Skip to content

Prevent stale running model evals #21

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
dzarlax merged 1 commit into from
Jun 16, 2026
Merged

Prevent stale running model evals #21

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 23 additions & 0 deletions internal/adminapi/adminapi_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -539,6 +539,29 @@ func TestModelEvalLoadLegacyAndDetailedReports(t *testing.T) {
}
}

func TestModelEvalNormalizesStaleRunningStatus(t *testing.T) {
s := newTestServer(t)
stale := map[string]modelEvalStatus{
modelCheckKey("openrouter", "stale/model:free"): {
Status: "running",
Provider: "openrouter",
ModelID: "stale/model:free",
StartedAt: time.Now().Add(-2 * time.Hour).Format(time.RFC3339),
UpdatedAt: time.Now().Add(-2 * time.Hour).Format(time.RFC3339),
},
}
data, err := json.Marshal(stale)
if err != nil {
t.Fatal(err)
}
s.settings = &fakeSettingsStore{values: map[string]string{settingKeyModelEvals: string(data)}}
evals := s.loadModelEvals(context.Background())
got := evals[modelCheckKey("openrouter", "stale/model:free")]
if got.Status != "error" || got.Error == "" || got.Failed != 1 {
t.Fatalf("stale running eval not normalized to error: %+v", got)
}
}

func TestWebModelCheckPersistsFreeModelStatus(t *testing.T) {
s, _ := newTGModelTestServer(t)
settings := &fakeSettingsStore{values: map[string]string{}}
Expand Down
32 changes: 29 additions & 3 deletions internal/adminapi/model_evals.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@ const (
settingKeyModelEvals = "recommendation.model_evals"
modelEvalDatasetPath = "evals/workload.json"
modelEvalTimeout = 45 * time.Second
modelEvalStaleAfter = modelEvalTimeout + 15*time.Second
)

type modelEvalStatus struct {
Expand Down Expand Up @@ -69,16 +70,20 @@ func (s *Server) handleModelEval(w http.ResponseWriter, r *http.Request) {
return
}
defer s.modelEvalMu.Unlock()
if err := s.saveModelEval(r.Context(), provider, modelID, runningModelEval(provider, modelID)); err != nil {
saveCtx, saveCancel := context.WithTimeout(context.Background(), 5*time.Second)
defer saveCancel()
if err := s.saveModelEval(saveCtx, provider, modelID, runningModelEval(provider, modelID)); err != nil {
http.Error(w, err.Error(), http.StatusServiceUnavailable)
return
}

caps := s.lookupCapsFor(r.Context(), provider, modelID)
ctx, cancel := context.WithTimeout(r.Context(), modelEvalTimeout)
ctx, cancel := context.WithTimeout(context.Background(), modelEvalTimeout)
defer cancel()
status := s.runModelEval(ctx, provider, modelID, caps)
if err := s.saveModelEval(ctx, provider, modelID, status); err != nil {
saveCtx, saveCancel = context.WithTimeout(context.Background(), 5*time.Second)
defer saveCancel()
if err := s.saveModelEval(saveCtx, provider, modelID, status); err != nil {
http.Error(w, err.Error(), http.StatusServiceUnavailable)
return
}
Expand Down Expand Up @@ -258,9 +263,30 @@ func normalizeSingleModelEval(provider, modelID string, status modelEvalStatus)
if status.CheckedAt == "" && status.FinishedAt != "" {
status.CheckedAt = status.FinishedAt
}
if status.Status == "running" && isStaleModelEval(status, time.Now()) {
now := time.Now().Format(time.RFC3339)
status.Status = "error"
status.FinishedAt = now
status.UpdatedAt = now
status.CheckedAt = now
Comment on lines +267 to +271

@chatgpt-codex-connector chatgpt-codex-connector Bot Jun 16, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Preserve the original stale timeout time

When a stale running eval is read, this stamps CheckedAt/UpdatedAt with time.Now(), but loadModelEvals only normalizes in memory and does not write the converted record back. For any stale record left after a restart or aborted handler, every /evals refresh reassigns a fresh timestamp, so buildModelOpsData keeps sorting that old timeout above genuinely recent evals and the UI reports it as having just finished. Use a deterministic timeout time derived from StartedAt or persist the normalized status once.

Useful? React with 👍 / 👎.

status.Failed = 1
status.Error = "model eval timed out before completion"
status.Failures = []string{status.Error}
}
return status
}

func isStaleModelEval(status modelEvalStatus, now time.Time) bool {
if status.StartedAt == "" {
return false
}
started, err := time.Parse(time.RFC3339, status.StartedAt)
if err != nil {
return false
}
return now.Sub(started) > modelEvalStaleAfter
}

type modelOpsData struct {
ActiveTab string
Checks []modelCheckStatus
Expand Down
Loading