Core concepts
The words Infrared uses, from the request you send to the decision your team makes, and how each one relates to the others.
The same few nouns appear in the API, in the dashboard and on every page of these docs. This page defines each one in a sentence or two and links to the page that explains it in full.
What you send
| Word | What it is |
|---|---|
| Moderation | One piece of content that was screened, with its result. You create one with POST /v1/moderations; its id starts with mod_. |
| Content type | What the content is: text or image. A whole form holds several fields of either. |
| Template | The configuration for one kind of content, such as a listing, a review or a message: its fields, the policies that run on it and the rules that decide. A request names one by its slug. |
| Basic check | The built-in template every project has. It runs when a request names no template, and it is all the Free plan runs. |
| Field | A named slot of a template (title, photos, price), with a type and validation rules. |
| Author | The person in your product who submitted the content. You name them with author_id, and Infrared keeps a history for each. |
How content is judged
| Word | What it is |
|---|---|
| Policy | One thing content is scored for, such as toxicity, personal_data or spam. Every policy that runs returns a score from 0 to 1. |
| Guideline | A policy of your own, written in plain language. It is scored like any other and appears as guideline/<slug>. |
| Wordlist | A policy of your own made of terms: competitor names, brands, slang. It appears as wordlist/<slug>. |
| AI instructions | Plain-language instructions on a template that are judged across all the fields of one submission. |
| Trigger | What a policy's score sets off once it crosses its threshold: flag marks the content FLAGGED, review asks for a human look. |
| Content rule | An if-then rule on a template. Rules are read top to bottom and the first one that matches decides. |
| Severity triage | What decides when no rule matches: the highest weighted score is compared with two thresholds. |
What comes back
A moderation carries two answers, and they are separate on purpose.
| Word | What it is |
|---|---|
| Verdict | flagged in the response: true is FLAGGED, false is SAFE. It says whether a policy set to flag has triggered. |
| Recommendation | recommendation in the response: allow, review or reject. It says what to do, and it comes from your content rules or from severity triage. |
| Status | How far the moderation got: screening while it runs, then completed, partial, failed or invalid. |
A listing can be FLAGGED and still allowed (a trusted author, a rule that says so), or SAFE and still sent to review. How a decision is made follows one moderation through every step.
What happens next
| Word | What it is |
|---|---|
| Review queue | Where every moderation with the recommendation review waits for a person. One queue per project. |
| Action | What a decision turns into: allow, reject, or an action you define, such as hiding a listing. Running one is logged and reported to your backend. |
| Webhook | A signed HTTP request Infrared sends to your endpoint when something happens: a moderation completes, a queue item is decided, an author is blocked. |
| Dry run | A mode in which everything is scored and queued and nothing is meant to be enforced, for trying Infrared on live traffic. |
| Simulation | A replay of the last seven days of decisions against rules you have not published yet. |
Who owns what
| Word | What it is |
|---|---|
| Organisation | The account. It holds the plan, the monthly checks and the members. The API spells it organization. |
| Project | One product or environment inside an organisation. Templates, guidelines, wordlists, API keys, webhook endpoints, authors and the review queue all belong to a project. |
| Member | A person in the organisation, with a role that says what they may do. |
| API key | What your backend authenticates with. A key belongs to one project and is either live or test. |
| Check | The unit a plan is measured in. A moderation uses at least one. |
Next
- How Infrared worksOne request goes through admission, validation, screening and a decision. What each stage does, what uses a check and what happens after the answer.
- The moderation objectThe verdict, the scores, the evidence in details and your own context. Every field of a moderation, and when each one is present.
- TemplatesA template is the configuration for one kind of content. It holds the fields, the policies and the rules, and requests name it by its slug.
Something missing or unclear on this page? Write to support@useinfrared.com.