Configuration Reference
VectorScaleDB is configured entirely via environment variables. All variables use the VSDB_ prefix.
Core Settings
| Variable |
Type |
Default |
Description |
VSDB_HOST |
string |
0.0.0.0 |
Listen address |
VSDB_PORT |
integer |
8080 |
Listen port |
VSDB_STORAGE_PATH |
string |
data/vsdb |
Storage path for embedded storage engine data |
VSDB_LOG_LEVEL |
string |
info |
Logging level (trace, debug, info, warn, error) |
VSDB_MAX_VECTOR_DIM |
integer |
4096 |
Maximum vector dimensionality |
VSDB_MAX_BATCH_SIZE |
integer |
10000 |
Maximum vectors per batch ingest |
VSDB_COMPRESSION_ENABLED |
boolean |
true |
Enable vector-aware compression |
VSDB_METRICS_ENABLED |
boolean |
true |
Enable Prometheus metrics endpoint |
Authentication & Security
| Variable |
Type |
Default |
Description |
VSDB_AUTH_ENABLED |
boolean |
false |
Enable API key authentication |
VSDB_AUTH_SECRET |
string |
-- |
JWT signing secret (required if auth enabled) |
VSDB_TLS_CERT |
string |
-- |
Path to TLS certificate file |
VSDB_TLS_KEY |
string |
-- |
Path to TLS private key file |
VSDB_RATE_LIMIT_RPS |
integer |
1000 |
Requests per second rate limit |
VSDB_AUDIT_ENABLED |
boolean |
false |
Enable SHA-256 audit chain |
Enterprise Features
| Variable |
Type |
Default |
Description |
VSDB_MULTI_TENANT |
boolean |
false |
Enable multi-tenant isolation |
VSDB_RBAC_ENABLED |
boolean |
false |
Enable role-based access control |
VSDB_GDPR_MODE |
boolean |
false |
Enable GDPR deletion compliance |
VSDB_SLA_MONITORING |
boolean |
false |
Enable SLA monitoring |
VSDB_METERING_ENABLED |
boolean |
false |
Enable usage metering |
VSDB_SHARDING_ENABLED |
boolean |
false |
Enable consistent-hash sharding |
Encryption at Rest
| Variable |
Type |
Default |
Description |
VSDB_ENCRYPTION_ENABLED |
boolean |
false |
Enable field-level encryption |
VSDB_ENCRYPTION_KEY_PROVIDER |
string |
local |
Key provider (local, env, aws-kms, azure-kv, hashicorp-vault) |
VSDB_ENCRYPTION_KEY_PATH |
string |
-- |
Path to encryption key file (for local provider) |
VSDB_ENCRYPTION_KEY_ENV |
string |
-- |
Environment variable containing key (for env provider) |
GraphQL
| Variable |
Type |
Default |
Description |
VSDB_GRAPHQL_ENABLED |
boolean |
false |
Enable GraphQL API |
VSDB_GRAPHQL_COMPLEXITY_LIMIT |
integer |
100 |
Maximum query complexity |
VSDB_GRAPHQL_DEPTH_LIMIT |
integer |
10 |
Maximum query depth |
Virtual Filesystem (VFS)
| Variable |
Type |
Default |
Description |
VSDB_VFS_ENABLED |
boolean |
false |
Enable virtual filesystem mount (ProjFS on Windows, FUSE on Linux/macOS) |
VSDB_VFS_MOUNT_PATH |
string |
-- |
Mount point for the virtual filesystem |
Coupling Matrix & Self-Awareness
| Variable |
Type |
Default |
Description |
VSDB_COUPLING_ENABLED |
boolean |
true |
Enable the coupling matrix for cross-entity behavioral prediction |
VSDB_DREAM_ENABLED |
boolean |
true |
Enable the dream daemon for offline consolidation and coupling discovery |
VSDB_CONVERGENCE_TRACKING |
boolean |
true |
Enable convergence tracking and Weissman compression scoring |
Webhooks & Streaming
| Variable |
Type |
Default |
Description |
VSDB_WEBHOOKS_ENABLED |
boolean |
false |
Enable webhook event notifications |
VSDB_STREAM_BUFFER_SIZE |
integer |
1000 |
Change stream buffer size |
Compile-Time Feature Flags
VectorScaleDB ships as a universal binary with all software features compiled in. The flags below are only relevant if building from source. Hardware-specific flags (gpu, pn-chip) enable optional acceleration that loads at runtime via dynamic modules.
