internal/process,server: fix unload regression (#828)

In v221 the shutdown behaviour was refactored so shutdown behaviour was
more reliable in stopping a process group. This exposed an existing bug
where the unload API had a timeout of 0 that snuck in during the big
refactor.

- set a default timeout of 10 seconds for unloads called via the API
- add logging around shutdown routine

updates: #807, #808
fixes: #827

internal/server/api.go

@@ -13,6 +13,9 @@ import (
 	"github.com/mostlygeek/llama-swap/internal/shared"
 )
 
+// apiUnloadTimeout is used by the API endpoints to stop processes
+const apiUnloadTimeout = 10 * time.Second
+
 // modelRecord is one entry in the OpenAI-compatible /v1/models listing.
 type modelRecord struct {
 	ID          string         `json:"id"`
@@ -94,7 +97,7 @@ type runningModel struct {
 // handleUnload stops every running local process. Peer models are remote and
 // unaffected.
 func (s *Server) handleUnload(w http.ResponseWriter, r *http.Request) {
-	s.local.Unload(0)
+	s.local.Unload(apiUnloadTimeout)
 	w.WriteHeader(http.StatusOK)
 	w.Write([]byte("OK"))
 }

internal/server/apigroup.go

@@ -66,7 +66,7 @@ func (s *Server) modelStatus() []apiModel {
 
 // handleAPIUnloadAll stops every running local process.
 func (s *Server) handleAPIUnloadAll(w http.ResponseWriter, r *http.Request) {
-	s.local.Unload(0)
+	s.local.Unload(apiUnloadTimeout)
 	w.Header().Set("Content-Type", "application/json")
 	json.NewEncoder(w).Encode(map[string]string{"msg": "ok"})
 }
@@ -83,7 +83,7 @@ func (s *Server) handleAPIUnloadModel(w http.ResponseWriter, r *http.Request) {
 		router.SendResponse(w, r, http.StatusNotFound, "no local server found for requested model")
 		return
 	}
-	s.local.Unload(0, realName)
+	s.local.Unload(apiUnloadTimeout, realName)
 	w.WriteHeader(http.StatusOK)
 	w.Write([]byte("OK"))
 }