diff --git a/.github/workflows/db-migrations.yaml b/.github/workflows/db-migrations.yaml
new file mode 100644
index 000000000..b36e9a721
--- /dev/null
+++ b/.github/workflows/db-migrations.yaml
@@ -0,0 +1,62 @@
+name: DB Migrations
+
+on:
+  push:
+    branches:
+      - develop
+  pull_request:
+  workflow_dispatch:
+
+jobs:
+  alembic-migrations:
+    strategy:
+      matrix:
+        python-version: ["3.10"]
+
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install dependencies
+        run: |
+          pip install --upgrade --upgrade-strategy eager -r requirements.txt
+          pip install --upgrade --upgrade-strategy eager -r requirements-dev.txt
+
+      - name: Set DB env vars
+        run: |
+          echo "DB_FILE=$RUNNER_TEMP/dashai-ci.sqlite3" >> "$GITHUB_ENV"
+          echo "DATABASE_URL=sqlite:///$RUNNER_TEMP/dashai-ci.sqlite3" >> "$GITHUB_ENV"
+          echo "Temp dir: $RUNNER_TEMP"
+
+      - name: Show Alembic info
+        run: |
+          alembic --version
+          echo "DB will be at: $DATABASE_URL"
+
+      - name: Prepare temp dir
+        run: |
+          mkdir -p "$(dirname "$DB_FILE")"
+          rm -f "$DB_FILE"
+
+      # upgrade to head
+      - name: Upgrade to head
+        run: |
+          alembic -x url="$DATABASE_URL" upgrade head
+
+      # Checks downgrade and upgrade again (reversibility)
+      - name: Downgrade to base and upgrade again (reversibility check)
+        run: |
+          alembic -x url="$DATABASE_URL" downgrade base
+          alembic -x url="$DATABASE_URL" upgrade head
+
+      - name: Check for pending autogenerate (python-based)
+        env:
+          PYTHONPATH: "${PYTHONPATH}:."
+        run: python -m scripts.ci_alembic_check
diff --git a/.github/workflows/pre-commit.yaml b/.github/workflows/pre-commit.yaml
index 0872ac1c4..4356ba6eb 100644
--- a/.github/workflows/pre-commit.yaml
+++ b/.github/workflows/pre-commit.yaml
@@ -1,26 +1,27 @@
 name: Pre-commit checks
-on:
-  push
+on: push
 jobs:
   pre-commit:
     runs-on: ubuntu-latest
     steps:
-      - name: 'Check out repository code'
+      - name: "Check out repository code"
         uses: actions/checkout@v3
 
       - name: Use Node.js
         uses: actions/setup-node@v3
         with:
-          node-version: '18.x'
-          cache: 'yarn'
+          node-version: "18.x"
+          cache: "yarn"
           cache-dependency-path: DashAI/front/yarn.lock
 
       - name: Install frontend deps
         working-directory: DashAI/front
         run: yarn install --frozen-lockfile
 
-      - name: Setup latest Python 3.x
-        uses: actions/setup-python@v3
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
 
       - name: Install pre-commit
         run: |
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 79c3bedcb..0fdb94455 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -15,13 +15,13 @@ repos:
     rev: v0.12.10
     hooks:
       - id: ruff
-        args: [--fix]
+        args: [--fix, "--show-fixes"]
       - id: ruff-format
   - repo: https://github.com/pre-commit/mirrors-prettier
     rev: "v4.0.0-alpha.8"
     hooks:
       - id: prettier
-        files: \.[jt]sx?$  # *.js, *.jsx, *.ts, *.tsx
+        files: \.[jt]sx?$ # *.js, *.jsx, *.ts, *.tsx
         types: [file]
         args:
           - "--config=DashAI/front/.prettierrc"
diff --git a/DashAI/alembic/env.py b/DashAI/alembic/env.py
new file mode 100644
index 000000000..0f3c44a1b
--- /dev/null
+++ b/DashAI/alembic/env.py
@@ -0,0 +1,91 @@
+import os
+from pathlib import Path
+from logging.config import fileConfig
+
+from sqlalchemy import engine_from_config, pool
+from alembic import context
+
+from DashAI.back.dependencies.database import Base
+
+config = context.config
+
+
+def _resolve_url() -> str:
+    """
+    Prioridad url sql:
+      1) alembic -x url=...
+      2) env var DATABASE_URL
+      3) value already present in alembic.ini (config.get_main_option)
+      4) fallback using the same logic as the application (~/.DashAI/db.sqlite)
+    """
+    # 1) cli -x url
+    xargs = context.get_x_argument(as_dictionary=True)
+    if xargs and xargs.get("url"):
+        return xargs["url"]
+
+    # 2) environment
+    env_url = os.getenv("DATABASE_URL")
+    if env_url:
+        return env_url
+
+    # 3) value from alembic.ini (if any)
+    cfg_url = config.get_main_option("sqlalchemy.url")
+    if cfg_url and cfg_url != "sqlite:///":
+        return cfg_url
+
+    # 4) fallback - same as application config
+    # Use the same default path as the application: ~/.DashAI/db.sqlite
+    from DashAI.back.config import DefaultSettings
+    settings = DefaultSettings()
+    local_path = Path(settings.LOCAL_PATH).expanduser().absolute()
+    db_path = local_path / settings.SQLITE_DB_PATH
+    return f"sqlite:///{db_path}"
+
+
+config.set_main_option("sqlalchemy.url", _resolve_url())
+
+if config.config_file_name is not None:
+    fileConfig(config.config_file_name)
+
+target_metadata = Base.metadata
+
+
+def run_migrations_offline():
+    url = config.get_main_option("sqlalchemy.url")
+    context.configure(
+        url=url,
+        target_metadata=target_metadata,
+        literal_binds=True,
+        compare_type=True,
+        compare_server_default=True,
+    )
+    with context.begin_transaction():
+        context.run_migrations()
+
+
+def run_migrations_online():
+    ini = dict(config.get_section(config.config_ini_section) or {})
+
+    connectable = engine_from_config(
+        ini,
+        prefix="sqlalchemy.",
+        poolclass=pool.NullPool,
+        future=True,
+    )
+
+    with connectable.connect() as connection:
+        context.configure(
+            connection=connection,
+            target_metadata=target_metadata,
+            compare_type=True,
+            compare_server_default=True,
+            render_as_batch=True,
+        )
+        with context.begin_transaction():
+            context.run_migrations()
+
+
+if context.is_offline_mode():
+    run_migrations_offline()
+else:
+    run_migrations_online()
diff --git a/alembic/script.py.mako b/DashAI/alembic/script.py.mako
similarity index 57%
rename from alembic/script.py.mako
rename to DashAI/alembic/script.py.mako
index 55df2863d..fbc4b07dc 100644
--- a/alembic/script.py.mako
+++ b/DashAI/alembic/script.py.mako
@@ -5,15 +5,17 @@ Revises: ${down_revision | comma,n}
 Create Date: ${create_date}
 