| Flag |
Description |
enterprise |
Rate limiting, RBAC, SLA monitoring, metering |
swagger |
OpenAPI UI at /swagger-ui |
cluster |
Distributed mode: coordinator, federation, router, rebalancer |
grpc |
gRPC API (requires protoc) |
graphql |
GraphQL endpoint at /graphql |
p2p |
P2P peer-to-peer transport (TCP + Noise + Yamux) |
ipfs |
Content-addressed segment archival and deduplication |
federation |
Cross-cluster behavioral template exchange |
topology |
Topological anomaly detection (structural pattern analysis) |
inventory |
Container/inventory management, audit log, reconstruction |
multi-rbac |
Granular per-entity/per-field RBAC with delegation |
portals |
Cross-tenancy portals, world management |
bci |
BCI cognitive layer, consent gate, skills, immersive reality |
trust |
Trust network: threshold signing, trust scoring, arbitration |
neuroscience |
Neuroscience-inspired pattern separation and operating modes |
unified-storage |
Vectors + documents + graphs + blobs + full-text search |
adaptive-deploy |
Adaptive deployment auto-configuration across the microcontroller-to-datacenter continuum |
self-repair |
Automatic integrity verification and corruption recovery |
cdn |
Semantic content-addressed delivery network |
dev-intel |
Developer intelligence: code symbols, builds, test results |
spectral |
Four-frequency data lifecycle management |
hierarchy |
Dynamic hierarchy overlay with hub topology |
Adaptive Deployment
VectorScaleDB is a single adaptive substrate that scales across a continuum from microcontroller-class hardware up to datacenter clusters — the same binary auto-tunes compression, indexing, and resource allocation to the host it runs on. Set a target profile via VSDB_DEPLOYMENT_PROFILE to bias the auto-tuning toward a class of hardware, or leave it unset to let the substrate detect and adapt at startup.
| Hardware class |
Typical RAM |
Use Case |
| Microcontroller / embedded |
tens of MB |
Microcontrollers, embedded IoT, wearables |
| Edge |
hundreds of MB |
Edge gateways, single-board computers |
| Workstation |
GBs |
Workstations, development, small production |
| Server |
tens of GBs |
Enterprise servers, multi-tenant production |
| Cluster / datacenter |
large |
High-throughput clusters, fleet-scale deployments |
Independently of hardware class, every deployment runs in one of three network modes: fully-federated (joins a coupling-aware cluster and exchanges behavioral templates across nodes), lightly-coupled (operates standalone but can opportunistically sync in island mode), or air-gapped (fully isolated, with attested offline bundle exchange when needed).
Cluster Settings
| Variable |
Type |
Default |
Description |
VSDB_CLUSTER_ENABLED |
boolean |
false |
Enable distributed cluster mode |
VSDB_NODE_ID |
string |
-- |
Unique node identifier |
VSDB_CLUSTER_SEEDS |
string |
-- |
Comma-separated seed node addresses |
VSDB_GOSSIP_PORT |
integer |
8081 |
Port for gossip-based membership protocol |
VSDB_P2P_ENABLED |
boolean |
false |
Enable P2P transport |
VSDB_CAS_API_URL |
string |
-- |
Content-addressed storage API URL for segment archival |
VSDB_FEDERATION_ENABLED |
boolean |
false |
Enable cross-cluster template federation |
Running VectorScaleDB
VectorScaleDB ships as a single self-contained binary. Production deployments run the native binary directly — no container runtime, external services, or dependencies are required. Download the binary, set the environment variables above, and run it:
# Configure via environment variables, then run the binary
export VSDB_AUTH_ENABLED=true
export VSDB_AUTH_SECRET="$(cat ./auth.secret)"
export VSDB_ENCRYPTION_ENABLED=true
export VSDB_STORAGE_PATH=/data/vsdb
./vsdb
Binary downloads at launch
Signed release binaries for Linux, macOS, and Windows are available at launch. Production runs the native binary — containers are an optional convenience for local development and cross-platform integration testing, not a production requirement.
Docker Compose (Dev Convenience)
For local development and integration testing, a Docker Compose configuration with authentication and encryption enabled is available at launch. Production deployments should run the native binary above rather than a container.
version: "3.8"
services:
vectorscaledb:
image: vectorscaledb/vectorscaledb:latest
ports:
- "8080:8080"
environment:
VSDB_STORAGE_PATH: /data/vsdb
VSDB_COMPRESSION_ENABLED: "true"
VSDB_AUTH_ENABLED: "true"
VSDB_AUTH_SECRET: "${VSDB_AUTH_SECRET}"
VSDB_ENCRYPTION_ENABLED: "true"
VSDB_ENCRYPTION_KEY_PROVIDER: env
VSDB_ENCRYPTION_KEY_ENV: VSDB_MASTER_KEY
volumes:
- vsdb-data:/data/vsdb
volumes:
vsdb-data:
Tip
Store secrets like VSDB_AUTH_SECRET and VSDB_MASTER_KEY in a .env file or a secrets manager -- never commit them to version control.