File size: 8,122 Bytes
722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e 66404dc 722753e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 |
"""API route handlers for stroke segmentation.
This module implements an async job queue pattern to handle long-running ML inference:
1. POST /api/segment creates a job and returns immediately (202 Accepted)
2. Background task runs the inference
3. Frontend polls GET /api/jobs/{job_id} for status/results
This pattern avoids HuggingFace Spaces' ~60s gateway timeout.
"""
from __future__ import annotations
import contextlib
import os
import uuid
from pathlib import Path
from fastapi import APIRouter, BackgroundTasks, HTTPException, Request
from stroke_deepisles_demo.api.job_store import JobStatus, get_job_store
from stroke_deepisles_demo.api.schemas import (
CasesResponse,
CreateJobResponse,
JobStatusResponse,
SegmentRequest,
SegmentResponse,
)
from stroke_deepisles_demo.core.logging import get_logger
from stroke_deepisles_demo.data import list_case_ids
from stroke_deepisles_demo.metrics import compute_volume_ml
from stroke_deepisles_demo.pipeline import run_pipeline_on_case
logger = get_logger(__name__)
router = APIRouter()
# Base directory for results
RESULTS_BASE = Path("/tmp/stroke-results")
def get_backend_base_url(request: Request) -> str:
"""Get the backend's public URL for building absolute file URLs.
Priority:
1. BACKEND_PUBLIC_URL env var (for production HF Spaces)
2. Request's base URL (for local development)
"""
env_url = os.environ.get("BACKEND_PUBLIC_URL", "").rstrip("/")
if env_url:
return env_url
return str(request.base_url).rstrip("/")
@router.get("/cases", response_model=CasesResponse)
def get_cases() -> CasesResponse:
"""List available cases from dataset.
Note: This is a sync def (not async) because list_case_ids() is synchronous.
FastAPI automatically runs sync endpoints in a threadpool to avoid blocking.
"""
try:
cases = list_case_ids()
return CasesResponse(cases=cases)
except HTTPException:
raise
except Exception:
logger.exception("Failed to list cases")
raise HTTPException(status_code=500, detail="Failed to retrieve cases") from None
@router.post(
"/segment",
response_model=CreateJobResponse,
status_code=202,
responses={
202: {"description": "Job created successfully"},
400: {"description": "Invalid request"},
500: {"description": "Internal server error"},
},
)
def create_segment_job(
request: Request,
body: SegmentRequest,
background_tasks: BackgroundTasks,
) -> CreateJobResponse:
"""Create an async segmentation job.
Returns immediately with a job ID. The actual ML inference runs in the background.
Poll GET /api/jobs/{jobId} for status updates and results.
This async pattern is required because:
- DeepISLES inference takes 30-60 seconds
- HuggingFace Spaces has a ~60s gateway timeout
- Returning immediately avoids timeout errors
"""
try:
# Use full UUID hex for uniqueness (no truncation)
job_id = uuid.uuid4().hex
store = get_job_store()
backend_url = get_backend_base_url(request)
# Create job record
store.create_job(job_id, body.case_id, body.fast_mode)
# Queue background task
background_tasks.add_task(
run_segmentation_job,
job_id=job_id,
case_id=body.case_id,
fast_mode=body.fast_mode,
backend_url=backend_url,
)
# Note: Don't log case_id as it may be sensitive (medical domain)
logger.info("Created segmentation job %s", job_id)
return CreateJobResponse(
jobId=job_id,
status="pending",
message=f"Segmentation job queued for {body.case_id}",
)
except Exception:
logger.exception("Failed to create segmentation job")
raise HTTPException(status_code=500, detail="Failed to create segmentation job") from None
@router.get(
"/jobs/{job_id}",
response_model=JobStatusResponse,
responses={
200: {"description": "Job status retrieved"},
404: {"description": "Job not found"},
},
)
def get_job_status(job_id: str) -> JobStatusResponse:
"""Get the status of a segmentation job.
Poll this endpoint to track job progress and retrieve results.
Returns:
Job status including progress percentage and results when completed.
Raises:
404: Job not found (may have expired or never existed)
"""
store = get_job_store()
job = store.get_job(job_id)
if job is None:
raise HTTPException(
status_code=404,
detail=f"Job not found: {job_id}. Jobs expire after 1 hour.",
)
# Build response from job data
response = JobStatusResponse(
jobId=job.id,
status=job.status.value,
progress=job.progress,
progressMessage=job.progress_message,
elapsedSeconds=round(job.elapsed_seconds, 2) if job.started_at else None,
result=None,
error=None,
)
# Include result if completed
if job.status == JobStatus.COMPLETED and job.result:
response.result = SegmentResponse(**job.result)
# Include error if failed
if job.status == JobStatus.FAILED and job.error:
response.error = job.error
return response
def run_segmentation_job(
job_id: str,
case_id: str,
fast_mode: bool,
backend_url: str,
) -> None:
"""Execute segmentation in background thread.
This function runs in a threadpool (not the main event loop) because
the ML inference is CPU/GPU-bound and blocking.
Updates job status and progress throughout execution, allowing the
frontend to show meaningful progress updates.
Args:
job_id: Unique job identifier
case_id: Case to process
fast_mode: Whether to use fast inference mode
backend_url: Base URL for constructing result file URLs
"""
store = get_job_store()
job = store.get_job(job_id)
if job is None:
logger.error("Job %s not found when starting execution", job_id)
return
try:
# Mark as running
store.start_job(job_id)
store.update_progress(job_id, 10, "Loading case data...")
# Set up output directory
output_dir = RESULTS_BASE / job_id
store.update_progress(job_id, 20, "Staging files for DeepISLES...")
# Run the pipeline
store.update_progress(job_id, 30, "Running DeepISLES inference...")
result = run_pipeline_on_case(
case_id,
output_dir=output_dir,
fast=fast_mode,
compute_dice=True,
cleanup_staging=True,
)
store.update_progress(job_id, 85, "Computing metrics...")
# Compute volume (may fail for edge cases)
volume_ml = None
with contextlib.suppress(Exception):
volume_ml = round(compute_volume_ml(result.prediction_mask, threshold=0.5), 2)
store.update_progress(job_id, 95, "Preparing results...")
# Build result data
dwi_filename = result.input_files["dwi"].name
pred_filename = result.prediction_mask.name
file_path_prefix = f"/files/{job_id}/{result.case_id}"
result_data = {
"caseId": result.case_id,
"diceScore": result.dice_score,
"volumeMl": volume_ml,
"elapsedSeconds": round(result.elapsed_seconds, 2),
"dwiUrl": f"{backend_url}{file_path_prefix}/{dwi_filename}",
"predictionUrl": f"{backend_url}{file_path_prefix}/{pred_filename}",
}
# Mark as completed
store.complete_job(job_id, result_data)
logger.info(
"Job %s completed: case=%s, dice=%.3f, time=%.1fs",
job_id,
case_id,
result.dice_score or 0,
result.elapsed_seconds,
)
except Exception:
logger.exception("Job %s failed", job_id)
# Sanitize error message - don't expose internal details to clients
store.fail_job(job_id, "Segmentation failed")
|