 """
+from typing import Sequence, Union
+
 from alembic import op
 import sqlalchemy as sa
 ${imports if imports else ""}
 
 # revision identifiers, used by Alembic.
-revision = ${repr(up_revision)}
-down_revision = ${repr(down_revision)}
-branch_labels = ${repr(branch_labels)}
-depends_on = ${repr(depends_on)}
+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:
diff --git a/DashAI/alembic/versions/57ff87150ae2_experiments_to_model_session.py b/DashAI/alembic/versions/57ff87150ae2_experiments_to_model_session.py
new file mode 100644
index 000000000..bd2815f5f
--- /dev/null
+++ b/DashAI/alembic/versions/57ff87150ae2_experiments_to_model_session.py
@@ -0,0 +1,76 @@
+"""Experiments to model session
+
+Revision ID: 57ff87150ae2
+Revises: 7662169fa0e0
+Create Date: 2026-01-19 16:14:03.739306
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import sqlite
+
+# revision identifiers, used by Alembic.
+revision: str = '57ff87150ae2'
+down_revision: Union[str, None] = '7662169fa0e0'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('model_session',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('dataset_id', sa.Integer(), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('task_name', sa.String(), nullable=False),
+    sa.Column('input_columns', sa.JSON(), nullable=False),
+    sa.Column('output_columns', sa.JSON(), nullable=False),
+    sa.Column('train_metrics', sa.JSON(), nullable=True),
+    sa.Column('validation_metrics', sa.JSON(), nullable=True),
+    sa.Column('test_metrics', sa.JSON(), nullable=True),
+    sa.Column('splits', sa.JSON(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.ForeignKeyConstraint(['dataset_id'], ['dataset.id'], name=op.f('fk_model_session_dataset_id_dataset')),
+    sa.PrimaryKeyConstraint('id', name=op.f('pk_model_session')),
+    sa.UniqueConstraint('name', name=op.f('uq_model_session_name'))
+    )
+    with op.batch_alter_table('run', schema=None) as batch_op:
+        batch_op.add_column(sa.Column('model_session_id', sa.Integer(), nullable=False))
+        batch_op.drop_constraint(batch_op.f('fk_run_experiment_id_experiment'), type_='foreignkey')
+        batch_op.create_foreign_key(batch_op.f('fk_run_model_session_id_model_session'), 'model_session', ['model_session_id'], ['id'], ondelete='CASCADE')
+        batch_op.drop_column('experiment_id')
+    op.drop_table('experiment')
+
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    with op.batch_alter_table('run', schema=None) as batch_op:
+        batch_op.add_column(sa.Column('experiment_id', sa.INTEGER(), nullable=False))
+        batch_op.drop_constraint(batch_op.f('fk_run_model_session_id_model_session'), type_='foreignkey')
+        batch_op.create_foreign_key(batch_op.f('fk_run_experiment_id_experiment'), 'experiment', ['experiment_id'], ['id'], ondelete='CASCADE')
+        batch_op.drop_column('model_session_id')
+
+    op.create_table('experiment',
+    sa.Column('id', sa.INTEGER(), nullable=False),
+    sa.Column('dataset_id', sa.INTEGER(), nullable=False),
+    sa.Column('name', sa.VARCHAR(), nullable=False),
+    sa.Column('task_name', sa.VARCHAR(), nullable=False),
+    sa.Column('input_columns', sqlite.JSON(), nullable=False),
+    sa.Column('output_columns', sqlite.JSON(), nullable=False),
+    sa.Column('train_metrics', sqlite.JSON(), nullable=True),
+    sa.Column('validation_metrics', sqlite.JSON(), nullable=True),
+    sa.Column('test_metrics', sqlite.JSON(), nullable=True),
+    sa.Column('splits', sqlite.JSON(), nullable=False),
+    sa.Column('created', sa.DATETIME(), nullable=False),
+    sa.Column('last_modified', sa.DATETIME(), nullable=False),
+    sa.ForeignKeyConstraint(['dataset_id'], ['dataset.id'], name=op.f('fk_experiment_dataset_id_dataset')),
+    sa.PrimaryKeyConstraint('id', name=op.f('pk_experiment')),
+    sa.UniqueConstraint('name', name=op.f('uq_experiment_name'))
+    )
+    op.drop_table('model_session')
+    # ### end Alembic commands ###
diff --git a/DashAI/alembic/versions/7662169fa0e0_initial_migrations.py b/DashAI/alembic/versions/7662169fa0e0_initial_migrations.py
new file mode 100644
index 000000000..82b1a4dc7
--- /dev/null
+++ b/DashAI/alembic/versions/7662169fa0e0_initial_migrations.py
@@ -0,0 +1,283 @@
+"""Initial migrations
+
+Revision ID: 7662169fa0e0
+Revises:
+Create Date: 2026-01-07 10:58:08.748292
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '7662169fa0e0'
+down_revision: Union[str, None] = None
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('dataset',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('huey_id', sa.String(), nullable=True),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.Column('file_path', sa.String(), nullable=False),
+    sa.Column('status', sa.Enum('NOT_STARTED', 'DELIVERED', 'STARTED', 'FINISHED', 'ERROR', name='datasetstatus'), nullable=False),
+    sa.PrimaryKeyConstraint('id'),
+    sa.UniqueConstraint('name')
+    )
+    op.create_table('generative_session',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.Column('task_name', sa.String(), nullable=False),
+    sa.Column('model_name', sa.String(), nullable=False),
+    sa.Column('parameters', sa.JSON(), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('description', sa.String(), nullable=True),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('global_explainer',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('run_id', sa.Integer(), nullable=False),
+    sa.Column('huey_id', sa.String(), nullable=True),
+    sa.Column('explainer_name', sa.String(), nullable=False),
+    sa.Column('explanation_path', sa.String(), nullable=True),
+    sa.Column('plot_path', sa.String(), nullable=True),
