fix(web): preserve legacy provider prefixes on model updates

fix(models): normalize explicit provider-prefixed ids

fix(api): preserve legacy model updates across providers

fix(agent): preserve config identity for explicit provider refs

pkg/agent/instance_test.go

@@ -229,6 +229,43 @@ func TestNewAgentInstance_PreservesDistinctLimiterIdentityForSharedResolvedModel
 	}
 }
 
+func TestNewAgentInstance_PreservesConfigIdentityForExplicitProviderModelRef(t *testing.T) {
+	tmpDir := t.TempDir()
+
+	cfg := &config.Config{
+		Agents: config.AgentsConfig{
+			Defaults: config.AgentDefaults{
+				Workspace: tmpDir,
+				ModelName: "nvidia/z-ai/glm-5.1",
+			},
+		},
+		ModelList: []*config.ModelConfig{
+			{
+				ModelName: "nvidia-glm",
+				Provider:  "nvidia",
+				Model:     "z-ai/glm-5.1",
+				RPM:       7,
+			},
+		},
+	}
+
+	agent := NewAgentInstance(nil, &cfg.Agents.Defaults, cfg, &mockProvider{})
+	if len(agent.Candidates) != 1 {
+		t.Fatalf("len(Candidates) = %d, want 1", len(agent.Candidates))
+	}
+
+	candidate := agent.Candidates[0]
+	if candidate.Provider != "nvidia" || candidate.Model != "z-ai/glm-5.1" {
+		t.Fatalf("candidate = %s/%s, want nvidia/z-ai/glm-5.1", candidate.Provider, candidate.Model)
+	}
+	if candidate.IdentityKey != "model_name:nvidia-glm" {
+		t.Fatalf("identity key = %q, want %q", candidate.IdentityKey, "model_name:nvidia-glm")
+	}
+	if candidate.RPM != 7 {
+		t.Fatalf("RPM = %d, want 7", candidate.RPM)
+	}
+}
+
 func TestNewAgentInstance_AllowsMediaTempDirForReadListAndExec(t *testing.T) {
 	workspace := t.TempDir()
 	mediaDir := media.TempDir()

pkg/agent/model_resolution.go

@@ -60,6 +60,12 @@ func lookupModelConfigByRef(cfg *config.Config, raw string) *config.ModelConfig
 		return mc
 	}
 
+	rawRef := providers.ParseModelRef(raw, "")
+	rawKey := ""
+	if rawRef != nil && strings.TrimSpace(rawRef.Provider) != "" && strings.TrimSpace(rawRef.Model) != "" {
+		rawKey = providers.ModelKey(rawRef.Provider, rawRef.Model)
+	}
+
 	for i := range cfg.ModelList {
 		mc := cfg.ModelList[i]
 		if mc == nil {
@@ -72,10 +78,13 @@ func lookupModelConfigByRef(cfg *config.Config, raw string) *config.ModelConfig
 		if fullModel == raw {
 			return mc
 		}
-		_, modelID := providers.ExtractProtocol(mc)
+		protocol, modelID := providers.ExtractProtocol(mc)
 		if modelID == raw {
 			return mc
 		}
+		if rawKey != "" && providers.ModelKey(protocol, modelID) == rawKey {
+			return mc
+		}
 	}
 
 	return nil

pkg/providers/factory_provider.go

@@ -107,7 +107,13 @@ func ExtractProtocol(cfg *config.ModelConfig) (protocol, modelID string) {
 
 	model := strings.TrimSpace(cfg.Model)
 	if provider := strings.TrimSpace(cfg.Provider); provider != "" {
-		return NormalizeProvider(provider), model
+		normalized := NormalizeProvider(provider)
+		if prefix, rest, found := strings.Cut(model, "/"); found {
+			if NormalizeProvider(prefix) == normalized && strings.TrimSpace(rest) != "" {
+				return normalized, strings.TrimSpace(rest)
+			}
+		}
+		return normalized, model
 	}
 	if model == "" {
 		return "", ""

pkg/providers/factory_provider_test.go

@@ -83,6 +83,18 @@ func TestExtractProtocol(t *testing.T) {
 			wantProtocol: "nvidia",
 			wantModelID:  "z-ai/glm-5.1",
 		},
+		{
+			name:         "explicit provider strips redundant matching prefix",
+			config:       &config.ModelConfig{Provider: "openai", Model: "openai/gpt-4o"},
+			wantProtocol: "openai",
+			wantModelID:  "gpt-4o",
+		},
+		{
+			name:         "explicit provider strips redundant aliased prefix",
+			config:       &config.ModelConfig{Provider: "qwen", Model: "qwen/qwen-plus"},
+			wantProtocol: "qwen-portal",
+			wantModelID:  "qwen-plus",
+		},
 		{
 			name:         "empty provider segment",
 			config:       &config.ModelConfig{Model: "/gpt-4o"},
@@ -152,6 +164,27 @@ func TestCreateProviderFromConfig_UsesExplicitProvider(t *testing.T) {
 	}
 }
 
