Trazio by CerebrixOS

← All guides

The guided setup

Open trazio.cerebrixos.com/console, sign in, and choose Set up your first pipeline. The setup asks what you want to watch, then shows only the steps and options that matter for it. Every step is checked before you can continue, and nothing is created until the last step.

Who can do what

Your role is shown as a badge at the top right of every page, and the What can I do here? panel explains it. The person who creates a workspace is its first admin. Members see a short page with the connector address and the pipelines they can read.

The steps

You pick a goal first. Your goal decides the steps.

A half-finished setup is saved for a week, so you can leave and come back. Start over on the review page clears it.

Data in

Choose which sources the pipeline accepts and what happens to events from anything else. You can only narrow the sources for your use case, never widen them. Optionally pin a protocol (such as otlp.logs) to a source, so everything arriving on it counts as that verified source. For robots you can also limit which recorded topics are indexed and how much of a recording is decoded.

Camera and Analysis (video and live)

Camera sets the video channel and how recorded video is processed. Analysis sets what a model does with live frames: frames per second (0.2 to 5), minimum confidence (0 to 1) and the longest acceptable delay (250 to 30,000 ms). The model recognises generic objects such as people and vehicles. It does not recognise robots or intent, and live delay isn't guaranteed. These options need approval, explained below.

The transport is fixed when you create a pipeline and can't be changed afterwards, so a live pipeline is created as a live pipeline from the start.

Rules

Pick ready-made rules, switch off individual rules, and add your own with the rule builder: choose where it looks, a field, a condition and a severity. Fields are limited to what each source allows, so a rule that could never match is refused with the reason. Trazio writes a matching and a non-matching test event for each rule and runs them through the platform's own validation before you save, so you see "matches" and "no match" before anything goes live. Custom rules are saved as versioned rule packs in your own namespace and are never overwritten.

Alerts

Add up to five destinations: Slack, Teams, Discord, PagerDuty, Opsgenie, Jira, ServiceNow, email, SMS or a webhook. Web addresses must be public https addresses: plain http, internal names, local addresses and IP addresses are refused. Keys for PagerDuty, Opsgenie and similar services are stored encrypted and never shown again. Set how serious something must be, and how long to wait before repeating the same alert (60 seconds to a day).

Advanced

Link related events (and choose what they share), open incidents from a severity, and set how long receipts and quarantined events are kept (1 to 7 days on the trial). AI summaries, search by meaning and analytics export are here too and need approval.

Review and Connect

The review page lists every choice, whether it applies now or waits for approval, and the result of a full check of the settings. If something is wrong it names the setting and links to the step. After you create the pipeline you see the write-only collector token once, then your next steps: send a first event, connect an assistant, invite your team.

Options that need approval

Some options cost money or need infrastructure that only the Trazio team can set up: the live video channel, live analysis and its settings, recorded-video processing, AI summaries, search by meaning and analytics export. Choosing one doesn't block you:

• Your pipeline is created now and everything else works.
• Each costly option becomes a request, shown on your pipeline page as Waiting for approval.
• Staff approve or decline each request. An approved setting is applied if it can be applied; if the infrastructure isn't ready it shows Approved, being set up until it is. Declined requests show the reason.
• Turning an option back off withdraws a waiting request.

Changing things later

Open a pipeline and choose a settings page. They are the same screens as the setup, saved straight to the pipeline. If someone else changed the pipeline since you opened it, you're asked to reload instead of overwriting their change. Pause stops a pipeline accepting data and keeps everything. Delete removes it for good.

From code

Everything here is also available from the SDK and the API: see Create pipelines from the console, code or an assistant.

Common messages