+    sa.Column('parameters', sa.JSON(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('status', sa.Enum('NOT_STARTED', 'DELIVERED', 'STARTED', 'FINISHED', 'ERROR', name='explainerstatus'), nullable=False),
+    sa.PrimaryKeyConstraint('id'),
+    sa.UniqueConstraint('name')
+    )
+    op.create_table('local_explainer',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('run_id', sa.Integer(), nullable=False),
+    sa.Column('huey_id', sa.String(), nullable=True),
+    sa.Column('explainer_name', sa.String(), nullable=False),
+    sa.Column('dataset_id', sa.Integer(), nullable=False),
+    sa.Column('explanation_path', sa.String(), nullable=True),
+    sa.Column('plots_path', sa.String(), nullable=True),
+    sa.Column('parameters', sa.JSON(), nullable=False),
+    sa.Column('fit_parameters', sa.JSON(), nullable=False),
+    sa.Column('scope', sa.JSON(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('status', sa.Enum('NOT_STARTED', 'DELIVERED', 'STARTED', 'FINISHED', 'ERROR', name='explainerstatus'), nullable=False),
+    sa.PrimaryKeyConstraint('id'),
+    sa.UniqueConstraint('name')
+    )
+    op.create_table('pipeline',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('steps', sa.JSON(), nullable=True),
+    sa.Column('edges', sa.JSON(), nullable=True),
+    sa.Column('exploration', sa.JSON(), nullable=True),
+    sa.Column('train', sa.JSON(), nullable=True),
+    sa.Column('prediction', sa.JSON(), nullable=True),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('plugin',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('author', sa.String(), nullable=False),
+    sa.Column('installed_version', sa.String(), nullable=False),
+    sa.Column('lastest_version', sa.String(), nullable=False),
+    sa.Column('status', sa.Enum('REGISTERED', 'INSTALLED', 'ERROR', name='pluginstatus'), nullable=False),
+    sa.Column('summary', sa.String(), nullable=False),
+    sa.Column('description', sa.String(), nullable=False),
+    sa.Column('description_content_type', sa.String(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.PrimaryKeyConstraint('id'),
+    sa.UniqueConstraint('name')
+    )
+    op.create_table('experiment',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('dataset_id', sa.Integer(), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('task_name', sa.String(), nullable=False),
+    sa.Column('input_columns', sa.JSON(), nullable=False),
+    sa.Column('output_columns', sa.JSON(), nullable=False),
+    sa.Column('train_metrics', sa.JSON(), nullable=True),
+    sa.Column('validation_metrics', sa.JSON(), nullable=True),
+    sa.Column('test_metrics', sa.JSON(), nullable=True),
+    sa.Column('splits', sa.JSON(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.ForeignKeyConstraint(['dataset_id'], ['dataset.id'], ),
+    sa.PrimaryKeyConstraint('id'),
+    sa.UniqueConstraint('name')
+    )
+    op.create_table('generative_process',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.Column('session_id', sa.Integer(), nullable=False),
+    sa.Column('status', sa.Enum('NOT_STARTED', 'DELIVERED', 'STARTED', 'FINISHED', 'ERROR', name='runstatus'), nullable=False),
+    sa.Column('delivery_time', sa.DateTime(), nullable=True),
+    sa.Column('start_time', sa.DateTime(), nullable=True),
+    sa.Column('end_time', sa.DateTime(), nullable=True),
+    sa.ForeignKeyConstraint(['session_id'], ['generative_session.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('notebook',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('dataset_id', sa.Integer(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.Column('file_path', sa.String(), nullable=False),
+    sa.Column('name', sa.String(), nullable=True),
+    sa.Column('description', sa.String(), nullable=True),
+    sa.ForeignKeyConstraint(['dataset_id'], ['dataset.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('parameter_history',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('session_id', sa.Integer(), nullable=False),
+    sa.Column('parameters', sa.JSON(), nullable=False),
+    sa.Column('modified_at', sa.DateTime(), nullable=False),
+    sa.ForeignKeyConstraint(['session_id'], ['generative_session.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('tag',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('plugin_id', sa.Integer(), nullable=False),
+    sa.Column('name', sa.Enum('DashAI', 'Package', 'Task', 'Model', 'Metric', 'Dataloader', 'Converter', 'Explainer', name='plugintag'), nullable=False),
+    sa.ForeignKeyConstraint(['plugin_id'], ['plugin.id'], ),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('converter_list',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('notebook_id', sa.Integer(), nullable=False),
+    sa.Column('huey_id', sa.String(), nullable=True),
+    sa.Column('converter', sa.String(), nullable=False),
+    sa.Column('parameters', sa.JSON(), nullable=False),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.Column('status', sa.Enum('NOT_STARTED', 'DELIVERED', 'STARTED', 'FINISHED', 'ERROR', name='converterliststatus'), nullable=False),
+    sa.Column('delivery_time', sa.DateTime(), nullable=True),
+    sa.Column('start_time', sa.DateTime(), nullable=True),
+    sa.Column('end_time', sa.DateTime(), nullable=True),
+    sa.ForeignKeyConstraint(['notebook_id'], ['notebook.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('explorer',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('notebook_id', sa.Integer(), nullable=False),
+    sa.Column('huey_id', sa.String(), nullable=True),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.Column('columns', sa.JSON(), nullable=False),
