Stabilize AI provider health checks

internal/httpapi/providers.go

@@ -14,9 +14,11 @@ type providerStatus struct {
 	Name       string `json:"name"`
 	Configured bool   `json:"configured"`
 	OK         bool   `json:"ok"`
+	Stale      bool   `json:"stale,omitempty"`
 	URL        string `json:"url,omitempty"`
 	Model      string `json:"model,omitempty"`
 	LatencyMS  int64  `json:"latency_ms,omitempty"`
+	LastOKAt   string `json:"last_ok_at,omitempty"`
 	Error      string `json:"error,omitempty"`
 }
 
@@ -25,6 +27,13 @@ type providersStatusResponse struct {
 	Providers []providerStatus `json:"providers"`
 }
 
+type providerHealthSnapshot struct {
+	at        time.Time
+	latencyMS int64
+}
+
+const providerStaleOKWindow = 2 * time.Minute
+
 func (s *Server) handleProviderStatus(w http.ResponseWriter, r *http.Request) {
 	ctx, cancel := contextWithTimeout(r, 8*time.Second)
 	defer cancel()
@@ -63,14 +72,15 @@ func (s *Server) checkLLM(ctx context.Context) providerStatus {
 	st.LatencyMS = time.Since(start).Milliseconds()
 	if err != nil {
 		st.Error = err.Error()
-		return st
+		return s.withStaleProviderOK("llm", st)
 	}
 	defer res.Body.Close()
 	if res.StatusCode >= 300 {
 		st.Error = fmt.Sprintf("http %d: %s", res.StatusCode, readSmallBody(res.Body))
-		return st
+		return s.withStaleProviderOK("llm", st)
 	}
 	st.OK = true
+	s.rememberProviderOK("llm", st.LatencyMS)
 	return st
 }
 
@@ -80,24 +90,63 @@ func (s *Server) checkWhisperX(ctx context.Context) providerStatus {
 	if !st.Configured {
 		return st
 	}
-	start := time.Now()
-	req, err := http.NewRequestWithContext(ctx, http.MethodGet, baseURL+"/health", nil)
-	if err != nil {
-		st.Error = err.Error()
+	paths := []string{"/health", "/healthz", "/readyz", "/"}
+	var lastErr string
+	for _, path := range paths {
+		cctx, cancel := context.WithTimeout(ctx, 2*time.Second)
+		start := time.Now()
+		req, err := http.NewRequestWithContext(cctx, http.MethodGet, baseURL+path, nil)
+		if err != nil {
+			cancel()
+			lastErr = err.Error()
+			continue
+		}
+		res, err := (&http.Client{Timeout: 2 * time.Second}).Do(req)
+		st.LatencyMS = time.Since(start).Milliseconds()
+		cancel()
+		if err != nil {
+			lastErr = err.Error()
+			continue
+		}
+		body := ""
+		if res.StatusCode >= 300 {
+			body = readSmallBody(res.Body)
+		}
+		_ = res.Body.Close()
+		if res.StatusCode >= 300 {
+			lastErr = fmt.Sprintf("%s http %d: %s", path, res.StatusCode, body)
+			continue
+		}
+		st.OK = true
+		s.rememberProviderOK("whisperx", st.LatencyMS)
 		return st
 	}
-	res, err := http.DefaultClient.Do(req)
-	st.LatencyMS = time.Since(start).Milliseconds()
-	if err != nil {
-		st.Error = err.Error()
-		return st
-	}
-	defer res.Body.Close()
-	if res.StatusCode >= 300 {
-		st.Error = fmt.Sprintf("http %d: %s", res.StatusCode, readSmallBody(res.Body))
+	st.Error = lastErr
+	return s.withStaleProviderOK("whisperx", st)
+}
+
+func (s *Server) rememberProviderOK(name string, latencyMS int64) {
+	s.providerMu.Lock()
+	defer s.providerMu.Unlock()
+	s.providerOKs[name] = providerHealthSnapshot{at: time.Now().UTC(), latencyMS: latencyMS}
+}
+
+func (s *Server) withStaleProviderOK(name string, st providerStatus) providerStatus {
+	s.providerMu.Lock()
+	snap, ok := s.providerOKs[name]
+	s.providerMu.Unlock()
+	if !ok || time.Since(snap.at) > providerStaleOKWindow {
 		return st
 	}
 	st.OK = true
+	st.Stale = true
+	if st.LatencyMS == 0 {
+		st.LatencyMS = snap.latencyMS
+	}
+	st.LastOKAt = snap.at.Format(time.RFC3339)
+	if st.Error != "" {
+		st.Error = "последняя проверка не ответила: " + st.Error
+	}
 	return st
 }