than most plaintext.

On wiki-all-1M at 99% recall, CyborgDB beats Weaviate, Milvus, Elasticsearch, Qdrant, LanceDB and pgvector — encrypted.

Recall (%) vs queries per second across vector databases.
Database	Recall (%)	QPS
CyborgDB (encrypted)	72.8	792
CyborgDB (encrypted)	76.3	744
CyborgDB (encrypted)	81.3	672
CyborgDB (encrypted)	85.9	638
CyborgDB (encrypted)	88.8	567
CyborgDB (encrypted)	91.8	499
CyborgDB (encrypted)	94.0	433
CyborgDB (encrypted)	96.2	351
CyborgDB (encrypted)	97.3	321
CyborgDB (encrypted)	98.2	266
CyborgDB (encrypted)	98.6	245
CyborgDB (encrypted)	99.0	220
CyborgDB (encrypted)	99.4	182
CyborgDB (encrypted)	99.8	151
Qdrant	89.5	113
Qdrant	95.2	89
Qdrant	97.8	68
Qdrant	99.1	45
Qdrant	99.4	35
Qdrant	99.7	27
Qdrant	99.9	17
Weaviate	77.2	840
Weaviate	83.4	737
Weaviate	86.9	715
Weaviate	91.3	600
Weaviate	93.3	533
Weaviate	95.9	430
Weaviate	97.5	349
Weaviate	98.6	266
Weaviate	98.9	224
Weaviate	99.4	176
Weaviate	99.8	108
Milvus	92.6	66
Milvus	96.7	61
Milvus	98.3	55
Milvus	99.1	47
Milvus	99.5	41
Milvus	99.7	29
Elasticsearch	81.4	303
Elasticsearch	83.4	293
Elasticsearch	89.6	278
Elasticsearch	90.8	276
Elasticsearch	94.6	241
Elasticsearch	95.2	240
Elasticsearch	96.6	218
Elasticsearch	97.5	198
Elasticsearch	97.8	184
Elasticsearch	98.7	161
Elasticsearch	99.2	131
Elasticsearch	99.2	129
Elasticsearch	99.4	120
Elasticsearch	99.4	117
pgvector	81.4	835
pgvector	89.2	497
pgvector	94.1	402
pgvector	96.7	283
pgvector	97.6	190
pgvector	98.5	130
pgvector	99.1	82
pgvector	99.5	47
pgvector	99.6	38
LanceDB	94.2	282
LanceDB	97.5	174
LanceDB	99.0	101
LanceDB	99.3	90

CyborgDB encrypted

Qdrant

Weaviate

Milvus

Elasticsearch

pgvector

LanceDB

QPS @ 99% recall

CyborgDB v0.17.0 DiskIVF encrypted 220qps

Weaviate v1.37.0 HNSW 214qps

Elasticsearch v9.3.4 HNSW 143qps

LanceDB v0.27.1 IVF-PQ 101qps

pgvector v0.8.2 HNSW 90qps

Milvus v2.6.15 HNSW 48qps

Qdrant v1.17.1 HNSW 47qps

higher = better

DATASET wiki-all-1M · 768 dims · 1M vectors · top-k = 10

Single-threaded runs on the ann-benchmarks harness, c8g.4xlarge · May 2026.

What teams build when the data can't leave their control.

All use cases → · Industries →

RAG on sensitive data Clinical notes, contracts, PII — never in plaintext at retrieval.

Agent memory & audit Per-user decryption, every retrieval written to the ledger.

Biometric vector search new Faces, voices, prints — matched on ciphertext, templates stay sealed.

Multi-tenant search Cryptographic isolation between tenants — not just row-level filtering.

Private embeddings Encrypt before you embed. Vectors never exist in the clear on the server.

Secure RAG,
at NVIDIA scale.

The Cyborg–NVIDIA Enterprise RAG Blueprint is a reference architecture for deploying secure AI applications at NVIDIA scale. Cyborg is one of eight companies NVIDIA invited to author an Enterprise Blueprint.

NVIDIA NIMs handle the inference. CyborgDB handles encrypted-in-use retrieval across multimodal corpora.

15× faster multimodal PDF extraction
50% fewer incorrect answers
Zero plaintext exposure of vector data

View on build.nvidia.com

Cyborg — NVIDIA Reference Architecture

Cyborg–NVIDIA Enterprise RAG Blueprint reference architecture diagram, showing the retrieval and extraction pipelines that combine NVIDIA NIMs with CyborgDB.

Pull, configure, ship.

A single Docker image, your existing storage and KMS, and one import change.

0m15m30m45m60m

Step 01 ~2 min

Install

shell

# Pull CyborgDB image

docker pull cyborg/cyborgdb:latest

# Launch CyborgDB service

docker run -p 8000:8000 cyborg/cyborgdb

Step 02 ~10 min

Configure

cyborg.yaml

backend:

kind: postgres

dsn: ${POSTGRES_URL}

keys:

kms: aws

key_id: alias/cyborg-prod

audit:

sink: s3://acme-audit/cyborg/

Step 03 ~45 min

Ship

from cyborgdb import Client

# key never leaves your app

index_key = client.generate_key()

index = client.create_index(index_name="patient-notes", index_key=index_key)

# search runs on ciphertext

index.upsert(items)

results = index.query(query_vectors=q, top_k=10)

Shipped Encrypted vector search running in your VPC, backed by your existing infra, keys in your KMS.

Start free Read the docs → Docs

Migrate your vector database.
Encrypted.

Already on Pinecone, Weaviate, or Milvus? Bring your workload to a thirty-minute call. We'll show you how CyborgDB performs against it. Free for up to 1M vectors.

Request a demo Start free