Developer tooling
SlideOps
An infrastructure operations platform the operator owns end to end
- Role
- Creator and lead developer
- Built with
- Go
- React
- TypeScript
- PostgreSQL
- Redis
- Lua
- Docker
- Caddy
- WebSockets
- SSH
A live, multi tenant infrastructure operations platform. An operator connects their own Linux servers over SSH, then discovers, configures, deploys, secures, verifies, and monitors them from one place. SlideOps orchestrates and explains the tools that already exist rather than replacing them, and it never owns the infrastructure. The operator does.
01
Why it exists
Server operations punish the people who are least equipped to absorb the punishment. The knowledge is real but it is scattered across manual pages, forum answers, and habit, and a wrong command is expensive. SlideOps makes the intent the thing an operator expresses, and takes responsibility for turning it into the right commands for the platform actually in front of it.
02
What was built
A Go backend of roughly four hundred source files exposing one hundred and thirteen REST routes, documented as an OpenAPI specification that the frontend consumes through a generated typed client rather than hand written calls.
A React and TypeScript frontend, built with Vite and shipped as an installable progressive web app, serving three areas from one application and one session: the public site, the operator dashboard across fifty five screens, and an internal admin control plane.
A marketplace of capabilities spanning databases, runtimes, web serving, messaging, caching, storage, search, networking, orchestration, deployment, backups, and security, each pairing a provider with a capability and installed per project.
A visual capability manager that makes every capability usable without any command line knowledge: a database explorer across PostgreSQL, MySQL, MariaDB, MongoDB, and Redis, drag and drop restore from upload, and a container manager built on the workload adoption backend.
Live billing with tiered quotas on nodes, projects, and seats, team collaboration, and node transfer between operators.
03
Architecture
The whole system turns on one separation: a Capability describes what the operator wants, and never names a tool, while a Provider knows how to achieve that outcome on a specific platform. Over forty provider implementations cover nine platform families, apt through zypper, with systemd or OpenRC underneath. One Capability, many Providers, and the choice between them is made by discovery rather than by the operator.
Every Capability runs the same non skippable lifecycle: discover, assess, recommend, plan, approve, execute, verify, observe, record. Discovery only observes. Approval is required before execution. Verification always follows execution, because an execution nobody verified is not finished. Everything meaningful emits an event, and history keeps all of it.
04
My contribution
Creator and lead developer. Domain design, the Go backend, the React frontend, the provider matrix, and the production deployment.
05
Engineering challenges
Realtime without blocking the interface: the backend publishes events to Redis and a WebSocket hub fans them out to the right operator, which is what makes live plan execution, a live terminal, and live node health possible at once.
Reading a Redis keyspace safely, which needs a server side Lua scan rather than a blocking key listing.
Multi tenancy that holds under pressure. Every state changing request is authorised against the owning tenant in the application layer, and the admin control plane is the only surface permitted to read across tenants. Building the visual explorer surfaced a real scoping bug, where a service scoped browse could return unscoped data because the check ran after the query instead of before it.
Credentials that are encrypted at rest with AES-256-GCM, argon2id passwords, optional TOTP, an append only audit log naming the actual acting operator, and a rotation flow that proves the new credential works before discarding the old.
06
Result and impact
Live in production, with payments switched on. Real services run through SlideOps on its own infrastructure, which means the deploy, reverse proxy, HTTPS, and state loop is proven outside of a test rather than in one. One of those services is the Amaizing inference backend, also his.
A real production incident, a wedged certificate job alongside a wrong port mapping, was diagnosed and fixed on the box using the platform's own monitoring primitives, which is the more honest evidence that they are load bearing.
07