.gitea/workflows/test.yml

@@ -62,7 +62,8 @@ jobs:
 
           docker compose -f "$COMPOSE_FILE" exec -T backend sh -lc "
             pip install -q -r requirements-dev.txt &&
-            python -m pytest tests -ra -vv --tb=short
+            python -m pytest tests -ra -vv --tb=short &&
+            python run_seeds.py --only seed_001_cleanup_pytest_artifacts --force
           "
           echo "✓ pytest OK"
 

README.md

@@ -103,7 +103,10 @@ Interaktive API-Doku (Dev): http://localhost:8097/api/docs
 
 ### Auth (AP0.2)
 
-Bootstrap (nur wenn noch kein User existiert — `.env` setzen):
+**Ersteinrichtung** (nur wenn noch kein User existiert):
+
+- UI-Registrierung unter http://localhost:3097 (Tab „Registrieren“), oder
+- Bootstrap per `.env` (wird als Data-Seed beim Start ausgeführt):
 
 ```env
 KAIRO_BOOTSTRAP_ADMIN_EMAIL=admin@kairo.local
@@ -111,6 +114,8 @@ KAIRO_BOOTSTRAP_ADMIN_PASSWORD=…
 KAIRO_BOOTSTRAP_TENANT_SLUG=default
 ```
 
+Migrationen & idempotente Data-Seeds: [docs/MIGRATIONS.md](docs/MIGRATIONS.md)
+
 | Endpoint | Methode | Auth | Beschreibung |
 |----------|---------|------|--------------|
 | `/api/auth/setup-status` | GET | — | Ob Erstregistrierung offen ist |

backend/main.py

@@ -21,9 +21,15 @@ else:
         print(f"[FAIL] Migrationen fehlgeschlagen (Exit {exit_code})")
         sys.exit(exit_code)
 
-    import bootstrap
+if os.getenv("SKIP_SEEDS", "").strip().lower() not in ("1", "true", "yes"):
+    import run_seeds
 
-    bootstrap.bootstrap_admin_if_needed()
+    exit_code = run_seeds.main()
+    if exit_code != 0:
+        print(f"[FAIL] Seeds fehlgeschlagen (Exit {exit_code})")
+        sys.exit(exit_code)
+else:
+    print("[SKIP_SEEDS] Data-Seeds übersprungen")
 
 allowed_origins = [
     origin.strip()

backend/migrations/003_data_seeds_tracking.sql

@@ -0,0 +1,8 @@
+-- Tracking für idempotente Data-Seeds (Checksum-basiert, bei Änderung erneut ausführbar)
+
+CREATE TABLE IF NOT EXISTS data_seeds (
+    seed_name VARCHAR(255) PRIMARY KEY,
+    checksum VARCHAR(64) NOT NULL,
+    executed_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    last_run_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);