+    sa.Column('exploration_type', sa.String(), nullable=False),
+    sa.Column('parameters', sa.JSON(), nullable=False),
+    sa.Column('exploration_path', sa.String(), nullable=True),
+    sa.Column('name', sa.String(), nullable=True),
+    sa.Column('delivery_time', sa.DateTime(), nullable=True),
+    sa.Column('start_time', sa.DateTime(), nullable=True),
+    sa.Column('end_time', sa.DateTime(), nullable=True),
+    sa.Column('status', sa.Enum('NOT_STARTED', 'DELIVERED', 'STARTED', 'FINISHED', 'ERROR', name='explorerstatus'), nullable=False),
+    sa.ForeignKeyConstraint(['notebook_id'], ['notebook.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('process_data',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('data', sa.String(), nullable=False),
+    sa.Column('data_type', sa.String(), nullable=False),
+    sa.Column('process_id', sa.Integer(), nullable=False),
+    sa.Column('is_input', sa.Boolean(), nullable=False),
+    sa.ForeignKeyConstraint(['process_id'], ['generative_process.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('run',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('experiment_id', sa.Integer(), nullable=False),
+    sa.Column('huey_id', sa.String(), nullable=True),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.Column('model_name', sa.String(), nullable=False),
+    sa.Column('parameters', sa.JSON(), nullable=False),
+    sa.Column('split_indexes', sa.JSON(), nullable=True),
+    sa.Column('optimizer_name', sa.String(), nullable=False),
+    sa.Column('optimizer_parameters', sa.JSON(), nullable=False),
+    sa.Column('plot_history_path', sa.String(), nullable=True),
+    sa.Column('plot_slice_path', sa.String(), nullable=True),
+    sa.Column('plot_contour_path', sa.String(), nullable=True),
+    sa.Column('plot_importance_path', sa.String(), nullable=True),
+    sa.Column('goal_metric', sa.String(), nullable=False),
+    sa.Column('artifacts', sa.JSON(), nullable=True),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('description', sa.String(), nullable=True),
+    sa.Column('run_path', sa.String(), nullable=True),
+    sa.Column('status', sa.Enum('NOT_STARTED', 'DELIVERED', 'STARTED', 'FINISHED', 'ERROR', name='runstatus'), nullable=False),
+    sa.Column('delivery_time', sa.DateTime(), nullable=True),
+    sa.Column('start_time', sa.DateTime(), nullable=True),
+    sa.Column('end_time', sa.DateTime(), nullable=True),
+    sa.ForeignKeyConstraint(['experiment_id'], ['experiment.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    op.create_table('metric',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('run_id', sa.Integer(), nullable=False),
+    sa.Column('split', sa.Enum('TRAIN', 'VALIDATION', 'TEST', name='splitenum'), nullable=False),
+    sa.Column('level', sa.Enum('LAST', 'TRIAL', 'STEP', 'EPOCH', name='levelenum'), nullable=False),
+    sa.Column('name', sa.String(), nullable=False),
+    sa.Column('value', sa.Float(), nullable=False),
+    sa.Column('step', sa.Integer(), nullable=False),
+    sa.Column('timestamp', sa.DateTime(), nullable=False),
+    sa.ForeignKeyConstraint(['run_id'], ['run.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    with op.batch_alter_table('metric', schema=None) as batch_op:
+        batch_op.create_index(batch_op.f('ix_metric_run_id'), ['run_id'], unique=False)
+        batch_op.create_index(batch_op.f('ix_metric_timestamp'), ['timestamp'], unique=False)
+
+    op.create_table('prediction',
+    sa.Column('id', sa.Integer(), nullable=False),
+    sa.Column('run_id', sa.Integer(), nullable=False),
+    sa.Column('dataset_id', sa.Integer(), nullable=True),
+    sa.Column('huey_id', sa.String(), nullable=True),
+    sa.Column('created', sa.DateTime(), nullable=False),
+    sa.Column('last_modified', sa.DateTime(), nullable=False),
+    sa.Column('results_path', sa.String(), nullable=True),
+    sa.Column('status', sa.Enum('NOT_STARTED', 'DELIVERED', 'STARTED', 'FINISHED', 'ERROR', name='predictionstatus'), nullable=False),
+    sa.Column('delivery_time', sa.DateTime(), nullable=True),
+    sa.Column('start_time', sa.DateTime(), nullable=True),
+    sa.Column('end_time', sa.DateTime(), nullable=True),
+    sa.ForeignKeyConstraint(['dataset_id'], ['dataset.id'], ),
+    sa.ForeignKeyConstraint(['run_id'], ['run.id'], ondelete='CASCADE'),
+    sa.PrimaryKeyConstraint('id')
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_table('prediction')
+    with op.batch_alter_table('metric', schema=None) as batch_op:
+        batch_op.drop_index(batch_op.f('ix_metric_timestamp'))
+        batch_op.drop_index(batch_op.f('ix_metric_run_id'))
+
+    op.drop_table('metric')
+    op.drop_table('run')
+    op.drop_table('process_data')
+    op.drop_table('explorer')
+    op.drop_table('converter_list')
+    op.drop_table('tag')
+    op.drop_table('parameter_history')
+    op.drop_table('notebook')
+    op.drop_table('generative_process')
+    op.drop_table('experiment')
+    op.drop_table('plugin')
+    op.drop_table('pipeline')
+    op.drop_table('local_explainer')
+    op.drop_table('global_explainer')
+    op.drop_table('generative_session')
+    op.drop_table('dataset')
+    # ### end Alembic commands ###
diff --git a/DashAI/back/app.py b/DashAI/back/app.py
index 3164801bb..71cc377dd 100644
--- a/DashAI/back/app.py
+++ b/DashAI/back/app.py
@@ -13,7 +13,7 @@
 from DashAI.back.api.front_api import router as app_router
 from DashAI.back.container import build_container
 from DashAI.back.dependencies.config_builder import build_config_dict
-from DashAI.back.dependencies.database.models import Base
+from DashAI.back.dependencies.database.migrate import migrate_on_startup
 