+func TestCreateProviderFromConfig_StripsRedundantExplicitProviderPrefix(t *testing.T) {
+	cfg := &config.ModelConfig{
+		ModelName: "test-openai",
+		Provider:  "openai",
+		Model:     "openai/gpt-4o",
+		APIBase:   "https://api.example.com/v1",
+	}
+	cfg.SetAPIKey("test-key")
+
+	provider, modelID, err := CreateProviderFromConfig(cfg)
+	if err != nil {
+		t.Fatalf("CreateProviderFromConfig() error = %v", err)
+	}
+	if provider == nil {
+		t.Fatal("CreateProviderFromConfig() returned nil provider")
+	}
+	if modelID != "gpt-4o" {
+		t.Fatalf("modelID = %q, want %q", modelID, "gpt-4o")
+	}
+}
+
 func TestCreateProviderFromConfig_DefaultAPIBase(t *testing.T) {
 	tests := []struct {
 		name     string

web/backend/api/models.go

@@ -6,6 +6,7 @@ import (
 	"io"
 	"net/http"
 	"strconv"
+	"strings"
 	"sync"
 
 	"github.com/sipeed/picoclaw/pkg/config"
@@ -50,6 +51,27 @@ type modelResponse struct {
 	IsVirtual bool   `json:"is_virtual"`
 }
 
+func normalizeExplicitProviderModel(mc *config.ModelConfig) {
+	if mc == nil {
+		return
+	}
+
+	provider := strings.TrimSpace(mc.Provider)
+	model := strings.TrimSpace(mc.Model)
+	if provider == "" || model == "" {
+		return
+	}
+
+	prefix, rest, found := strings.Cut(model, "/")
+	if !found || strings.TrimSpace(rest) == "" {
+		return
+	}
+
+	if providers.NormalizeProvider(prefix) == providers.NormalizeProvider(provider) {
+		mc.Model = strings.TrimSpace(rest)
+	}
+}
+
 // handleListModels returns all model_list entries with masked API keys.
 //
 //	GET /api/models
@@ -139,6 +161,7 @@ func (h *Handler) handleAddModel(w http.ResponseWriter, r *http.Request) {
 	if mc.APIKey != "" {
 		mc.ModelConfig.SetAPIKey(mc.APIKey)
 	}
+	normalizeExplicitProviderModel(&mc.ModelConfig)
 
 	cfg, err := config.LoadConfig(h.configPath)
 	if err != nil {
@@ -241,7 +264,31 @@ func (h *Handler) handleUpdateModel(w http.ResponseWriter, r *http.Request) {
 	// the new field yet, while still allowing explicit clearing via "".
 	if _, ok := rawFields["provider"]; !ok {
 		mc.Provider = cfg.ModelList[idx].Provider
+		// Older clients still round-trip the legacy model field only. When the
+		// stored config encodes provider/model in Model and has no explicit
+		// Provider field yet, continue preserving that hidden provider prefix.
+		// This keeps provider-omitted updates backward-compatible even when an
+		// older client edits the visible model ID.
+		if strings.TrimSpace(cfg.ModelList[idx].Provider) == "" {
+			existingProtocol, existingModelID := providers.ExtractProtocol(cfg.ModelList[idx])
+			existingRawModel := strings.TrimSpace(cfg.ModelList[idx].Model)
+			incomingModel := strings.TrimSpace(mc.Model)
+			if existingRawModel != "" && existingRawModel != existingModelID && incomingModel != "" {
+				if incomingModel == existingModelID {
+					mc.Model = existingRawModel
+				} else if strings.Contains(incomingModel, "/") && !strings.Contains(existingModelID, "/") {
+					// Older clients never saw the hidden provider prefix for simple
+					// legacy entries such as "openai/gpt-4o". If they now send an
+					// explicit provider/model string, treat it as the caller's full
+					// intent instead of re-applying the old hidden prefix.
+					mc.Model = incomingModel
+				} else if !strings.HasPrefix(incomingModel, existingProtocol+"/") {
+					mc.Model = existingProtocol + "/" + incomingModel
+				}
+			}
+		}
 	}
+	normalizeExplicitProviderModel(&mc.ModelConfig)
 
 	cfg.ModelList[idx] = &mc.ModelConfig
 

web/backend/api/models_test.go

@@ -508,6 +508,41 @@ func TestHandleAddModel_PersistsProvider(t *testing.T) {
 	}
 }
 