backend/run_seeds.py

@@ -0,0 +1,273 @@
+#!/usr/bin/env python3
+"""Apply idempotent data seeds with checksum-based tracking and re-run on change."""
+
+from __future__ import annotations
+
+import hashlib
+import importlib.util
+import os
+import re
+import sys
+import time
+from typing import Callable, List, Optional, Tuple
+
+import psycopg2
+import sqlparse
+
+from db import db_params, get_connection
+
+_SEED_PREFIX = re.compile(r"^seed_(\d+)_(.+)$")
+_DEV_MARKER = ".dev."
+
+
+def is_production() -> bool:
+    return os.getenv("ENVIRONMENT", "development").strip().lower() == "production"
+
+
+def seeds_directory() -> str:
+    docker_path = "/app/seeds"
+    if os.path.isdir(docker_path):
+        return docker_path
+    return os.path.join(os.path.dirname(os.path.abspath(__file__)), "seeds")
+
+
+def init_data_seeds_table(conn) -> None:
+    with conn.cursor() as cur:
+        cur.execute(
+            """
+            CREATE TABLE IF NOT EXISTS data_seeds (
+                seed_name VARCHAR(255) PRIMARY KEY,
+                checksum VARCHAR(64) NOT NULL,
+                executed_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+                last_run_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
+            )
+            """
+        )
+    conn.commit()
+
+
+def _seed_sort_key(filename: str) -> Tuple[int, str]:
+    stem = filename
+    for suffix in (".sql", ".py"):
+        if stem.endswith(suffix):
+            stem = stem[: -len(suffix)]
+            break
+    match = _SEED_PREFIX.match(stem)
+    if match:
+        return (int(match.group(1)), stem)
+    return (0, stem)
+
+
+def _seed_stem(filename: str) -> str:
+    if filename.endswith(".dev.sql"):
+        return filename[: -len(".dev.sql")]
+    if filename.endswith(".sql"):
+        return filename[: -len(".sql")]
+    if filename.endswith(".py"):
+        return filename[: -len(".py")]
+    return filename
+
+
+def seed_files(seeds_dir: str) -> List[Tuple[str, str, str]]:
+    """Return (seed_name, filepath, kind) sorted by numeric prefix."""
+    rows: List[Tuple[str, str, str]] = []
+    if not os.path.isdir(seeds_dir):
+        return rows
+
+    for filename in os.listdir(seeds_dir):
+        if filename.startswith("_") or not filename.startswith("seed_"):
+            continue
+        if filename.endswith(".sql"):
+            kind = "sql"
+        elif filename.endswith(".py"):
+            kind = "py"
+        else:
+            continue
+        stem = _seed_stem(filename)
+        rows.append((stem, os.path.join(seeds_dir, filename), kind))
+
+    rows.sort(key=lambda item: _seed_sort_key(item[0]))
+    return rows
+
+
+def seed_applies_in_environment(filename: str) -> bool:
+    if _DEV_MARKER in filename and is_production():
+        return False
+    return True
+
+
+def file_checksum(filepath: str) -> str:
+    digest = hashlib.sha256()
+    with open(filepath, "rb") as handle:
+        for chunk in iter(lambda: handle.read(65536), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def applied_seed_checksum(conn, seed_name: str) -> Optional[str]:
+    with conn.cursor() as cur:
+        cur.execute("SELECT checksum FROM data_seeds WHERE seed_name = %s", (seed_name,))
+        row = cur.fetchone()
+        return row[0] if row else None
+
+
+def record_seed(conn, seed_name: str, checksum: str) -> None:
+    with conn.cursor() as cur:
+        cur.execute(
+            """
+            INSERT INTO data_seeds (seed_name, checksum, executed_at, last_run_at)
+            VALUES (%s, %s, NOW(), NOW())
+            ON CONFLICT (seed_name) DO UPDATE SET
+                checksum = EXCLUDED.checksum,
+                last_run_at = NOW()
+            """,
+            (seed_name, checksum),
+        )
+
+
+def _split_statements(sql_text: str) -> List[str]:
+    parts = sqlparse.split(sql_text.strip())
+    return [part.strip() for part in parts if part and part.strip()]
+
+
+def run_sql_seed(conn, filepath: str) -> None:
+    with open(filepath, "r", encoding="utf-8") as handle:
+        body = handle.read()
+    statements = _split_statements(body)
+    with conn.cursor() as cur:
+        for stmt in statements:
+            cur.execute(stmt)
+
+