 logger = logging.getLogger(__name__)
 
@@ -81,15 +81,16 @@ def create_app(
     logger.debug("3. Creating app container and setting up dependency injection.")
     container = build_container(config=config)
 
-    logger.debug("5. Creating database.")
-    Base.metadata.create_all(bind=container["engine"])
-
-    logger.debug("6. Initializing FastAPI application.")
+    logger.debug("4. Applying database migrations.")
+    migrate_on_startup(
+        sqlite_file_path=pathlib.Path(config["SQLITE_DB_PATH"]),
+    )
+    logger.debug("5. Initializing FastAPI application.")
     app = FastAPI(title="DashAI")
     api_v0 = FastAPI(title="DashAI API v0")
     api_v1 = FastAPI(title="DashAI API v1")
 
-    logger.debug("7. Mounting API routers.")
+    logger.debug("6. Mounting API routers.")
     api_v0.include_router(api_router_v0)
     api_v1.include_router(api_router_v1)
 
diff --git a/DashAI/back/dependencies/database/migrate.py b/DashAI/back/dependencies/database/migrate.py
new file mode 100644
index 000000000..ffcbda073
--- /dev/null
+++ b/DashAI/back/dependencies/database/migrate.py
@@ -0,0 +1,66 @@
+import logging
+import os
+import shutil
+import time
+from pathlib import Path
+
+from alembic import command
+from alembic.config import Config
+
+from DashAI.back.dependencies.database.utils import resolve_db_url
+
+logger = logging.getLogger(__name__)
+
+
+def alembic_config(db_url: str) -> Config:
+    package_root = Path(__file__).absolute().parents[3]
+    script_location = package_root / "alembic"
+
+    cfg = Config()
+
+    cfg.set_main_option("script_location", str(script_location))
+    cfg.set_main_option("sqlalchemy.url", db_url)
+    cfg.set_main_option("prepend_sys_path", str(package_root))
+    cfg.set_main_option("version_path_separator", "os")
+    return cfg
+
+
+def run_migrations(db_url: str) -> None:
+    cfg = alembic_config(db_url=db_url)
+    command.upgrade(cfg, "head")
+
+
+def backup_and_recreate_db(db_url: str, sqlite_file_path: Path) -> None:
+    # Only try to backup if it's a local SQLite file and DATABASE_URL env var not set
+    env_url = os.getenv("DATABASE_URL")
+    if not env_url and sqlite_file_path.exists():
+        ts = time.strftime("%Y%m%d-%H%M%S")
+        backup = sqlite_file_path.with_suffix(f".bak-{ts}.sqlite3")
+        shutil.copy2(sqlite_file_path, backup)
+        sqlite_file_path.unlink()
+
+    run_migrations(db_url=db_url)
+
+
+def migrate_on_startup(sqlite_file_path: Path) -> None:
+    db_url = resolve_db_url(sqlite_file_path)
+
+    try:
+        logger.info(f"Running migrations on database at {db_url}")
+        run_migrations(db_url=db_url)
+    except Exception as exc:
+        logger.error(
+            (
+                f"Error during migration: {exc}. "
+                "Attempting to backup and recreate the database."
+            )
+        )
+        try:
+            backup_and_recreate_db(db_url=db_url, sqlite_file_path=sqlite_file_path)
+        except Exception as backup_exc:
+            logger.error(
+                f"Error during backup and recreate: {backup_exc}. "
+                f"Original migration error: {exc}."
+            )
+            raise backup_exc from exc
+        raise exc
diff --git a/DashAI/back/dependencies/database/models.py b/DashAI/back/dependencies/database/models.py
index 2f564f079..84433a742 100644
--- a/DashAI/back/dependencies/database/models.py
+++ b/DashAI/back/dependencies/database/models.py
@@ -11,6 +11,7 @@
     Float,
     ForeignKey,
     Integer,
+    MetaData,
     String,
 )
 from sqlalchemy.ext.declarative import declarative_base
@@ -31,7 +32,16 @@
 logger = logging.getLogger(__name__)
 
 
-Base = declarative_base()
+naming_convention = {
+    "ix": "ix_%(table_name)s_%(column_0_name)s",
+    "uq": "uq_%(table_name)s_%(column_0_name)s",
+    "ck": "ck_%(constraint_name)s",
+    "fk": "fk_%(table_name)s_%(column_0_name)s_%(referred_table_name)s",
+    "pk": "pk_%(table_name)s",
+}
+
+metadata = MetaData(naming_convention=naming_convention)
+Base = declarative_base(metadata=metadata)
 
 
 class Dataset(Base):
diff --git a/DashAI/back/dependencies/database/sqlite_database.py b/DashAI/back/dependencies/database/sqlite_database.py
index 6601f72de..e450a184f 100644
--- a/DashAI/back/dependencies/database/sqlite_database.py
+++ b/DashAI/back/dependencies/database/sqlite_database.py
@@ -7,6 +7,8 @@
 from sqlalchemy.engine import Engine
 from sqlalchemy.orm import sessionmaker
 
