isa-model 0.3.9__py3-none-any.whl → 0.4.0__py3-none-any.whl

isa_model/serving/api/routes/deployments.py

@@ -0,0 +1,339 @@
+"""
+Deployments API Routes
+
+Handles automated HuggingFace model deployment to Modal
+"""
+
+from fastapi import APIRouter, HTTPException, BackgroundTasks
+from pydantic import BaseModel
+from typing import Optional, List, Dict, Any
+import logging
+import asyncio
+import json
+from pathlib import Path
+
+from isa_model.deployment.services.auto_hf_modal_deployer import HuggingFaceModalDeployer
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter()
+
+# Request/Response models
+class AnalyzeModelRequest(BaseModel):
+    model_id: str
+
+class DeployModelRequest(BaseModel):
+    model_id: str
+    service_name: Optional[str] = None
+    auto_deploy: bool = False
+
+class DeploymentResponse(BaseModel):
+    success: bool
+    deployment_id: Optional[str] = None
+    model_id: str
+    config: Optional[Dict[str, Any]] = None
+    service_file: Optional[str] = None
+    deployment_command: Optional[str] = None
+    estimated_cost_per_hour: Optional[float] = None
+    deployed: bool = False
+    error: Optional[str] = None
+
+# Global deployer instance
+deployer = HuggingFaceModalDeployer()
+
+# In-memory deployment tracking (in production, use a database)
+deployments = {}
+
+@router.post("/analyze", response_model=Dict[str, Any])
+async def analyze_model(request: AnalyzeModelRequest):
+    """
+    Analyze a HuggingFace model for deployment compatibility
+    """
+    try:
+        logger.info(f"Analyzing model: {request.model_id}")
+        
+        # Analyze the model
+        config = deployer.analyze_model(request.model_id)
+        
+        return {
+            "success": True,
+            "model_id": config.model_id,
+            "model_type": config.model_type,
+            "architecture": config.architecture,
+            "parameters": config.parameters,
+            "gpu_requirements": config.gpu_requirements,
+            "memory_gb": config.memory_gb,
+            "container_memory_mb": config.container_memory_mb,
+            "dependencies": config.dependencies,
+            "capabilities": config.capabilities,
+            "estimated_cost_per_hour": config.estimated_cost_per_hour
+        }
+        
+    except Exception as e:
+        logger.error(f"Model analysis failed for {request.model_id}: {e}")
+        raise HTTPException(status_code=400, detail=f"Model analysis failed: {str(e)}")
+
+@router.post("/deploy", response_model=DeploymentResponse)
+async def deploy_model(request: DeployModelRequest, background_tasks: BackgroundTasks):
+    """
+    Deploy a HuggingFace model to Modal
+    """
+    try:
+        logger.info(f"Starting deployment for model: {request.model_id}")
+        
+        # Generate unique deployment ID
+        import time
+        import uuid
+        deployment_id = f"deploy_{uuid.uuid4().hex[:8]}_{int(time.time())}"
+        
+        # Add to deployments tracking
+        deployments[deployment_id] = {
+            "id": deployment_id,
+            "model_id": request.model_id,
+            "service_name": request.service_name,
+            "status": "pending",
+            "created_at": time.time(),
+            "auto_deploy": request.auto_deploy
+        }
+        
+        # Start deployment in background
+        background_tasks.add_task(
+            perform_deployment, 
+            deployment_id, 
+            request.model_id, 
+            request.service_name,
+            request.auto_deploy
+        )
+        
+        return DeploymentResponse(
+            success=True,
+            deployment_id=deployment_id,
+            model_id=request.model_id,
+            deployed=False
+        )
+        
+    except Exception as e:
+        logger.error(f"Deployment initiation failed for {request.model_id}: {e}")
+        raise HTTPException(status_code=500, detail=f"Deployment failed: {str(e)}")
+
+async def perform_deployment(deployment_id: str, model_id: str, service_name: Optional[str], auto_deploy: bool):
+    """
+    Perform the actual deployment in the background
+    """
+    import time
+    
+    try:
+        logger.info(f"Performing deployment {deployment_id} for model {model_id}")
+        
+        # Update status
+        deployments[deployment_id]["status"] = "deploying"
+        deployments[deployment_id]["progress"] = "Analyzing model"
+        
+        # Deploy the model
+        result = deployer.deploy_model(model_id, deploy=auto_deploy)
+        
+        if result["success"]:
+            deployments[deployment_id].update({
+                "status": "completed" if result.get("deployed") else "generated",
+                "progress": "Deployment completed",
+                "config": result["config"],
+                "service_file": result["service_file"],
+                "deployment_command": result["deployment_command"],
+                "estimated_cost_per_hour": result["estimated_cost_per_hour"],
+                "deployed": result.get("deployed", False),
+                "completed_at": time.time()
+            })
+        else:
+            deployments[deployment_id].update({
+                "status": "failed",
+                "progress": "Deployment failed",
+                "error": result.get("error", "Unknown error"),
+                "failed_at": time.time()
+            })
+            
+    except Exception as e:
+        logger.error(f"Deployment {deployment_id} failed: {e}")
+        deployments[deployment_id].update({
+            "status": "failed",
+            "progress": "Deployment failed",
+            "error": str(e),
+            "failed_at": time.time()
+        })
+
+@router.get("/")
+async def list_deployments():
+    """
+    List all deployments
+    """
+    try:
+        # Convert deployments to list format
+        deployment_list = []
+        
+        for deployment_id, deployment in deployments.items():
+            deployment_list.append({
+                "id": deployment_id,
+                "name": deployment.get("service_name") or f"{deployment['model_id'].split('/')[-1]} Service",
+                "model_id": deployment["model_id"],
