Initial public ModelForge release

2026-09-01 21:30:16 +02:00
commit 7082ab955a
490 changed files with 104252 additions and 0 deletions
@@ -0,0 +1,74 @@
+"""M9 multimodal capability evaluation foundation.
+
+Revision ID: 20260826_0013
+Revises: 20260825_0012
+"""
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+from alembic import op
+
+revision: str = "20260826_0013"
+down_revision: str | None = "20260825_0012"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "capability_evaluation_suites",
+        sa.Column("id", sa.Uuid(), nullable=False),
+        sa.Column("capability_contract_id", sa.Uuid(), nullable=False),
+        sa.Column("key", sa.String(length=128), nullable=False),
+        sa.Column("evaluation_type", sa.String(length=32), nullable=False),
+        sa.Column("revision", sa.String(length=128), nullable=False),
+        sa.Column("dataset_revision", sa.String(length=128), nullable=False),
+        sa.Column("definition_digest", sa.String(length=64), nullable=False),
+        sa.Column("metric_definitions", sa.JSON(), nullable=False),
+        sa.Column("case_definitions", sa.JSON(), nullable=False),
+        sa.Column("thresholds", sa.JSON(), nullable=False),
+        sa.Column("created_at", sa.DateTime(timezone=True), server_default=sa.func.now(), nullable=False),
+        sa.Column("immutable_at", sa.DateTime(timezone=True), server_default=sa.func.now(), nullable=False),
+        sa.ForeignKeyConstraint(["capability_contract_id"], ["capability_contracts.id"], ondelete="RESTRICT"),
+        sa.PrimaryKeyConstraint("id"),
+        sa.UniqueConstraint("capability_contract_id", "key", "revision", name="uq_capability_eval_suite"),
+        sa.UniqueConstraint("definition_digest", name="uq_capability_eval_definition_digest"),
+    )
+    op.create_index("ix_capability_evaluation_suites_capability_contract_id", "capability_evaluation_suites", ["capability_contract_id"])
+    op.create_index("ix_capability_evaluation_suites_evaluation_type", "capability_evaluation_suites", ["evaluation_type"])
+    op.create_table(
+        "capability_evaluation_runs",
+        sa.Column("id", sa.Uuid(), nullable=False),
+        sa.Column("suite_id", sa.Uuid(), nullable=False),
+        sa.Column("capability_deployment_id", sa.Uuid(), nullable=False),
+        sa.Column("status", sa.String(length=32), nullable=False),
+        sa.Column("metric_values", sa.JSON(), nullable=False),
+        sa.Column("case_results", sa.JSON(), nullable=False),
+        sa.Column("resource_metrics", sa.JSON(), nullable=False),
+        sa.Column("environment_fingerprint", sa.String(length=64), nullable=False),
+        sa.Column("evidence_digest", sa.String(length=64), nullable=False),
+        sa.Column("evidence", sa.JSON(), nullable=False),
+        sa.Column("started_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("completed_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("created_at", sa.DateTime(timezone=True), server_default=sa.func.now(), nullable=False),
+        sa.ForeignKeyConstraint(["suite_id"], ["capability_evaluation_suites.id"], ondelete="RESTRICT"),
+        sa.ForeignKeyConstraint(["capability_deployment_id"], ["capability_deployments.id"], ondelete="RESTRICT"),
+        sa.PrimaryKeyConstraint("id"),
+        sa.UniqueConstraint("evidence_digest", name="uq_capability_eval_evidence"),
+    )
+    op.create_index("ix_capability_evaluation_runs_suite_id", "capability_evaluation_runs", ["suite_id"])
+    op.create_index("ix_capability_evaluation_runs_capability_deployment_id", "capability_evaluation_runs", ["capability_deployment_id"])
+    op.create_index("ix_capability_evaluation_runs_status", "capability_evaluation_runs", ["status"])
+    op.create_index("ix_capability_evaluation_runs_environment_fingerprint", "capability_evaluation_runs", ["environment_fingerprint"])
+
+
+def downgrade() -> None:
+    op.drop_index("ix_capability_evaluation_runs_environment_fingerprint", table_name="capability_evaluation_runs")
+    op.drop_index("ix_capability_evaluation_runs_status", table_name="capability_evaluation_runs")
+    op.drop_index("ix_capability_evaluation_runs_capability_deployment_id", table_name="capability_evaluation_runs")
+    op.drop_index("ix_capability_evaluation_runs_suite_id", table_name="capability_evaluation_runs")
+    op.drop_table("capability_evaluation_runs")
+    op.drop_index("ix_capability_evaluation_suites_evaluation_type", table_name="capability_evaluation_suites")
+    op.drop_index("ix_capability_evaluation_suites_capability_contract_id", table_name="capability_evaluation_suites")
+    op.drop_table("capability_evaluation_suites")