+from DashAI.back.dependencies.database.utils import resolve_db_url
+
 logger = logging.getLogger(__name__)
 
 
@@ -32,8 +34,7 @@ def setup_sqlite_db(config: Dict[str, str]) -> Tuple[Engine, sessionmaker]:
                 for creating database sessions.
     """
 
-    if not str(config["SQLITE_DB_PATH"]).startswith("sqlite:///"):
-        db_url = "sqlite:///" + str(config["SQLITE_DB_PATH"])
+    db_url = resolve_db_url(config["SQLITE_DB_PATH"])
 
     logger.info("Using %s as SQLite path.", db_url)
 
diff --git a/DashAI/back/dependencies/database/utils.py b/DashAI/back/dependencies/database/utils.py
index e9bbbd991..5fe4b3d44 100644
--- a/DashAI/back/dependencies/database/utils.py
+++ b/DashAI/back/dependencies/database/utils.py
@@ -1,4 +1,6 @@
 import logging
+import os
+from pathlib import Path
 
 from fastapi import Depends
 from kink import di, inject
@@ -276,3 +278,23 @@ def find_entity_by_huey_id(huey_id: str) -> dict:
             }
 
         return None
+
+
+def resolve_db_url(sqlite_file_path: Path) -> str:
+    """
+    Resolve database URL with the same priority as alembic:
+      1) env var DATABASE_URL
+      2) fallback to provided sqlite file path
+
+    This is mainly to use in tests, where we set the env var to a temp path.
+    """
+    # 1) environment variable
+    env_url = os.getenv("DATABASE_URL")
+    if env_url:
+        return env_url
+
+    if not str(sqlite_file_path).startswith("sqlite:///"):
+        return f"sqlite:///{sqlite_file_path}"
+
+    # 2) fallback to provided sqlite file path
+    return sqlite_file_path
diff --git a/MANIFEST.in b/MANIFEST.in
index 51353ff30..a649ed52f 100644
--- a/MANIFEST.in
+++ b/MANIFEST.in
@@ -1,8 +1,11 @@
 include requirements.txt
 include requirements-dev.txt
+include DashAI/alembic.ini
+recursive-include DashAI/alembic *
 recursive-include DashAI/front/build *
 recursive-include DashAI/back/models/parameters/models_schemas *
 recursive-include DashAI/back/dataloaders/dataloaders_schemas *
 recursive-include DashAI/back/dataloaders/description_schemas *
 recursive-include DashAI/back/dataloaders/params_schemas *
+recursive-include DashAI/back/types *
 prune tests/
diff --git a/README.rst b/README.rst
index be6b32dae..1e2169a1a 100644
--- a/README.rst
+++ b/README.rst
@@ -199,6 +199,97 @@ You can check all available options through the command:
     $ python -m DashAI --help
 
 
+Database Migrations
+==================
+
+Migrations are managed through `Alembic <https://alembic.sqlalchemy.org/en/latest/>`_.
+
+They are automatically executed when starting DashAI. However, if you want to
+run them manually, you can do so using the following command (inside the
+`DashAI/` folder):
+
+.. code-block:: bash
+
+    $ alembic upgrade head
+
+This command applies all pending migrations up to the latest revision.
+
+---
+
+Creating a New Migration
+------------------------
+
+After modifying the database models, a new migration can be generated using:
+
+.. code-block:: bash
+
+    $ alembic revision --autogenerate -m "<<Your message here>>"
+
+Where ``<<Your message here>>`` is a brief description of the changes introduced
+(e.g., *add model metadata table*, *update dataset schema*).
+
+Generated migrations are located in the ``alembic/versions`` directory and
+**must be committed to the repository**.
+
+It is strongly recommended to review the autogenerated migration file before
+applying it, as Alembic may not always detect complex changes correctly.
+
+---
+
+Applying Migrations
+-------------------
+
+To apply all pending migrations:
+
+.. code-block:: bash
+
+    $ alembic upgrade head
+
+To upgrade to a specific revision:
+
+.. code-block:: bash
+
+    $ alembic upgrade <revision_id>
+
+---
+
+Downgrading Migrations
+----------------------
+
+If you need to revert database changes, migrations can be downgraded using:
+
+.. code-block:: bash
+
+    $ alembic downgrade -1
+
+This command reverts the last applied migration.
+
+To downgrade to a specific revision:
+
+.. code-block:: bash
+
+    $ alembic downgrade <revision_id>
+
+---
+
+Checking Migration Status
+-------------------------
+
+To view the current migration applied to the database:
+
+.. code-block:: bash
+
+    $ alembic current
+
+To list the full migration history:
+
+.. code-block:: bash
+
+    $ alembic history
+
+---
+
+
 Testing
 =======
 
diff --git a/alembic.ini b/alembic.ini
index 8ee69ae36..e71144cf7 100644
--- a/alembic.ini
+++ b/alembic.ini
@@ -2,7 +2,7 @@
 
 [alembic]
 # path to migration scripts
-script_location = alembic
+script_location = DashAI/alembic
 
 # template used to generate migration file names; The default value is %%(rev)s_%%(slug)s
 # Uncomment the line below if you want the files to be prepended with date and time
@@ -60,7 +60,9 @@ version_path_separator = os  # Use os.pathsep. Default configuration used for ne
 # are written from script.py.mako
 # output_encoding = utf-8
 
-sqlalchemy.url = sqlite:///DashAI/back/database/DashAI.sqlite
+# SQLite URL points to the same database used by the application
+# Note: This will be overridden by environment variables or CLI arguments if provided
+sqlalchemy.url = sqlite:///
 