+                "model_type": "text",  # Would be determined from analysis
+                "status": deployment["status"],
+                "gpu": "A10G",  # Would be from config
+                "cost_per_hour": "1.20",  # Would be from config
+                "created_at": deployment["created_at"],
+                "deployed_at": deployment.get("completed_at"),
+                "error": deployment.get("error")
+            })
+        
+        # Add some fallback deployments for demo
+        if not deployment_list:
+            deployment_list = [
+                {
+                    "id": "qwen2-vl-7b",
+                    "name": "Qwen2.5-VL Service",
+                    "model_id": "Qwen/Qwen2.5-VL-7B-Instruct",
+                    "model_type": "vision",
+                    "status": "active",
+                    "gpu": "A100",
+                    "cost_per_hour": "4.00",
+                    "created_at": 1705312200,
+                    "deployed_at": 1705312800
+                },
+                {
+                    "id": "embed-service",
+                    "name": "BGE Embed Service",
+                    "model_id": "BAAI/bge-base-en-v1.5",
+                    "model_type": "embedding",
+                    "status": "active",
+                    "gpu": "A10G",
+                    "cost_per_hour": "1.20",
+                    "created_at": 1705225800,
+                    "deployed_at": 1705226400
+                }
+            ]
+        
+        return deployment_list
+        
+    except Exception as e:
+        logger.error(f"Failed to list deployments: {e}")
+        raise HTTPException(status_code=500, detail=f"Failed to list deployments: {str(e)}")
+
+@router.get("/{deployment_id}")
+async def get_deployment(deployment_id: str):
+    """
+    Get deployment details
+    """
+    try:
+        if deployment_id not in deployments:
+            raise HTTPException(status_code=404, detail="Deployment not found")
+        
+        return deployments[deployment_id]
+        
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Failed to get deployment {deployment_id}: {e}")
+        raise HTTPException(status_code=500, detail=f"Failed to get deployment: {str(e)}")
+
+@router.delete("/{deployment_id}")
+async def cancel_deployment(deployment_id: str):
+    """
+    Cancel a pending deployment
+    """
+    try:
+        if deployment_id not in deployments:
+            raise HTTPException(status_code=404, detail="Deployment not found")
+        
+        deployment = deployments[deployment_id]
+        
+        if deployment["status"] == "pending":
+            deployment["status"] = "cancelled"
+            deployment["cancelled_at"] = time.time()
+            return {"success": True, "message": "Deployment cancelled"}
+        else:
+            raise HTTPException(status_code=400, detail="Cannot cancel deployment in current status")
+            
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Failed to cancel deployment {deployment_id}: {e}")
+        raise HTTPException(status_code=500, detail=f"Failed to cancel deployment: {str(e)}")
+
+@router.post("/{deployment_id}/retry")
+async def retry_deployment(deployment_id: str, background_tasks: BackgroundTasks):
+    """
+    Retry a failed deployment
+    """
+    try:
+        if deployment_id not in deployments:
+            raise HTTPException(status_code=404, detail="Deployment not found")
+        
+        deployment = deployments[deployment_id]
+        
+        if deployment["status"] == "failed":
+            # Reset deployment status
+            deployment["status"] = "pending"
+            deployment["error"] = None
+            deployment["progress"] = "Retrying deployment"
+            
+            # Start deployment in background
+            background_tasks.add_task(
+                perform_deployment,
+                deployment_id,
+                deployment["model_id"],
+                deployment.get("service_name"),
+                deployment.get("auto_deploy", False)
+            )
+            
+            return {"success": True, "message": "Deployment retry started"}
+        else:
+            raise HTTPException(status_code=400, detail="Cannot retry deployment in current status")
+            
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Failed to retry deployment {deployment_id}: {e}")
+        raise HTTPException(status_code=500, detail=f"Failed to retry deployment: {str(e)}")
+
+@router.get("/{deployment_id}/code")
+async def get_service_code(deployment_id: str):
+    """
+    Download the generated service code for a deployment
+    """
+    try:
+        if deployment_id not in deployments:
+            raise HTTPException(status_code=404, detail="Deployment not found")
+        
+        deployment = deployments[deployment_id]
+        service_file = deployment.get("service_file")
+        
+        if not service_file or not Path(service_file).exists():
+            raise HTTPException(status_code=404, detail="Service code not found")
+        
+        # Read the service code file
+        with open(service_file, 'r') as f:
+            service_code = f.read()
+        
+        from fastapi.responses import PlainTextResponse
+        return PlainTextResponse(
+            content=service_code,
+            headers={
+                "Content-Disposition": f"attachment; filename={Path(service_file).name}"
+            }
+        )
+        
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Failed to get service code for {deployment_id}: {e}")
+        raise HTTPException(status_code=500, detail=f"Failed to get service code: {str(e)}")
+
+# Health check for deployments service
+@router.get("/health")
+async def deployments_health():
+    """Health check for deployments service"""
+    return {
+        "status": "healthy",
+        "service": "deployments",
+        "active_deployments": len([d for d in deployments.values() if d["status"] == "active"]),
+        "pending_deployments": len([d for d in deployments.values() if d["status"] == "pending"]),
+        "total_deployments": len(deployments)
+    }