Aegis, Containerised
Content Intelligence Platform
Content analysis jobs take minutes, not milliseconds. Aegis is the architecture that follows from taking that seriously — six services, a job queue, and no request left hanging.
Six services, one job queue
Aegis splits content processing across six containerised services orchestrated with Docker Compose. A gateway accepts the request, an orchestrator manages the flow, a worker performs the long-running scraping and file generation, and a content-intelligence service handles the AI analysis. PostgreSQL stores results; Redis carries live progress and shared state; nginx fronts the whole thing on a bridged network.
The reason for the split is simple: content jobs are slow. Running them inside the request that triggered them means timeouts, no progress reporting, and a system that falls over the moment two people use it at once.
A slow job inside a web request is a design error
Scraping a set of pages and generating output from them takes minutes. Do that inside an HTTP handler and every layer between the browser and the code — proxy, load balancer, browser itself — is entitled to give up first.
Worse, there is nothing to show the user while they wait, and no way to recover if the process dies halfway through.
Accept fast, work in the background
The gateway accepts a job and returns immediately with an identifier. The orchestrator places the work on a queue; workers pick it up and report progress into Redis as they go, which is what the dashboard reads. Results land in PostgreSQL, so a completed job survives a restart of any service that produced it.
Because each concern is its own container, the slow part scales independently of the part that answers the browser.
What it does
Architecture
- Six services, Docker Compose
- nginx reverse proxy on a bridged network
- Independent scaling of the slow path
Long jobs
- Asynchronous job acceptance
- Live progress reporting
- Results survive a service restart
Operation
- Operator dashboard
- Job status and history
- Export gateway for generated files
Actual stack
Status
⚠️ A working prototype, not a deployed product
The architecture runs and the dashboard reports real job state, but this has not been operated under production load. There is no authentication layer, no horizontal-scaling story beyond "add worker containers", and no uptime or throughput figure is claimed.
The architecture is the point
Nothing here is exotic. It is the shape a system takes once you accept that the work is slower than the request that asks for it.