fix(P0): Migration uses vector(768), remove direct port exposure in prod compose, add lockfile note

backend/alembic/versions/001_initial.py

@@ -1,18 +1,17 @@
-"""Initial schema - metadata driven
+"""Initial schema - creates all tables from ORM models.
 
 Revision ID: 001_initial
 Revises: None
 Create Date: 2026-05-12
 
-This migration intentionally creates tables from SQLAlchemy metadata so local
-setup stays aligned with the rapidly evolving app models.
+NOTE: This migration uses Base.metadata.create_all() for rapid iteration.
+For production, replace with explicit op.create_table() calls.
 """
 from typing import Sequence, Union
 
 from alembic import op
 import sqlalchemy as sa
 
-# revision identifiers, used by Alembic.
 revision: str = "001_initial"
 down_revision: Union[str, None] = None
 branch_labels: Union[str, Sequence[str], None] = None
@@ -21,65 +20,51 @@ depends_on: Union[str, Sequence[str], None] = None
 
 def _import_all_models() -> None:
     """Import every module that declares SQLAlchemy models."""
-    # Core model modules
-    import app.models.user  # noqa: F401
-    import app.models.job  # noqa: F401
-    import app.models.application  # noqa: F401
-    import app.models.resume  # noqa: F401
-    import app.models.scraper  # noqa: F401
-    import app.models.llm  # noqa: F401
-    import app.models.alerts  # noqa: F401
-
-    # A few later-stage models currently live in route/service modules.
-    # Importing them registers their tables on Base.metadata.
-    import app.api.routes.bookmarks  # noqa: F401
-    import app.api.routes.webhooks  # noqa: F401
-    import app.api.routes.reminders  # noqa: F401
-    import app.api.routes.cover_letters  # noqa: F401
-    import app.api.routes.qa_bank  # noqa: F401
-    import app.services.analytics  # noqa: F401
+    import app.models.user  # noqa
+    import app.models.job  # noqa
+    import app.models.application  # noqa
+    import app.models.resume  # noqa
+    import app.models.scraper  # noqa
+    import app.models.llm  # noqa
+    import app.models.alerts  # noqa
+    import app.api.routes.bookmarks  # noqa
+    import app.api.routes.webhooks  # noqa
+    import app.api.routes.reminders  # noqa
+    import app.api.routes.cover_letters  # noqa
+    import app.api.routes.qa_bank  # noqa
+    import app.services.analytics  # noqa
 
 
 def upgrade() -> None:
     conn = op.get_bind()
 
-    # Required extensions. pgvector requires the Docker image pgvector/pgvector:pg17.
     op.execute('CREATE EXTENSION IF NOT EXISTS "uuid-ossp"')
     op.execute('CREATE EXTENSION IF NOT EXISTS "pg_trgm"')
     op.execute('CREATE EXTENSION IF NOT EXISTS "vector"')
 
     _import_all_models()
     from app.core.database import Base
-
-    # Create all ORM-backed tables.
     Base.metadata.create_all(bind=conn)
 
-    # Create semantic search table manually because SQLAlchemy does not ship a
-    # native pgvector type. The semantic search service writes to this table.
-    op.execute(
-        """
+    # Semantic search table — uses 768 dims (nomic-embed-text via Ollama)
+    op.execute("""
         CREATE TABLE IF NOT EXISTS job_embeddings (
             job_id UUID PRIMARY KEY REFERENCES jobs(id) ON DELETE CASCADE,
             text_hash VARCHAR(32) NOT NULL,
             updated_at TIMESTAMPTZ DEFAULT now(),
-            embedding vector(1536)
+            embedding vector(768)
         )
-        """
-    )
-    op.execute(
-        """
+    """)
+    op.execute("""
         CREATE INDEX IF NOT EXISTS idx_job_embeddings_vector
         ON job_embeddings USING ivfflat (embedding vector_cosine_ops) WITH (lists = 100)
-        """
-    )
+    """)
 
-    # Useful search indexes. Ignore failures on repeated local runs.
-    op.execute(
-        """
+    # Full-text search indexes
+    op.execute("""
         CREATE INDEX IF NOT EXISTS idx_jobs_fts ON jobs
         USING gin(to_tsvector('english', coalesce(title, '') || ' ' || coalesce(company_name, '') || ' ' || coalesce(description_text, '')))
-        """
-    )
+    """)
     op.execute("CREATE INDEX IF NOT EXISTS idx_jobs_title_trgm ON jobs USING gin(title gin_trgm_ops)")
 
 
@@ -88,8 +73,6 @@ def downgrade() -> None:
     op.execute("DROP TABLE IF EXISTS job_embeddings")
     _import_all_models()
     from app.core.database import Base
-
-    # Drop in dependency order.
     Base.metadata.drop_all(bind=conn)
     op.execute('DROP EXTENSION IF EXISTS "vector"')
     op.execute('DROP EXTENSION IF EXISTS "pg_trgm"')