+def run_python_seed(filepath: str) -> None:
+    module_name = f"kairo_seed_{hashlib.md5(filepath.encode()).hexdigest()[:12]}"
+    spec = importlib.util.spec_from_file_location(module_name, filepath)
+    if spec is None or spec.loader is None:
+        raise RuntimeError(f"Seed-Modul nicht ladbar: {filepath}")
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    run_fn: Callable[[], None] | None = getattr(module, "run", None)
+    if run_fn is None:
+        raise RuntimeError(f"Seed {filepath} hat keine run()-Funktion")
+    run_fn()
+
+
+def connect_with_retry(max_retries: int = 30):
+    p = db_params()
+    for attempt in range(max_retries):
+        try:
+            conn = get_connection()
+            conn.autocommit = False
+            print(f"[OK] Connected to database: {p['dbname']}")
+            return conn
+        except psycopg2.OperationalError:
+            if attempt >= max_retries - 1:
+                raise
+            print(f"Waiting for database... ({attempt + 1}/{max_retries})")
+            time.sleep(2)
+
+
+def run_seed(conn, seed_name: str, filepath: str, kind: str) -> tuple[bool, object]:
+    print(f"Running seed: {seed_name}")
+    try:
+        if kind == "sql":
+            run_sql_seed(conn, filepath)
+        else:
+            conn.commit()
+            conn.close()
+            run_python_seed(filepath)
+            conn = connect_with_retry(max_retries=5)
+
+        checksum = file_checksum(filepath)
+        record_seed(conn, seed_name, checksum)
+        conn.commit()
+        print(f"  [OK] {seed_name}")
+        return True, conn
+    except Exception as exc:
+        try:
+            conn.rollback()
+        except Exception:
+            pass
+        print(f"  [FAIL] {seed_name}: {exc}")
+        return False, conn
+
+
+def pending_seeds(
+    conn,
+    seeds_dir: str,
+    *,
+    only: Optional[List[str]] = None,
+    force: bool = False,
+) -> List[Tuple[str, str, str]]:
+    selected: List[Tuple[str, str, str]] = []
+    for seed_name, filepath, kind in seed_files(seeds_dir):
+        filename = os.path.basename(filepath)
+        if not seed_applies_in_environment(filename):
+            print(f"  [SKIP] {seed_name} (nur Nicht-Prod)")
+            continue
+        if only and seed_name not in only:
+            continue
+        checksum = file_checksum(filepath)
+        if force or applied_seed_checksum(conn, seed_name) != checksum:
+            selected.append((seed_name, filepath, kind))
+    return selected
+
+
+def main(argv: Optional[List[str]] = None) -> int:
+    argv = argv if argv is not None else sys.argv[1:]
+    only: Optional[List[str]] = None
+    force = False
+
+    idx = 0
+    while idx < len(argv):
+        arg = argv[idx]
+        if arg == "--only" and idx + 1 < len(argv):
+            only = [part.strip() for part in argv[idx + 1].split(",") if part.strip()]
+            idx += 2
+            continue
+        if arg == "--force":
+            force = True
+            idx += 1
+            continue
+        print(f"[FAIL] Unbekanntes Argument: {arg}")
+        return 1
+
+    print("=" * 60)
+    print("Jinkendo Kairo — Data Seeds")
+    print("=" * 60)
+
+    seeds_dir = seeds_directory()
+    if not os.path.isdir(seeds_dir):
+        print(f"[OK] Kein seeds-Verzeichnis ({seeds_dir}) — nichts zu tun.")
+        return 0
+
+    try:
+        conn = connect_with_retry()
+        init_data_seeds_table(conn)
+        to_run = pending_seeds(conn, seeds_dir, only=only, force=force)
+
+        if not to_run:
+            print("[OK] Alle Seeds aktuell — nichts auszuführen.")
+            conn.close()
+            return 0
+
+        print(f"{len(to_run)} Seed(s) ausstehend:")
+        for seed_name, _, _ in to_run:
+            print(f"  - {seed_name}")
+
+        for seed_name, filepath, kind in to_run:
+            ok, conn = run_seed(conn, seed_name, filepath, kind)
+            if not ok:
+                conn.close()
+                return 1
+
+        conn.close()
+        print("[OK] Seeds abgeschlossen.")
+        return 0
+    except Exception as exc:
+        print(f"[FAIL] {exc}")
+        return 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())

