README

Google Analytics for fancy-flow — as four imported, versioned packages, one per runtime. Not vendored source: a copy cannot be upgraded, and third-party APIs change.

The ui package is the editor surface and is React on every host — a PHP or Python project installs it and its own runtime package, and never the js one.

What it costs you

One dependency: @particle-academy/fancy-connector-core (or particle-academy/fancy-connector-core on Composer), which the js and php packages pull in themselves. The Python package has zero runtime dependencies.

No Google Analytics SDK. Plain HTTP, deliberately: a vendor SDK is third-party code subject to the kit's full approval bar, and one per provider is hundreds of dependencies nobody is tracking.

Setting it up

Everything below is generated from provider/manifest.json, so it cannot disagree with what the packages do.

On Google Analytics's side

1. Create (or reuse) the Google Cloud project for your app's name — Every Google connector shares ONE consent grant (oauth.grant "google"), so use one project and one OAuth client for all of them: one consent and one refresh token per person. (developers.google.com)
2. Enable the Google Analytics Data API — Console: APIs & Services, Library; open the Google Analytics Data API and click Enable. (developers.google.com)
3. Configure the consent screen — Console: Google Auth platform, Branding (app name your app's name, support email, contact email), then Audience: Internal for people inside your Google Workspace organisation only, or External (while it is in testing, add each tester under Test users). Then Data Access, Add or Remove Scopes, and add https://www.googleapis.com/auth/analytics.readonly beside the scopes of any other Google connector you use. You cannot remove the consent screen once it is configured. (developers.google.com)
4. Create a Web application OAuth client — Console: Google Auth platform, Clients, Create Client; type Web application. Under Authorized redirect URIs click Add URI and paste your host's OAuth redirect URL, then Create. Copy the client ID and client secret when the client is created. (developers.google.com)
5. Connect an account and check it works — Connect one Google account through your app's name's OAuth flow, then run the check below. (developers.google.com)

Every step in full — who does it, what to paste, which credentials it yields, and how you know it worked — is in SETUP.md.

Credentials

A Google Analytics connection holds 4 values.

Two kinds of value, and mixing them up matters. A provider credential is ONE value for the whole installation — an OAuth app's client secret serves every connected account. An account credential is one per connected account. A host that stores the second where it stores the first lets one account's credentials reach another's.

Authorising

Google Analytics uses OAuth2 (authorization_code). The package DECLARES the exchange; the HOST performs it — a consent screen needs a browser, a redirect URI and somewhere to persist the result, and all three belong to the host.

• Authorize URL — https://accounts.google.com/o/oauth2/v2/auth
• Token URL — https://oauth2.googleapis.com/token
• Scopes — https://www.googleapis.com/auth/analytics.readonly
• Access token lifetime — 3600 seconds (1 hours). A host that never refreshes works all afternoon and is broken by morning.

One consent grant, shared (grant: "google"). Every connector declaring the same grant can run on ONE OAuth client, with one consent and one refresh token per person.

• Authorize with access_type=offline, include_granted_scopes=true. Ask for a feature's scopes when it is first used; the new token covers every scope granted before. A refresh token comes back only on the FIRST exchange, so keep the stored one when an exchange returns none.
• Revoke — POST https://oauth2.googleapis.com/revoke with token. It removes the WHOLE grant: every connector sharing it, not only this one. A per-feature disconnect stops using that feature's scopes locally; only a full disconnect revokes.
• Identify the account by sub: request openid, then verify the token response's id_token (issuer https://accounts.google.com or accounts.google.com, keys at https://www.googleapis.com/oauth2/v3/certs, audience = clientId). Never key a connection by email.
• No restricted scopes in this connector.

The refresh tokens do not rotate: the same one is reusable, so a refresh may safely be retried and may run concurrently. Stated rather than assumed, because the opposite — a provider that spends the token and revokes the grant on a replay — looks identical until it happens.

The estate

Google Analytics has no test estate, and somebody checked. Everything this connector does is real. Use the faker to build against it.

Google Analytics has no sandbox. A report reads a real property's real data and spends real quota. The faker returns a plausibly-shaped report so a flow can be built without a Google account.

What it can do

Actions

report_run — Google Analytics report

Run a report against a Google Analytics 4 property.

POST /v1beta/{+property}:runReport · reads only — safe to replay

Run it before you have credentials

Every operation ships a faker, whether or not Google Analytics has a sandbox. Set a node's mode to fake and it returns the shape Google Analytics actually publishes — the same field names, deterministically — so you can wire the downstream nodes before touching an account, a key, or a network.

This repository is generated

provider/ is the source. Everything under packages/ is emitted from it and must not be hand-edited — CI regenerates and diffs on every push, and the next protocol sync destroys anything it finds. See AGENTS.md.

Two namespaces, which do not match on purpose

The repo is github.com/Fancy-Friends/google-analytics; the packages publish under particle-academy. Nothing derives one from the other — the names come from weaver's friends.json and nowhere else.

Licence

MIT.