docs(examples): Alembic migration scaffold for v3 reference seller (#390)

* docs(examples): Alembic migration scaffold for v3 reference seller

Refs #382. Adds production migration tooling while keeping create_all
as the default fast-iteration boot path (per DX review).

https://claude.ai/code/session_01SnEYzkwXDnJEFpj7zNVSY6

* fix(examples): address pre-PR review blockers in Alembic scaffold

- env.py: narrow _db_url to str via os.environ[key] (no mypy Any)
- migrate.py: redact DB password before printing URL
- test_migrations.py: drop psycopg2 assumption; use sqlalchemy async
  engine for all schema inspection (asyncpg already installed)
- test_migrations.py: ensure downgrade test starts from head state

https://claude.ai/code/session_01SnEYzkwXDnJEFpj7zNVSY6

* fix(examples): add 0002 migration covering broadening-cycle additions

PR #408 merged three schema additions that were not covered by the
0001 initial migration: `media_buys.invoice_recipient` (JSON, nullable),
the `creatives` table (idempotency-keyed on tenant_id + creative_id),
and the `performance_feedback` table (FK'd to media_buys.id).

Without this migration, `alembic upgrade head` against a database
created after #408 would produce drift that `alembic check` flags in
CI and on every production deploy.

Adds `0002_broadening_cycle.py` (revises 0001) covering all three
additions, and updates the integration test to expect all eight tables.

Refs #382

https://claude.ai/code/session_0121TgqHyzzjyZnQiGGmjrRu

---------

Co-authored-by: Claude <noreply@anthropic.com>

Author: bokelley

examples/v3_reference_seller/README.md

@@ -147,15 +147,79 @@ exist; the reference seller wires the simpler defaults:
   in `src/app.py` for production durability. Both classes ship in
   the SDK; this seller's `app.py` uses the in-memory variants for
   fast iteration.
-- **Alembic migrations** — `Base.metadata.create_all` runs at boot
-  (idempotent on table existence — it does NOT detect column
-  renames or type changes on existing tables). Adopters who
-  prototyped against earlier branches and pulled new column
-  changes should drop and recreate the dev database; production
-  sellers wire Alembic and version their schema changes.
 - **Admin CRUD API** — separate Starlette app for tenant / agent
   CRUD. Patterns to come; for now use `seed.py` and direct SQL.
 
+## Migrations
+
+The app boots with `Base.metadata.create_all` — idempotent on table
+existence, but **blind to column renames, type changes, and new columns
+on existing tables**.  For local fast-iteration this is fine.  Once you
+have production data, use Alembic to evolve the schema safely.
+
+> ⚠️ **`create_all` is unsafe for schema evolution once production data
+> exists.** Column renames and type changes applied after first boot
+> will not be detected and will silently leave the schema stale.
+
+### Install Alembic
+
+```bash
+pip install alembic
+# or, if using a requirements file:
+echo "alembic" >> requirements.txt && pip install -r requirements.txt
+```
+
+### Apply migrations
+
+```bash
+cd examples/v3_reference_seller
+
+# Apply all pending migrations (run after every git pull that touches models).
+DATABASE_URL=postgresql+asyncpg://postgres@localhost/adcp python -m migrate
+
+# Equivalent direct alembic invocation:
+DATABASE_URL=postgresql+asyncpg://postgres@localhost/adcp alembic upgrade head
+```
+
+### Generate a new migration after changing models
+
+```bash
+cd examples/v3_reference_seller
+DATABASE_URL=postgresql+asyncpg://postgres@localhost/adcp \
+  alembic revision --autogenerate -m "describe your change"
+```
+
+Alembic compares the live database to `Base.metadata` and emits a
+migration file under `alembic/versions/`.  **Always review the generated
+file before committing** — autogenerate misses some constructs (partial
+index predicates, custom CHECK constraints, server defaults).
+
+> ⚙️ **Adding a new model file?** Import it in `alembic/env.py` alongside
+> `src.models` and `src.audit`, or autogenerate will silently omit its
+> tables from the migration.
+
+### Roll back
+
+```bash
+# Roll back one step.
+DATABASE_URL=postgresql+asyncpg://postgres@localhost/adcp alembic downgrade -1
+
+# Roll back to before any migrations (drops all tables defined in this schema).
+DATABASE_URL=postgresql+asyncpg://postgres@localhost/adcp alembic downgrade base
+```
+
+> ⚠️ **`downgrade` in production is irreversible without a data backup.**
+> Take a snapshot before running downgrade against any database that
+> holds real data.
+
+### Run migration integration tests
+
+```bash
+# Uses a throw-away database (adcp_test) so the migration run starts clean.
+DATABASE_URL=postgresql+asyncpg://postgres@localhost/adcp_test \
+  pytest examples/v3_reference_seller/tests/test_migrations.py -m integration -v
+```
+
 ## Customization
 
