Github: https://github.com/WinmemWitness/Winmem-Witness/tree/Winmem-v1.2.2

Documentation: https://github.com/WinmemWitness/Winmem-Witness/blob/Winmem-v1.2.2/DEPLOYMENT.md

1. Overview

Winmem is designed to run as long-lived infrastructure. Deployments are expected
to remain operational even when the original development team is no longer
active.

Supported environments:

• Local development

• Self-hosted servers

• Kubernetes clusters

• Helm-managed environments

2. Quickstart: Docker Compose

Requirements

• Docker >= 24

• Docker Compose v2

• Node.js 20+ (optional, for CLI usage)

Steps

cp .env.example .env docker compose -f infra/docker/docker-compose.yml up -d

Services

• API: http://localhost:8080

• Web UI: http://localhost:3000

• Grafana: http://localhost:3001

• Prometheus: http://localhost:9090

Docker Compose is recommended for local testing and small-scale deployments.

3. Self-Hosting

Winmem is fully self-hostable and does not require external SaaS services.

Core Components

• API service

• Worker service

• Web UI

• PostgreSQL

• Redis

Optional components:

• ClickHouse (high-throughput event storage)

• Grafana / Prometheus / Loki / Tempo

Configuration

All configuration is driven by environment variables and YAML configuration files.
Refer to .env.example and winmem.yaml for the full configuration surface.

4. Kubernetes Deployment

Winmem supports Kubernetes via raw manifests and Kustomize overlays.

Structure

• infra/k8s/base: base manifests

• infra/k8s/overlays: environment-specific overlays (dev, prod)

Deploy

kubectl apply -k infra/k8s/overlays/prod

Features

• Horizontal Pod Autoscaling

• ConfigMaps and Secrets

• Ingress support

• Stateless API and Worker services

5. Helm Deployment

Winmem provides an official Helm chart.

Install

helm install winmem infra/helm/winmem -f values.prod.yaml

Configuration

All services can be configured via values.yaml. Secrets should be injected
through external secret managers or Kubernetes Secrets.

6. Solana RPC Recommendations

Winmem is RPC-intensive by design.

Best Practices

• Use multiple RPC providers

• Enable RPC pooling and rotation

• Avoid free public RPC endpoints in production

Supported Providers

• Helius

• Triton

• QuickNode

• Self-hosted Solana RPC nodes

7. Scaling Strategy

API Service

• Stateless

• Horizontally scalable

Worker Service

• Queue-based processing

• Scale by increasing replica count

Storage

• PostgreSQL: vertical scaling + read replicas

• Redis: optional cluster mode

8. Backups and Recovery

PostgreSQL

• Daily logical backups

• Point-in-time recovery recommended

Object Storage

• Snapshot exports should be versioned and immutable

Restore

Database restoration can be performed using the provided scripts:

• scripts/backup-db.sh

• scripts/restore-db.sh

9. Operational Notes

• Always pin Docker image versions in production

• Monitor RPC error rates and latency

• Treat archive snapshots as immutable artifacts

• Store audit proofs separately from live databases

10. Deployment Philosophy

Winmem is designed to remain operational even when unattended.

A correct deployment:

• Requires minimal human intervention

• Produces verifiable outputs

• Can be audited long after active development ends

This is not a hosting guide for short-lived demos.
It is infrastructure for long-term memory preservation.