From 17512ab3966313d95605da540ea68043f0a85d6c Mon Sep 17 00:00:00 2001
From: Derek <derek@hyperi.io>
Date: Mon, 2 Nov 2026 01:05:32 +1100
Subject: [PATCH] fix(ci): run the full test tier nightly and before every
 release

Releases require a full-tier run (`test.full.required_for_release: true`). CI runs the full tier nightly at 16:19 UTC. The CI caller forwards `test-tier`, so manual runs can select full.

tests/TESTING.md says what each tier runs and which tests use the shared ClickHouse and Kafka, the PostgreSQL and MySQL containers, or no container. Core and full run the same tests because `addopts` deselects nothing.
---
 .github/workflows/ci.yml |  9 +++++++++
 .hyperi-ci.yaml          |  3 +++
 tests/TESTING.md         | 14 ++++++++++++++
 3 files changed, 26 insertions(+)
 create mode 100644 tests/TESTING.md

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 9fd7c17..61c2292 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -8,6 +8,8 @@
 name: CI
 
 on:
+  schedule:
+    - cron: '19 16 * * *'
   push:
     branches: ["**"]
   pull_request:
@@ -29,6 +31,12 @@ on:
         required: false
         default: "auto"
         description: "Version resolution for from-head: auto | patch | minor | X.Y.Z"
+      test-tier:
+        type: choice
+        required: false
+        default: "core"
+        options: [core, full]
+        description: "core | full. full runs the full test tier on this run."
 
 permissions:
   contents: write
@@ -44,4 +52,5 @@ jobs:
       tag: ${{ inputs.tag || '' }}
       from-head: ${{ inputs.from-head || '' }}
       bump: ${{ inputs.bump || 'auto' }}
+      test-tier: ${{ inputs.test-tier || 'core' }}
     secrets: inherit
diff --git a/.hyperi-ci.yaml b/.hyperi-ci.yaml
index 14952b4..f56523c 100644
--- a/.hyperi-ci.yaml
+++ b/.hyperi-ci.yaml
@@ -11,6 +11,9 @@ test:
   min_coverage: 75  # currently 80%; ratchet up as coverage improves
   python:
     parallel: auto
+  # The full tier (tests/TESTING.md) runs nightly and before every release.
+  full:
+    required_for_release: true
 
 quality:
   # examples/ are illustrative snippets, not scanned code
diff --git a/tests/TESTING.md b/tests/TESTING.md
new file mode 100644
index 0000000..db1fd72
--- /dev/null
+++ b/tests/TESTING.md
@@ -0,0 +1,14 @@
+# Testing
+
+The general rules are hyperi-ai's testing standard (`standards/universal/testing.md` in hyperi-ai). This page covers what logreducer does on top of them.
+
+| Level | Runs | Command |
+|---|---|---|
+| core | locally, every push and PR | `hyperi-ci run test`, or `uv run pytest` |
+| full | nightly (`schedule` in `.github/workflows/ci.yml`) and before every release (`test.full.required_for_release: true`) | `hyperi-ci run test --tier full` |
+
+`addopts` deselects nothing, so core and full run the same tests today. The full tier is what a release waits on. A test that grows slow enough to move out of core gets a marker and an `addopts -m` deselect, and full still runs it.
+
+ClickHouse and Kafka are the run's shared services from scalo-py's test plugin, loaded in `tests/conftest.py`. That's one container per service per run, shared by every xdist worker, with a database or topics of its own for each test. The images, the naming and the clean-up are scalo-py's, in its `tests/TESTING.md`. `HYPERI_TEST_*` settings point a local run at an existing ClickHouse or Kafka instead.
+
+PostgreSQL (`postgres:16.6`) and MySQL (`mysql:8.4.5`) are throwaway testcontainers in `tests/integration/conftest.py`, because the plugin doesn't provide them. Each starts once per session and stops when it ends. `LOGREDUCER_KEEP_CONTAINERS=1` leaves them running, and without Docker those tests skip. Unit tests (`tests/unit/`) use no container.