enVector provides secure vector search using ES2 (Encrypted Similarity Search) based on Fully Homomorphic Encryption (FHE). This repository contains self-hosted deployment assets and Python SDK notebooks.
- End-to-end encrypted vector search
- Docker Compose and Kubernetes (Helm) deployment support
- Multi-service architecture for scaling and HA
- Python SDK notebooks for quick start and API flow
enVector consists of five main services:
- envector-endpoint: API gateway and client entrypoint
- envector-backend: metadata/service management
- envector-orchestrator: request coordination and scheduling
- envector-compute: encrypted vector compute workers
- envector-shaper: shard split/merge and storage shaping tasks
Infrastructure dependencies:
- PostgreSQL (metadata)
- Object storage (S3-compatible or GCS via Helm values)
envector-deployment/
├── docker-compose/
│ ├── docker-compose.envector.yml
│ ├── docker-compose.infra.yml
│ ├── docker-compose.gpu.yml
│ ├── start_envector.sh
│ └── README.md
├── kubernetes-manifests/
│ ├── helm/
│ │ ├── Chart.yaml
│ │ ├── values.yaml
│ │ └── templates/
│ └── README.md
└── notebooks/
├── 00-quick-start.ipynb
├── 01-api-flow.ipynb
├── 02-simple-rag.ipynb
├── 03-rag-with-langchain.ipynb
├── 04-ann-api-flow.ipynb
└── 05-insert-load-capacity.ipynb
Important: Docker images are in private repositories. Contact heaan for access.
See full guide: docker-compose/README.md
git clone https://github.com/CryptoLabInc/envector-deployment.git
cd envector-deployment/docker-compose
# optional (script auto-creates from .env.example if missing)
cp .env.example .env
# preflight + up (docker, image access, license token)
./start_envector.shUseful variants:
# print merged compose config only
./start_envector.sh --config
# GPU override
./start_envector.sh --gpu
# scale compute/orchestrator
./start_envector.sh --num-compute 4 --num-orchestrator 2
# stop stack
./start_envector.sh --downSee full guide: kubernetes-manifests/README.md
git clone https://github.com/CryptoLabInc/envector-deployment.git
cd envector-deployment/kubernetes-manifests
# install chart
helm install envector ./helmFor production, configure before install/upgrade:
externalServices.metadatadb/externalServices.storagecompute.license(token secret creation orexistingSecret)externalSecrets.*if using External Secrets Operator
| Variable | Description | Default |
|---|---|---|
ENVECTOR_ENDPOINT_HOST_PORT |
Host port for gRPC endpoint | 50050 |
ENVECTOR_HTTP_HEALTH_HOST_PORT |
Host port for HTTP health/admin | 18080 |
ENVECTOR_ADMIN_API_ENABLED |
Enable admin API on endpoint | true |
ENVECTOR_LOG_LEVEL |
Service log level | INFO |
ENVECTOR_COMPUTE_TAG |
Compute image tag | latest |
ENVECTOR_LICENSE_TOKEN |
In-container license token path | /envector/license/token.jwt |
ENVECTOR_DB_* |
PostgreSQL connection parts | varies |
ENVECTOR_STORAGE_* |
Storage connection settings | varies |
kubernetes-manifests/helm/values.yaml 주요 항목:
endpoint.*,backend.*,orchestrator.*,compute.*,shaper.*externalServices.metadatadb.*,externalServices.storage.*compute.license.*(createSecret/existingSecret/mountAsFile/injectAsEnv)externalSecrets.*(ESO 기반 시크릿 주입)ingress.*(TLS/HTTPS)
pip install pyenvectorBasic init example:
import pyenvector as ev
ev.init(
address="localhost:50050",
key_path="./keys",
key_id="my_key",
)Notebook examples are in notebooks/.
This project is proprietary software. For licensing information, contact heaan.
- Product Information: heaan
- Technical Support: es2.support@heaan.com