 Adopters typically change:

examples/v3_reference_seller/alembic.ini

@@ -0,0 +1,50 @@
+# Alembic configuration for the v3 reference seller example.
+#
+# Run from the examples/v3_reference_seller/ directory:
+#
+#   DATABASE_URL=postgresql+asyncpg://... alembic upgrade head
+#
+# When embedding this example inside a larger repo, update
+# script_location to an absolute path (e.g. /path/to/alembic) so
+# Alembic can find the migration scripts regardless of cwd.
+
+[alembic]
+script_location = alembic
+
+# DATABASE_URL is read from the environment in env.py — leave this
+# blank so it is never accidentally hardcoded in version control.
+sqlalchemy.url =
+
+[loggers]
+keys = root,sqlalchemy,alembic
+
+[handlers]
+keys = console
+
+[formatters]
+keys = generic
+
+[logger_root]
+level = WARN
+handlers = console
+qualname =
+
+[logger_sqlalchemy]
+level = WARN
+handlers =
+qualname = sqlalchemy.engine
+
+[logger_alembic]
+level = INFO
+handlers =
+qualname = alembic
+
+[handler_console]
+class = StreamHandler
+args = (sys.stderr,)
+level = NOTSET
+formatter = generic
+
+[formatter_generic]
+format = %(levelname)-5.5s [%(name)s] %(message)s
+datefmt = %H:%M:%S

examples/v3_reference_seller/alembic/env.py

@@ -0,0 +1,104 @@
+"""Alembic environment for the v3 reference seller.
+
+Uses SQLAlchemy's async engine (asyncpg) via the standard Alembic
+async pattern.  Run from the examples/v3_reference_seller/ directory:
+
+    DATABASE_URL=postgresql+asyncpg://... alembic upgrade head
+
+For autogenerate to capture every table, both src.models and src.audit
+must be imported before target_metadata is read.  Missing either import
+silently omits that module's tables from the generated migration.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import os
+import sys
+from logging.config import fileConfig
+from pathlib import Path
+
+from alembic import context
+from sqlalchemy.ext.asyncio import create_async_engine
+
+# ---------------------------------------------------------------------------
+# Path wiring — make ``src.*`` importable when env.py is executed from the
+# examples/v3_reference_seller/ directory by the ``alembic`` CLI.
+# ---------------------------------------------------------------------------
+_HERE = Path(__file__).resolve().parent.parent  # examples/v3_reference_seller/
+if str(_HERE) not in sys.path:
+    sys.path.insert(0, str(_HERE))
+
+# Import all ORM modules so their tables appear in Base.metadata.
+# Adding a new model file?  Import it here or autogenerate will miss it.
+import src.audit  # noqa: E402, F401 — registers AuditEventRow on Base.metadata
+from src.models import Base  # noqa: E402
+
+target_metadata = Base.metadata
+
+# ---------------------------------------------------------------------------
+# Alembic config
+# ---------------------------------------------------------------------------
+config = context.config
+
+if config.config_file_name is not None:
+    fileConfig(config.config_file_name)
+
+# DATABASE_URL comes from the environment; never hardcode it here.
+try:
+    _db_url: str = os.environ["DATABASE_URL"]
+except KeyError:
+    raise RuntimeError(
+        "DATABASE_URL environment variable is not set. "
+        "Example: DATABASE_URL=postgresql+asyncpg://postgres@localhost/adcp alembic upgrade head"
+    ) from None
+config.set_main_option("sqlalchemy.url", _db_url)
+
+
+# ---------------------------------------------------------------------------
+# Migration helpers
+# ---------------------------------------------------------------------------
+
+def run_migrations_offline() -> None:
+    """Emit SQL to stdout rather than connecting to the DB.
+
+    Useful for generating a migration script to review or apply manually.
+    """
+    url = config.get_main_option("sqlalchemy.url")
+    context.configure(
+        url=url,
+        target_metadata=target_metadata,
+        literal_binds=True,
+        dialect_opts={"paramstyle": "named"},
+        compare_type=True,
+    )
+    with context.begin_transaction():
+        context.run_migrations()
+
+
+def do_run_migrations(connection) -> None:
+    context.configure(
+        connection=connection,
+        target_metadata=target_metadata,
+        compare_type=True,
+    )
+    with context.begin_transaction():
+        context.run_migrations()
+
+
+async def run_migrations_online() -> None:
+    """Create an async engine and run migrations inside a sync wrapper.
+
+    Alembic's migration functions are synchronous; ``run_sync`` bridges
+    the gap so we can use an asyncpg engine end-to-end.
+    """
+    connectable = create_async_engine(_db_url, echo=False)
+    async with connectable.connect() as connection:
+        await connection.run_sync(do_run_migrations)
+    await connectable.dispose()
+
+
+if context.is_offline_mode():
+    run_migrations_offline()
+else:
+    asyncio.run(run_migrations_online())

examples/v3_reference_seller/alembic/script.py.mako

@@ -0,0 +1,28 @@
+"""${message}
+
+Revision ID: ${up_revision}
+Revises: ${down_revision | comma,n}
+Create Date: ${create_date}
+
+"""
+from __future__ import annotations
+
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+${imports if imports else ""}
+
+# revision identifiers, used by Alembic.
+revision: str = ${repr(up_revision)}
+down_revision: Union[str, None] = ${repr(down_revision)}
+branch_labels: Union[str, Sequence[str], None] = ${repr(branch_labels)}
+depends_on: Union[str, Sequence[str], None] = ${repr(depends_on)}
+
+
+def upgrade() -> None:
+    ${upgrades if upgrades else "pass"}
+
+
+def downgrade() -> None:
+    ${downgrades if downgrades else "pass"}