backend/seeds/seed_001_cleanup_pytest_artifacts.dev.sql

@@ -0,0 +1,32 @@
+-- Dev/CI: pytest-Artefakte entfernen (idempotent, sicher mehrfach ausführbar).
+-- Nur in Nicht-Prod-Umgebungen (Dateiname *.dev.sql).
+
+-- Abhängigkeiten vor User-Löschung (human actors: user_id NOT NULL constraint)
+DELETE FROM actors
+WHERE user_id IN (SELECT id FROM users WHERE email ~* '@example\.com$');
+
+DELETE FROM tenant_memberships
+WHERE user_id IN (SELECT id FROM users WHERE email ~* '@example\.com$');
+
+DELETE FROM sessions
+WHERE user_id IN (SELECT id FROM users WHERE email ~* '@example\.com$');
+
+DELETE FROM audit_log
+WHERE user_id IN (SELECT id FROM users WHERE email ~* '@example\.com$');
+
+DELETE FROM users WHERE email ~* '@example\.com$';
+
+-- Verwaiste pytest-Tenants (Factory-Slug t-{10 hex}), ohne Memberships
+DELETE FROM actors a
+USING tenants t
+WHERE a.tenant_id = t.id
+  AND t.slug ~ '^t-[a-f0-9]{10}$'
+  AND NOT EXISTS (
+      SELECT 1 FROM tenant_memberships tm WHERE tm.tenant_id = t.id
+  );
+
+DELETE FROM tenants t
+WHERE t.slug ~ '^t-[a-f0-9]{10}$'
+  AND NOT EXISTS (
+      SELECT 1 FROM tenant_memberships tm WHERE tm.tenant_id = t.id
+  );

backend/seeds/seed_002_bootstrap_admin.py

@@ -0,0 +1,9 @@
+"""Bootstrap portal admin from KAIRO_BOOTSTRAP_* when no users exist."""
+
+from __future__ import annotations
+
+
+def run() -> None:
+    from bootstrap import bootstrap_admin_if_needed
+
+    bootstrap_admin_if_needed()

backend/tests/conftest.py

@@ -7,6 +7,7 @@ import os
 import pytest
 
 os.environ.setdefault("SKIP_DB_MIGRATE", "1")
+os.environ.setdefault("SKIP_SEEDS", "1")
 os.environ.setdefault("SKIP_BOOTSTRAP", "1")
 
 
@@ -37,6 +38,16 @@ def _run_migrations():
     assert run_migrations.main() == 0
 
 
+@pytest.fixture(scope="session", autouse=True)
+def _seed_hygiene():
+    """Vor/nach Tests pytest-Artefakte aus der geteilten Dev-DB entfernen."""
+    import run_seeds
+
+    run_seeds.main(["--only", "seed_001_cleanup_pytest_artifacts"])
+    yield
+    run_seeds.main(["--only", "seed_001_cleanup_pytest_artifacts"])
+
+
 @pytest.fixture()
 def client():
     import importlib

backend/tests/test_migrations.py

@@ -33,6 +33,7 @@ def test_migration_runner_finds_migrations():
     names = [name for name, _ in files]
     assert "001_init_core" in names
     assert "002_auth_identity_tenant_actor" in names
+    assert "003_data_seeds_tracking" in names
 
 
 def test_migration_runner_is_idempotent():
@@ -44,6 +45,7 @@ def test_migration_runner_is_idempotent():
     conn.close()
     assert "001_init_core" in executed
     assert "002_auth_identity_tenant_actor" in executed
+    assert "003_data_seeds_tracking" in executed
 
 
 def test_core_table_exists():

backend/tests/test_seeds.py

@@ -0,0 +1,37 @@
+"""Data seed runner tests (require PostgreSQL)."""
+
+from __future__ import annotations
+
+import uuid
+
+import run_seeds
+from tests.factories import create_user
+
+
+def test_seed_runner_finds_seeds():
+    seeds_dir = run_seeds.seeds_directory()
+    names = [name for name, _, _ in run_seeds.seed_files(seeds_dir)]
+    assert "seed_001_cleanup_pytest_artifacts" in names
+    assert "seed_002_bootstrap_admin" in names
+
+
+def test_cleanup_seed_removes_example_com_users():
+    suffix = uuid.uuid4().hex[:8]
+    user = create_user(email=f"seed-test-{suffix}@example.com")
+
+    assert run_seeds.main(["--only", "seed_001_cleanup_pytest_artifacts", "--force"]) == 0
+
+    from db import get_connection
+
+    conn = get_connection()
+    try:
+        with conn.cursor() as cur:
+            cur.execute("SELECT COUNT(*) FROM users WHERE id = %s", (user["id"],))
+            assert cur.fetchone()[0] == 0
+    finally:
+        conn.close()
+
+
+def test_cleanup_seed_is_idempotent():
+    assert run_seeds.main(["--only", "seed_001_cleanup_pytest_artifacts"]) == 0
+    assert run_seeds.main(["--only", "seed_001_cleanup_pytest_artifacts"]) == 0

