
Security News
Agent Baseline Publishes Security Reference Architecture for Software Factories
Agent Baseline’s new reference architecture shows how to secure software factories with four boundaries enforced outside the AI agent.
@polyxd/analytics
Advanced tools
Sends Polyxd's semantic events to your own analytics: PostHog, Segment, Google Analytics 4 or your endpoint, with an allow-list guard. No dependencies; Polyxd receives nothing
Sends Polyxd's semantic events to your own analytics. The renderers emit them (onEvent on <PolyxdSurface>, polyxd-event on <polyxd-surface>); this package turns each one into a call to PostHog, Segment, Google Analytics 4 or your own endpoint. No dependencies. Polyxd itself receives none of these events, unless you point toFetch at Studio Insights (below).
import posthog from "posthog-js";
import { toPostHog } from "@polyxd/analytics";
const onEvent = toPostHog(posthog);
<PolyxdSurface document={doc} onEvent={onEvent} />;
// posthog.capture("polyxd action.taken", { surface_id: "send", intent: "money.send", capability: "transfer.review", … })
| Export | What it does |
|---|---|
toPostHog(posthog, { groups?, eventName? }) | posthog.capture("polyxd " + type, flatten(event)), with $groups when you give groups (fixed, or a function of the event) |
toSegment(analytics, { eventName? }) | analytics.track("polyxd " + type, flatten(event)) |
toGA4(gtag, { eventName? }) | gtag("event", "polyxd_" + type with dots as underscores, flatten(event)), names and values cut to GA4's limits |
toFetch(url, { batchSize?, flushInterval?, headers?, fetch?, onError? }) | POSTs { "events": [...] } to your endpoint in batches (20 events, or 5 seconds, or when the page is hidden), with keepalive. Returns a handler with flush() and close() |
redact(event) | Rebuilds an event from the schema's allow-list: any property the schema doesn't define is dropped, at every level, and so is a reason that isn't a short code. Every adapter runs it first |
flatten(event) | One level of snake_case properties: surface_id, intent, pattern, journey, spec_version, generator, direction, experiment_<key>, actor, assistive_tech, component, component_key, component_type, capability, checkpoint, duration_ms, steps, reason, rating, polyxd_session_id |
EVENT_TYPES, ALLOWED_PROPERTIES, defaultEventName, ga4EventName | The event types and the allow-list (both checked against schema/event.schema.json in the tests), and the default names |
A client that throws never breaks the surface: the adapter catches it.
Which events there are, when each fires and what it carries: Capabilities, journeys and events.
Polyxd Studio counts these events for a workspace, if you send them there. Make an ingest key in Studio, then:
const studio = toFetch("https://studio.polyxd.com/api/w/<workspace>/events", { headers: { "x-polyxd-key": "<ingest key>" } });
Studio keeps daily counts only: never the events, the values, session ids or timestamps.
npm test -w @polyxd/analytics runs every adapter against a fake client with real events from @polyxd/core, and checks the allow-list and the redacted events against the spec's schema.
FAQs
Sends Polyxd's semantic events to your own analytics: PostHog, Segment, Google Analytics 4 or your endpoint, with an allow-list guard. No dependencies; Polyxd receives nothing
The npm package @polyxd/analytics receives a total of 751 weekly downloads. As such, @polyxd/analytics popularity was classified as not popular.
We found that @polyxd/analytics demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.

Security News
Agent Baseline’s new reference architecture shows how to secure software factories with four boundaries enforced outside the AI agent.

Security News
arXiv now limits authors to two submissions a month as AI slop overwhelms moderators, delays good papers, and sparks debate over applying the limit to everyone.

Research
/Security News
A new GhostAction wave hits hundreds of GitHub repos, expanding CI/CD secret theft to cloud and AI credentials in source code and git history.