projectdavid-orm 1.3.0__tar.gz → 1.4.0__tar.gz

{projectdavid_orm-1.3.0 → projectdavid_orm-1.4.0}/CHANGELOG.md

@@ -1,3 +1,10 @@
+# [1.4.0](https://github.com/project-david-ai/projectdavid-orm/compare/v1.3.0...v1.4.0) (2026-03-23)
+
+
+### Features
+
+* Phase 5a ([7213c69](https://github.com/project-david-ai/projectdavid-orm/commit/7213c698beb56eddb5230bbcf609f250e6b20c5e))
+
 # [1.3.0](https://github.com/project-david-ai/projectdavid-orm/compare/v1.2.0...v1.3.0) (2026-03-23)
 
 

{projectdavid_orm-1.3.0/src/projectdavid_orm.egg-info → projectdavid_orm-1.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: projectdavid-orm
-Version: 1.3.0
+Version: 1.4.0
 Summary: Shared SQLAlchemy ORM models for ProjectDavid services
 Author-email: "Francis N." <francis.neequaye@projectdavid.co.uk>
 License: Polyform

{projectdavid_orm-1.3.0 → projectdavid_orm-1.4.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "projectdavid-orm"
-version = "1.3.0"
+version = "1.4.0"
 description = "Shared SQLAlchemy ORM models for ProjectDavid services"
 readme = "README.md"
 authors = [

{projectdavid_orm-1.3.0 → projectdavid_orm-1.4.0}/src/projectdavid_orm/projectdavid_orm/models.py

@@ -821,8 +821,7 @@ class TrainingJob(Base):
     output_path = Column(String(512), nullable=True)
 
     # Phase 4: node_id stores the Ray node ID (hex string) for traceability.
-    # FK to compute_nodes retained for legacy compatibility — will be dropped
-    # in Phase 5 when compute_nodes table is removed entirely.
+    # FK to compute_nodes retained for legacy compatibility — dropped in Phase 5.
     node_id = Column(String(64), ForeignKey("compute_nodes.id", ondelete="SET NULL"), nullable=True)
 
     dataset = relationship("Dataset", back_populates="training_jobs", lazy="select")
@@ -881,8 +880,8 @@ class ComputeNode(Base):
     """
     Phase 4 status: LEGACY.
     Node registration and heartbeat are now handled by Ray natively.
-    This table is retained for FK compatibility with TrainingJob and
-    FineTunedModel. Will be dropped in Phase 5 cleanup.
+    Retained for FK compatibility with TrainingJob and FineTunedModel.
+    Will be dropped in Phase 5 cleanup.
     """
 
     __tablename__ = "compute_nodes"
@@ -898,15 +897,13 @@ class ComputeNode(Base):
     allocations = relationship("GPUAllocation", back_populates="node", cascade="all, delete-orphan")
     training_jobs = relationship("TrainingJob", back_populates="node")
     active_models = relationship("FineTunedModel", back_populates="node")
-    # Phase 4: InferenceDeployment no longer has a FK back to compute_nodes
-    # so the deployments relationship is removed from this side too.
 
 
 class GPUAllocation(Base):
     """
     Phase 4 status: LEGACY.
     VRAM reservation is now handled by Ray resource scheduling.
-    Table retained for schema compatibility. Will be dropped in Phase 5.
+    Retained for schema compatibility. Will be dropped in Phase 5.
     """
 
     __tablename__ = "gpu_allocations"
@@ -941,9 +938,7 @@ class InferenceDeployment(Base):
     __tablename__ = "inference_deployments"
     id = Column(String(64), primary_key=True, index=True)
 
-    # Phase 4: node_id is now a plain string storing the Ray node ID (hex).
-    # FK to compute_nodes and uq_node_port_deployment unique constraint have
-    # been dropped from the DB — this column is no longer constrained.
+    # Phase 4: plain string storing the Ray node ID (hex). No FK constraint.
     node_id = Column(String(128), nullable=True)
 
     internal_hostname = Column(String(128), nullable=True)
@@ -954,6 +949,10 @@ class InferenceDeployment(Base):
     current_throughput = Column(Float, default=0.0)
     last_seen = Column(BigInteger, default=lambda: int(time.time()))
 
-    # Phase 4: node relationship removed — node_id is no longer a FK.
+    # Inference sharding: number of GPUs this deployment spans.
+    # 1 = single GPU (default, backward compatible).
+    # N > 1 = tensor parallel across N GPUs via vLLM --tensor-parallel-size.
+    tensor_parallel_size = Column(Integer, nullable=False, default=1)
+
     base_model = relationship("BaseModel")
     fine_tuned_model = relationship("FineTunedModel")

{projectdavid_orm-1.3.0 → projectdavid_orm-1.4.0/src/projectdavid_orm.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: projectdavid-orm
-Version: 1.3.0
+Version: 1.4.0
 Summary: Shared SQLAlchemy ORM models for ProjectDavid services
 Author-email: "Francis N." <francis.neequaye@projectdavid.co.uk>
 License: Polyform