backend/version.py

@@ -1,3 +1,3 @@
 APP_VERSION = "0.2.0-ap0.2"
-DB_SCHEMA_VERSION = "002"
+DB_SCHEMA_VERSION = "003"
 APP_NAME = "jinkendo-kairo"

docs/MIGRATIONS.md

@@ -0,0 +1,51 @@
+# Migrationen & Data-Seeds
+
+Kairo trennt **Schema-Migrationen** (einmalig) von **Data-Seeds** (idempotent, bei Änderung erneut ausführbar).
+
+## Schema-Migrationen
+
+| | |
+|---|---|
+| **Pfad** | `backend/migrations/NNN_beschreibung.sql` |
+| **Tracking** | Tabelle `schema_migrations` |
+| **Ausführung** | Beim Backend-Start (`run_migrations.py`), überspringbar via `SKIP_DB_MIGRATE=1` |
+| **Regel** | Jede Datei wird **genau einmal** angewendet. Änderungen → neue nummerierte Datei. |
+
+## Data-Seeds
+
+| | |
+|---|---|
+| **Pfad** | `backend/seeds/seed_NNN_beschreibung.sql` oder `.py` |
+| **Tracking** | Tabelle `data_seeds` (Name + SHA256-Checksum) |
+| **Ausführung** | Nach Schema-Migrationen (`run_seeds.py`), überspringbar via `SKIP_SEEDS=1` |
+| **Regel** | Seed läuft erneut, wenn die Datei **geändert** wurde (Checksum abweicht) oder `--force` gesetzt ist. |
+
+### Umgebungsfilter
+
+Dateien mit **`.dev.`** im Namen (z. B. `seed_001_…​.dev.sql`) laufen **nicht in Production** (`ENVIRONMENT=production`).
+
+### Seeds manuell ausführen
+
+```bash
+docker compose -f docker-compose.dev-env.yml exec backend python run_seeds.py
+docker compose -f docker-compose.dev-env.yml exec backend python run_seeds.py --only seed_001_cleanup_pytest_artifacts --force
+```
+
+## Aktuelle Seeds
+
+| Seed | Typ | Zweck |
+|------|-----|--------|
+| `seed_001_cleanup_pytest_artifacts` | SQL (dev) | Entfernt pytest/CI-User (`*@example.com`) und verwaiste Test-Tenants |
+| `seed_002_bootstrap_admin` | Python | Legt Systemadmin aus `KAIRO_BOOTSTRAP_*` an, wenn noch kein User existiert |
+
+## Neuen Seed anlegen
+
+1. Datei `backend/seeds/seed_NNN_kurzname.sql` oder `.py` anlegen (Python: `def run() -> None:`).
+2. SQL idempotent halten (`DELETE … WHERE …`, `INSERT … ON CONFLICT`, etc.).
+3. Nur Dev/CI: `.dev.sql` / `.dev.py` Suffix verwenden.
+4. Nach Deploy prüfen: `python run_seeds.py` — bei geänderter Datei wird der Seed automatisch erneut ausgeführt.
+
+## Tests & CI
+
+- pytest setzt `SKIP_SEEDS=1` beim App-Import; Cleanup läuft über Fixture + nach CI-pytest.
+- Geteilte Dev-DB wird nach jedem Test-Lauf bereinigt (`run_seeds.py --only seed_001_cleanup_pytest_artifacts --force`).