1. Executive summary

Solynth Network already exists as a functioning cloud-based collection of applications, publishing workflows, research practices and technical proofs of concept. The next step is not to discard that work, but to place a controlled, enterprise-capable infrastructure layer beneath it for customers who require on-premises operation, stronger data control, local AI processing, or contractual isolation.

Document purpose

This is deliberately a strawman rather than the final engineering manual. It is intended to let founders, advisers, customers and prospective investors review the concept, understand the deployment sequence, challenge assumptions and approve the next level of detailed design.

2. What exists today

The present Solynth implementation is predominantly cloud-based. It has been used to demonstrate practical capability rather than merely describe a future product.

3. Solynth ecosystem scope

The Solynth Network is intended to be an enabling ecosystem rather than a single application. The diagrams below show the breadth of the present technical vocabulary and one possible specialist application domain. They are included to communicate scope, not to imply that every named component has already been implemented or contracted.

Current capability and target integration landscape

Future concept — games and interactive media

4. Customer 1: first on-premises implementation

5. The Solynth on-premises reference system

The reference configuration is a repeatable baseline, not a single mandatory bill of materials. Dell servers, AMD EPYC processors, enterprise ECC memory, NVIDIA professional GPUs, redundant storage and Proxmox form the current reference because they offer credible support, broad availability and a practical balance between open software and enterprise hardware.

6. Reference Operations Centre rollout

A ROC is a remotely operated Solynth infrastructure and support hub. It is not necessarily the sole production centre, and customer sites are not designed to fail merely because a ROC is unavailable.

7. WAN and customer connectivity

Default: encrypted public Internet

The default inter-site connection should use business-grade Internet circuits with WireGuard or IPsec VPN tunnels. This approach is inexpensive, portable, widely supportable and adequate for monitoring, software distribution, administrative access and controlled data exchange.

Private connectivity only when required

MPLS, carrier Ethernet, private wavelength, cloud direct-connect services or customer-provided private circuits should be treated as optional contract requirements. Solynth should accommodate them, but should not absorb their cost unless the customer’s commercial agreement explicitly includes them.

8. Installation timing and knowledge transfer

Assumption: all equipment has arrived at the site, rack space and power are ready, firmware and licence entitlements are available, WAN service is live, and no building or electrical work is required.

A safe project allowance for the first Customer 1 phase-one build is two working weeks elapsed: approximately ten person-days of technical effort plus room for customer change windows, access delays and defect correction. After automation and templating, a standard phase-one system should be installable in three to five elapsed working days, with a separate soak and acceptance period.

Training estimate

9. Dark-room operations and staffing

Solynth systems are designed for lights-out operation. No permanent operator is expected at the customer site or ROC. Monitoring, alerting, patch management, capacity checks, backup verification and most corrective actions are performed remotely.

Initial staffing model

One experienced platform engineer, supported by good automation and vendor hardware contracts, can initially operate several installations. The limiting factor will not be raw server count; it will be customer-specific integrations, support response commitments and change-management load. The first commercial target should therefore be service quality, not an unsupported claim that a tiny team can operate one hundred customers.

10. Customer deployment and commercial model

Future hardware supply

Customer 1 purchased or controlled its hardware independently. Future contracts may offer hardware through Solynth or a nominated reseller, creating a modest equipment margin while giving the customer one accountable implementation route. Hardware margin should remain transparent and secondary to the higher-value architecture, integration, support and managed-service work.

11. Data consistency and storage model

PostgreSQL is the canonical system of record for identity, metadata, workflow, permissions, provenance, audit, configuration, searchable text and vector indexes. Large durable artefacts belong in object storage, while local filesystems are used only for temporary staging and caches.

Appendix D — detailed phase-one implementation estimate

Appendix E — recommended additional diagrams for the full document

The following SVG-style figures would add value in later revisions without turning this strawman into a 200-page manual:

1. Security trust zones: Internet edge, management, customer application, database, AI/GPU, backup and out-of-band management networks.
2. Release promotion path: development → test → customer approval → signed package → staged rollout → rollback.
3. Monitoring and incident flow: local agent → ROC dashboard → alert triage → remote remediation → hands-and-feet → vendor escalation.
4. Identity and access model: Solynth operators, customer administrators, applications, service accounts and break-glass access.
5. Backup and recovery map: local snapshots, PBS, object replication, off-site copy and tape, including recovery-time objectives.
6. Customer deployment decision tree: Solynth-hosted, on-premises or hybrid based on residency, latency, integration, budget and control requirements.

Appendix F — decisions still open

13. Recommended next document stage

After review, this strawman can be split into three controlled documents rather than expanded indiscriminately: