docs: remove GA4 Data API data source

Reflects codelibs/recotem#109, which removed the `source.type: ga4`
data source. Drops the GA4 source page, recipe `source.type: ga4`
section, `recotem[ga4]` install extra, `RECOTEM_GA4_MAX_PAGES` env var,
and the GA4 sidebar entry, in both English and Japanese v2 docs.

BigQuery-based GA4 usage (`events_*` export via `type: bigquery`) is
preserved.

Also repoints a pre-existing dead link in environment-variables.md
(`./data-sources/` -> `./recipe-reference#source`) so the build passes.

Author: marevol

.vitepress/config.ts

@@ -119,7 +119,6 @@ function v2DocsSidebar(lang: 'en' | 'ja'): DefaultTheme.SidebarItem[] {
         { text: 'CSV / Parquet', link: `${prefix}/data-sources/csv` },
         { text: 'BigQuery', link: `${prefix}/data-sources/bigquery` },
         { text: 'SQL', link: `${prefix}/data-sources/sql` },
-        { text: 'GA4', link: `${prefix}/data-sources/ga4` },
         { text: lang === 'ja' ? 'プラグイン' : 'Plugins', link: `${prefix}/data-sources/plugins` },
       ],
     },

docs/data-sources/ga4.md

@@ -89,14 +89,13 @@ These variables configure storage paths, locking, metadata field filtering, and
 
 ## Data source
 
-These variables tune behaviour of specific data sources. They are read only by `recotem train` and only when the corresponding source is used. See the [Data sources](./data-sources/) reference for full context.
+These variables tune behaviour of specific data sources. They are read only by `recotem train` and only when the corresponding source is used. See the [Data sources](./recipe-reference#source) reference for full context.
 
 | Variable | Default | Scope | Clamping | Description |
 |---|---|---|---|---|
 | `RECOTEM_BQ_REQUIRE_STORAGE_API` | (unset) | train | — | Truthy values: `1`, `true`, `yes`, `on`. When set, the BigQuery source raises `DataSourceError` (exit 3) instead of silently falling back to the slower REST API when the BigQuery Storage Read API fails (e.g. missing `bigquery.readSessions.create` IAM permission). Use this to surface IAM gaps rather than accepting degraded throughput. |
 | `RECOTEM_MAX_SQL_ROWS` | `50_000_000` | train | [1_000, 500_000_000] | Hard cap on the number of rows returned by the SQL data source. Exceeding the cap raises `DataSourceError` (exit 3). Caps **row count**, not DataFrame resident memory — see [SQL source — memory bound caveat](./data-sources/sql#memory-bound-caveat). |
 | `RECOTEM_SQL_ALLOW_PRIVATE` | (unset) | train | — | Truthy values: `1`, `true`, `yes`, `on`. Opts the SQL source into accepting private/loopback DSN hosts (default deny, for SSRF). Covers every driver-routing form — netloc, `?host=`, `?hostaddr=`, `?service=`, `?unix_socket=`, absolute-path host, and network DSNs with no host info — all default-deny without this flag. Also disables the DNS-rebinding re-check before each probe/fetch — opting in means trusting the host end-to-end. |
-| `RECOTEM_GA4_MAX_PAGES` | `500` | train | [1, 10_000] | Hard ceiling on GA4 Data API pagination loops. Reached when a property is too large for the default; raise after confirming quota. |
 
 ## Recipe expansion
 

docs/environment-variables.md

@@ -47,7 +47,7 @@ A recipe is the single source of truth for a model:
 ```
 
 The recipe captures:
-- **Where to get data** (`source` block — CSV, Parquet, BigQuery, SQL, GA4, or plugin)
+- **Where to get data** (`source` block — CSV, Parquet, BigQuery, SQL, or plugin)
 - **How to map columns** (`schema` block — user ID, item ID, optional timestamp)
 - **Data quality gates** (`cleansing` block — null-drop, dedup, minimum thresholds)
 - **What to train** (`training` block — algorithms, Optuna budget, split scheme)
@@ -152,7 +152,6 @@ This separation means:
 - [CSV / Parquet Source](./data-sources/csv) — local, object-storage, and HTTP source options
 - [BigQuery Source](./data-sources/bigquery) — authentication, parameter binding, GA4 patterns
 - [SQL Source](./data-sources/sql) — PostgreSQL / MySQL / MariaDB / SQLite via SQLAlchemy 2
-- [GA4 Source](./data-sources/ga4) — Google Analytics 4 Data API, skipping the BigQuery Export hop
 - [Plugin Data Sources](./data-sources/plugins) — extend `source.type` with custom plugins
 - Deployment guides — Docker, Kubernetes, cron scheduling
 - Operations — key rotation, recovery, sizing, troubleshooting

docs/index.md

@@ -11,7 +11,7 @@ A recipe is a YAML file that defines what data to fetch, how to train, and where
 | Field | Type | Required | Description |
 |-------|------|----------|-------------|
 | `name` | string | yes | Endpoint name. Pattern: `^[A-Za-z0-9_-]{1,64}$`. Used in endpoint paths such as `/v1/recipes/{name}:recommend`. |
-| `source` | object | yes | Data source config. `type` field is the discriminator (`csv`, `parquet`, `bigquery`, `sql`, `ga4`, or any plugin). Validated in two stages: the rest of the recipe is parsed first, then the source dict is dispatched to the plugin's `Config` class. As a result, errors in `source.*` surface *after* errors elsewhere in the recipe; an unknown `source.type` raises a `DataSourceError` listing all registered type names. |
+| `source` | object | yes | Data source config. `type` field is the discriminator (`csv`, `parquet`, `bigquery`, `sql`, or any plugin). Validated in two stages: the rest of the recipe is parsed first, then the source dict is dispatched to the plugin's `Config` class. As a result, errors in `source.*` surface *after* errors elsewhere in the recipe; an unknown `source.type` raises a `DataSourceError` listing all registered type names. |
 | `schema` | object | yes | Column mapping. |
 | `cleansing` | object | no | Data quality gates. |
 | `item_metadata` | object | no | Metadata joined into predict responses. |
@@ -100,37 +100,6 @@ source:
 
 Install one extra: `pip install "recotem[postgres]"`, `recotem[mysql]`, or `recotem[sqlite]`. Full reference: [SQL source](./data-sources/sql).
 
-### `source.type: ga4`
-
-```yaml
-source:
-  type: ga4
-  property_id: "123456789"
-  user_dimension: userPseudoId
-  item_dimension: itemId
-  time_dimension: date
-  event_names: [purchase, view_item, add_to_cart]
-  lookback_days: 90               # XOR with start_date + end_date
-  max_rows: 1_000_000
-  weight_column: event_count
-  api_timeout_seconds: 60
-```
-
-| Field | Type | Default | Notes |
-|-------|------|---------|-------|
-| `property_id` | string | required | Numeric only (`^\d+$`). Not the `G-XXXX` measurement ID. |
-| `user_dimension` | string | required | `userId` or `userPseudoId`. |
-| `item_dimension` | string | `itemId` | Any GA4 item-scoped dimension. |
-| `time_dimension` | string | `date` | `date` / `dateHour` / `dateHourMinute`. |
-| `event_names` | list[string] | required | 1–50 names; each matches `^[A-Za-z_][A-Za-z0-9_]{0,39}$`. |
-| `lookback_days` | int | XOR | 1–3650. Rolling window ending yesterday. |
-| `start_date` / `end_date` | string (ISO) | XOR | Both required if either is set. |
-| `max_rows` | int | required | Valid range `[1, 50_000_000]`. |
-| `weight_column` | string | `event_count` | Must not collide with the dimension keys or the literal `eventName`. |
-| `api_timeout_seconds` | int | `60` | Valid range `[5, 600]`. |
-
-Install the extra: `pip install "recotem[ga4]"`. Full reference: [GA4 source](./data-sources/ga4).
-
 ---
 
 ## `schema`

docs/recipe-reference.md

@@ -31,7 +31,6 @@ The core package ships with CSV and Parquet data sources. Install extras for add
 | PostgreSQL data source | `pip install "recotem[postgres]"` | Read interaction data from PostgreSQL via psycopg |
 | MySQL / MariaDB data source | `pip install "recotem[mysql]"` | Read interaction data from MySQL or MariaDB via PyMySQL |
 | SQLite data source | `pip install "recotem[sqlite]"` | Read interaction data from SQLite (uses stdlib `sqlite3`) |
-| Google Analytics 4 data source | `pip install "recotem[ga4]"` | Read interaction events from GA4 via the Data API |
 | Amazon S3 | `pip install "recotem[s3]"` | Read/write artifacts and data from S3 |
 | Google Cloud Storage | `pip install "recotem[gcs]"` | Read/write artifacts and data from GCS |
 | Azure Blob Storage | `pip install "recotem[azure]"` | Read/write artifacts and data from Azure |