 
 [post_write_hooks]
diff --git a/alembic/env.py b/alembic/env.py
deleted file mode 100644
index 7907b3085..000000000
--- a/alembic/env.py
+++ /dev/null
@@ -1,67 +0,0 @@
-from logging.config import fileConfig
-
-from sqlalchemy import engine_from_config, pool
-
-from alembic import context
-from DashAI.back.database.models import Base
-
-# this is the Alembic Config object, which provides
-# access to the values within the .ini file in use.
-config = context.config
-
-# Interpret the config file for Python logging.
-# This line sets up loggers basically.
-if config.config_file_name is not None:
-    fileConfig(config.config_file_name)
-
-target_metadata = Base.metadata
-
-
-def run_migrations_offline() -> None:
-    """Run migrations in 'offline' mode.
-
-    This configures the context with just a URL
-    and not an Engine, though an Engine is acceptable
-    here as well.  By skipping the Engine creation
-    we don't even need a DBAPI to be available.
-
-    Calls to context.execute() here emit the given string to the
-    script output.
-
-    """
-    url = config.get_main_option("sqlalchemy.url")
-    context.configure(
-        url=url,
-        target_metadata=target_metadata,
-        literal_binds=True,
-        dialect_opts={"paramstyle": "named"},
-    )
-
-    with context.begin_transaction():
-        context.run_migrations()
-
-
-def run_migrations_online() -> None:
-    """Run migrations in 'online' mode.
-
-    In this scenario we need to create an Engine
-    and associate a connection with the context.
-
-    """
-    connectable = engine_from_config(
-        config.get_section(config.config_ini_section, {}),
-        prefix="sqlalchemy.",
-        poolclass=pool.NullPool,
-    )
-
-    with connectable.connect() as connection:
-        context.configure(connection=connection, target_metadata=target_metadata)
-
-        with context.begin_transaction():
-            context.run_migrations()
-
-
-if context.is_offline_mode():
-    run_migrations_offline()
-else:
-    run_migrations_online()
diff --git a/alembic/versions/2f997776b9a1_add_splits_metadata_to_experiment_model_.py b/alembic/versions/2f997776b9a1_add_splits_metadata_to_experiment_model_.py
deleted file mode 100644
index cb96a15fe..000000000
--- a/alembic/versions/2f997776b9a1_add_splits_metadata_to_experiment_model_.py
+++ /dev/null
@@ -1,25 +0,0 @@
-"""Add splits metadata to Experiment model on DB
-
-Revision ID: 2f997776b9a1
-Revises: 7ba93a3d42a8
-Create Date: 2023-10-09 12:34:12.786244
-
-"""
-
-# revision identifiers, used by Alembic.
-revision = "2f997776b9a1"
-down_revision = "7ba93a3d42a8"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    pass
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    pass
-    # ### end Alembic commands ###
diff --git a/alembic/versions/4bce47006c88_add_splits_metadata_to_experiment_model_.py b/alembic/versions/4bce47006c88_add_splits_metadata_to_experiment_model_.py
deleted file mode 100644
index c8c7a000f..000000000
--- a/alembic/versions/4bce47006c88_add_splits_metadata_to_experiment_model_.py
+++ /dev/null
@@ -1,28 +0,0 @@
-"""Add splits metadata to Experiment model on DB
-
-Revision ID: 4bce47006c88
-Revises: 2f997776b9a1
-Create Date: 2023-10-09 12:57:13.075246
-
-"""
-import sqlalchemy as sa
-
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "4bce47006c88"
-down_revision = "2f997776b9a1"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column("experiment", sa.Column("splits", sa.JSON(), nullable=False))
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_column("experiment", "splits")
-    # ### end Alembic commands ###
diff --git a/alembic/versions/649c500185b8_delete_task_name_and_adds_dataloader_.py b/alembic/versions/649c500185b8_delete_task_name_and_adds_dataloader_.py
deleted file mode 100644
index dd781e15c..000000000
--- a/alembic/versions/649c500185b8_delete_task_name_and_adds_dataloader_.py
+++ /dev/null
@@ -1,32 +0,0 @@
-"""Delete task_name and adds dataloader name to Dataset model
-
-Revision ID: 649c500185b8
-Revises: 80faf11605a2
-Create Date: 2023-08-09 08:43:29.112965
-
-"""
-import sqlalchemy as sa
-
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "649c500185b8"
-down_revision = "80faf11605a2"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column("dataset", sa.Column("feature_names", sa.JSON(), nullable=False))
-    op.add_column("dataset", sa.Column("dataset_name", sa.String(), nullable=False))
-    op.drop_column("dataset", "task_name")
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column("dataset", sa.Column("task_name", sa.VARCHAR(), nullable=False))
-    op.drop_column("dataset", "dataset_name")
-    op.drop_column("dataset", "feature_names")
-    # ### end Alembic commands ###
diff --git a/alembic/versions/7ba93a3d42a8_delete_feature_names_to_experiment_.py b/alembic/versions/7ba93a3d42a8_delete_feature_names_to_experiment_.py
deleted file mode 100644
index 7ddac412d..000000000
--- a/alembic/versions/7ba93a3d42a8_delete_feature_names_to_experiment_.py
+++ /dev/null
@@ -1,31 +0,0 @@
-"""Delete feature_names to Experiment model on DB
-
-Revision ID: 7ba93a3d42a8
-Revises: 7e990d4a75ec
-Create Date: 2023-09-25 10:03:29.526284
-
-"""
-import sqlalchemy as sa
-from sqlalchemy.dialects import sqlite
-
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "7ba93a3d42a8"
-down_revision = "7e990d4a75ec"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_column("experiment", "feature_names")
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column(
