Add jobs router

app/routers/jobs.py

@@ -0,0 +1,78 @@
+"""Training job routes."""
+from fastapi import APIRouter, Depends, HTTPException
+from sqlalchemy.orm import Session
+
+from .. import models, schemas
+from ..database import get_db
+from ..deps import get_current_user
+from ..services import training
+
+router = APIRouter(prefix="/jobs", tags=["jobs"])
+
+
+@router.post("", response_model=schemas.JobOut, status_code=201)
+def create_job(body: schemas.JobCreate, user=Depends(get_current_user), db: Session = Depends(get_db)):
+    epochs = int((body.config or {}).get("epochs", 20))
+    job = models.TrainingJob(
+        name=body.name or f"run-{models._id()[:6]}",
+        owner=user.username,
+        base_model=body.base_model,
+        dataset=body.dataset,
+        job_type=body.job_type,
+        status="queued",
+        config=body.config or {},
+        metrics={},
+        total_epochs=epochs,
+        current_epoch=0,
+    )
+    db.add(job)
+    db.commit()
+    db.refresh(job)
+    training.launch(job.id)
+    return job
+
+
+@router.get("", response_model=list[schemas.JobOut])
+def list_jobs(user=Depends(get_current_user), db: Session = Depends(get_db)):
+    return (
+        db.query(models.TrainingJob)
+        .filter(models.TrainingJob.owner == user.username)
+        .order_by(models.TrainingJob.created_at.desc())
+        .all()
+    )
+
+
+@router.get("/{job_id}", response_model=schemas.JobOut)
+def get_job(job_id: str, user=Depends(get_current_user), db: Session = Depends(get_db)):
+    j = db.get(models.TrainingJob, job_id)
+    if j is None:
+        raise HTTPException(404, "Job not found")
+    if j.owner != user.username:
+        raise HTTPException(403, "Not your job")
+    return j
+
+
+@router.get("/{job_id}/metrics", response_model=list[schemas.MetricPoint])
+def get_metrics(job_id: str, user=Depends(get_current_user), db: Session = Depends(get_db)):
+    j = db.get(models.TrainingJob, job_id)
+    if j is None:
+        raise HTTPException(404, "Job not found")
+    if j.owner != user.username:
+        raise HTTPException(403, "Not your job")
+    losses = (j.metrics or {}).get("loss") or []
+    return [schemas.MetricPoint(epoch=i + 1, loss=float(v)) for i, v in enumerate(losses)]
+
+
+@router.post("/{job_id}/cancel", response_model=schemas.JobOut)
+def cancel_job(job_id: str, user=Depends(get_current_user), db: Session = Depends(get_db)):
+    j = db.get(models.TrainingJob, job_id)
+    if j is None:
+        raise HTTPException(404, "Job not found")
+    if j.owner != user.username:
+        raise HTTPException(403, "Not your job")
+    training.cancel(job_id)
+    if not training.is_running(job_id) and j.status not in ("completed", "failed", "cancelled"):
+        j.status = "cancelled"
+        db.commit()
+        db.refresh(j)
+    return j