Stream job logs from current offset

backend/app/agents/evaluation_agent.py

@@ -59,6 +59,11 @@ def evaluate_project() -> dict:
         and "setResults(resultsNext)" in frontend_text
         and "slice(0, 240)" not in frontend_text,
         "job_progress_ui": "JobProgressBar" in frontend_text and "progressTrack" in frontend_text,
+        "live_log_stream_ui": "EventSource" in frontend_text
+        and "eventSourceRef" in frontend_text
+        and "log_size" in frontend_text
+        and "events?offset=" in frontend_text,
+        "live_log_offset_api": "log_size" in backend_text and "offset: int = Query(0, ge=0)" in backend_text,
         "runtime_readiness_ui": "runtimeReadiness" in frontend_text and "环境就绪" in frontend_text,
         "capability_matrix_ui": "capabilities" in frontend_text and "全功能矩阵" in frontend_text,
         "dataset_api": "/api/datasets" in backend_text and "api_upload_dataset_files" in backend_text,

backend/app/main.py

@@ -39,6 +39,8 @@ app.add_middleware(
 def _job_with_progress(job: dict, include_log_tail: bool = False) -> dict:
     enriched = dict(job)
     max_bytes = 65536 if include_log_tail else 32768
+    log_path = Path(enriched["log_path"])
+    enriched["log_size"] = log_path.stat().st_size if log_path.exists() else 0
     enriched["progress"] = progress_from_log_path(enriched["log_path"], enriched["status"], max_bytes=max_bytes)
     if include_log_tail:
         enriched["log_tail"] = db.log_tail(enriched["log_path"], max_bytes=max_bytes)
@@ -183,9 +185,9 @@ def api_cancel_job(job_id: str) -> dict:
 
 
 @app.get("/api/jobs/{job_id}/events")
-async def api_job_events(job_id: str):
+async def api_job_events(job_id: str, offset: int = Query(0, ge=0)):
     async def stream():
-        last_size = 0
+        last_size = offset
         while True:
             job = db.get_job(job_id)
             if not job:
@@ -196,6 +198,8 @@ async def api_job_events(job_id: str):
             chunk = ""
             if path.exists():
                 size = path.stat().st_size
+                if last_size > size:
+                    last_size = size
                 if size > last_size:
                     with path.open("rb") as handle:
                         handle.seek(last_size)

backend/tests/test_jobs.py

@@ -1,6 +1,38 @@
+from fastapi.testclient import TestClient
+
 from app.jobs import default_conda_env_for_job
+from app.main import _job_with_progress, app
 
 
 def test_mmseg_jobs_use_mmseg_conda_env_by_default():
     assert default_conda_env_for_job("mmseg.train") == "seg_mmcv"
     assert default_conda_env_for_job("segmodel.train") == "seg_smp"
+
+
+def test_job_progress_reports_log_size(tmp_path):
+    log_path = tmp_path / "job.log"
+    log_path.write_text("line one\nline two\n", encoding="utf-8")
+    job = {"id": "job1", "status": "running", "log_path": str(log_path)}
+
+    enriched = _job_with_progress(job, include_log_tail=True)
+
+    assert enriched["log_size"] == log_path.stat().st_size
+    assert enriched["log_tail"] == "line one\nline two\n"
+
+
+def test_job_events_respect_log_offset(tmp_path, monkeypatch):
+    from app import main
+
+    log_path = tmp_path / "job.log"
+    old_chunk = "old chunk\n"
+    log_path.write_text(old_chunk + "new chunk\n", encoding="utf-8")
+
+    def fake_get_job(job_id):
+        return {"id": job_id, "status": "success", "log_path": str(log_path)}
+
+    monkeypatch.setattr(main.db, "get_job", fake_get_job)
+    response = TestClient(app).get(f"/api/jobs/job1/events?offset={len(old_chunk.encode('utf-8'))}")
+
+    assert response.status_code == 200
+    assert "new chunk" in response.text
+    assert "old chunk" not in response.text