-        "experiment", sa.Column("feature_names", sqlite.JSON(), nullable=False)
-    )
-    # ### end Alembic commands ###
diff --git a/alembic/versions/7e990d4a75ec_add_metadata_to_experiment_model_on_db.py b/alembic/versions/7e990d4a75ec_add_metadata_to_experiment_model_on_db.py
deleted file mode 100644
index 216084ee6..000000000
--- a/alembic/versions/7e990d4a75ec_add_metadata_to_experiment_model_on_db.py
+++ /dev/null
@@ -1,30 +0,0 @@
-"""Add metadata to Experiment model on DB
-
-Revision ID: 7e990d4a75ec
-Revises: 8307d4c14824
-Create Date: 2023-09-22 22:45:34.448464
-
-"""
-import sqlalchemy as sa
-
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "7e990d4a75ec"
-down_revision = "8307d4c14824"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column("experiment", sa.Column("input_columns", sa.JSON(), nullable=False))
-    op.add_column("experiment", sa.Column("output_columns", sa.JSON(), nullable=False))
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_column("experiment", "output_columns")
-    op.drop_column("experiment", "input_columns")
-    # ### end Alembic commands ###
diff --git a/alembic/versions/80faf11605a2_initial_migration.py b/alembic/versions/80faf11605a2_initial_migration.py
deleted file mode 100644
index d3da13249..000000000
--- a/alembic/versions/80faf11605a2_initial_migration.py
+++ /dev/null
@@ -1,30 +0,0 @@
-"""Initial migration
-
-Revision ID: 80faf11605a2
-Revises:
-Create Date: 2023-08-08 09:13:58.794747
-
-"""
-import sqlalchemy as sa
-
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "80faf11605a2"
-down_revision = None
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_column("experiment", "step")
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column(
-        "experiment", sa.Column("step", sa.VARCHAR(length=19), nullable=False)
-    )
-    # ### end Alembic commands ###
diff --git a/alembic/versions/8307d4c14824_add_metadata_to_experiment_model_on_db.py b/alembic/versions/8307d4c14824_add_metadata_to_experiment_model_on_db.py
deleted file mode 100644
index 39c1b41d7..000000000
--- a/alembic/versions/8307d4c14824_add_metadata_to_experiment_model_on_db.py
+++ /dev/null
@@ -1,28 +0,0 @@
-"""Add metadata to Experiment model on DB
-
-Revision ID: 8307d4c14824
-Revises: d267eb740e0f
-Create Date: 2023-09-22 21:46:37.521272
-
-"""
-import sqlalchemy as sa
-
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "8307d4c14824"
-down_revision = "d267eb740e0f"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column("experiment", sa.Column("feature_names", sa.JSON(), nullable=False))
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_column("experiment", "feature_names")
-    # ### end Alembic commands ###
diff --git a/alembic/versions/d267eb740e0f_delete_dataloader_name_in_dataset_model.py b/alembic/versions/d267eb740e0f_delete_dataloader_name_in_dataset_model.py
deleted file mode 100644
index 13ef6cc77..000000000
--- a/alembic/versions/d267eb740e0f_delete_dataloader_name_in_dataset_model.py
+++ /dev/null
@@ -1,28 +0,0 @@
-"""Delete dataloader name in Dataset model
-
-Revision ID: d267eb740e0f
-Revises: 649c500185b8
-Create Date: 2023-08-09 20:45:43.229983
-
-"""
-import sqlalchemy as sa
-
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "d267eb740e0f"
-down_revision = "649c500185b8"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_column("dataset", "dataset_name")
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column("dataset", sa.Column("dataset_name", sa.VARCHAR(), nullable=False))
-    # ### end Alembic commands ###
diff --git a/alembic/versions/e40a1eeee0ef_delete_feature_names_from_datasets.py b/alembic/versions/e40a1eeee0ef_delete_feature_names_from_datasets.py
deleted file mode 100644
index ac89ee989..000000000
--- a/alembic/versions/e40a1eeee0ef_delete_feature_names_from_datasets.py
+++ /dev/null
@@ -1,29 +0,0 @@
-"""Delete feature_names from datasets
-
-Revision ID: e40a1eeee0ef
-Revises: 4bce47006c88
-Create Date: 2023-11-28 15:58:18.913263
-
-"""
-import sqlalchemy as sa
-from sqlalchemy.dialects import sqlite
-
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "e40a1eeee0ef"
-down_revision = "4bce47006c88"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_column("dataset", "feature_names")
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.add_column("dataset", sa.Column("feature_names", sqlite.JSON(), nullable=False))
-    # ### end Alembic commands ###
diff --git a/scripts/ci_alembic_check.py b/scripts/ci_alembic_check.py
new file mode 100644
index 000000000..84dbdd7bc
--- /dev/null
+++ b/scripts/ci_alembic_check.py
@@ -0,0 +1,34 @@
+#!/usr/bin/env python3
+import os
+import sys
+
+from alembic.autogenerate import compare_metadata
+from alembic.migration import MigrationContext
+from sqlalchemy import create_engine
+
+try:
+    from DashAI.back.dependencies.database.models import Base
+
+    metadata = Base.metadata
+except Exception as e:
+    print("ERROR importing project metadata:", e, file=sys.stderr)
+    sys.exit(2)
+
+database_url = os.environ.get("DATABASE_URL")
+if not database_url:
+    print("DATABASE_URL is not defined", file=sys.stderr)
+    sys.exit(2)
+
+engine = create_engine(database_url)
+with engine.connect() as conn:
+    ctx = MigrationContext.configure(conn)
+    diffs = compare_metadata(ctx, metadata)
+
+if diffs:
+    print("¡Detected schema drift! Autogenerate produced changes:")
+    for d in diffs:
+        print(d)
+    sys.exit(1)
+
+print("No pending schema changes detected.")
+sys.exit(0)