+func TestHandleAddModel_NormalizesRedundantExplicitProviderPrefix(t *testing.T) {
+	configPath, cleanup := setupOAuthTestEnv(t)
+	defer cleanup()
+
+	h := NewHandler(configPath)
+	mux := http.NewServeMux()
+	h.RegisterRoutes(mux)
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodPost, "/api/models", bytes.NewBufferString(`{
+		"model_name":"openai-gpt",
+		"provider":"openai",
+		"model":"openai/gpt-4o-mini",
+		"api_key":"sk-openai"
+	}`))
+	req.Header.Set("Content-Type", "application/json")
+	mux.ServeHTTP(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusOK, rec.Body.String())
+	}
+
+	cfg, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	added := cfg.ModelList[len(cfg.ModelList)-1]
+	if got := added.Provider; got != "openai" {
+		t.Fatalf("provider = %q, want %q", got, "openai")
+	}
+	if got := added.Model; got != "gpt-4o-mini" {
+		t.Fatalf("model = %q, want %q", got, "gpt-4o-mini")
+	}
+}
+
 func TestHandleAddModel_PersistsCustomHeaders(t *testing.T) {
 	configPath, cleanup := setupOAuthTestEnv(t)
 	defer cleanup()
@@ -699,6 +734,210 @@ func TestHandleUpdateModel_PreservesProviderWhenOmitted(t *testing.T) {
 	}
 }
 
+func TestHandleUpdateModel_NormalizesRedundantExplicitProviderPrefix(t *testing.T) {
+	configPath, cleanup := setupOAuthTestEnv(t)
+	defer cleanup()
+
+	cfg, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	cfg.ModelList = []*config.ModelConfig{{
+		ModelName: "editable",
+		Model:     "gpt-4o",
+		Provider:  "openai",
+	}}
+	if err := config.SaveConfig(configPath, cfg); err != nil {
+		t.Fatalf("SaveConfig() error = %v", err)
+	}
+
+	h := NewHandler(configPath)
+	mux := http.NewServeMux()
+	h.RegisterRoutes(mux)
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodPut, "/api/models/0", bytes.NewBufferString(`{
+		"model_name":"editable",
+		"provider":"openai",
+		"model":"openai/gpt-5.4"
+	}`))
+	req.Header.Set("Content-Type", "application/json")
+	mux.ServeHTTP(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusOK, rec.Body.String())
+	}
+
+	updated, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	if got := updated.ModelList[0].Provider; got != "openai" {
+		t.Fatalf("provider = %q, want %q", got, "openai")
+	}
+	if got := updated.ModelList[0].Model; got != "gpt-5.4" {
+		t.Fatalf("model = %q, want %q", got, "gpt-5.4")
+	}
+}
+
+func TestHandleUpdateModel_PreservesLegacyModelPrefixWhenProviderOmitted(t *testing.T) {
+	configPath, cleanup := setupOAuthTestEnv(t)
+	defer cleanup()
+
+	cfg, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	cfg.ModelList = []*config.ModelConfig{{
+		ModelName: "legacy-openrouter",
+		Model:     "openrouter/openai/gpt-5.4",
+	}}
+	if err := config.SaveConfig(configPath, cfg); err != nil {
+		t.Fatalf("SaveConfig() error = %v", err)
+	}
+
+	h := NewHandler(configPath)
+	mux := http.NewServeMux()
+	h.RegisterRoutes(mux)
+
+	// Simulate an older client: it reads GET /api/models, ignores the new
+	// provider field, then PUTs the visible model string back unchanged.
+	recList := httptest.NewRecorder()
+	reqList := httptest.NewRequest(http.MethodGet, "/api/models", nil)
+	mux.ServeHTTP(recList, reqList)
+
+	if recList.Code != http.StatusOK {
+		t.Fatalf("list status = %d, want %d, body=%s", recList.Code, http.StatusOK, recList.Body.String())
+	}
+
+	var listResp struct {
+		Models []modelResponse `json:"models"`
+	}
+	if err := json.Unmarshal(recList.Body.Bytes(), &listResp); err != nil {
+		t.Fatalf("Unmarshal() error = %v", err)
+	}
+	if len(listResp.Models) != 1 {
+		t.Fatalf("len(models) = %d, want 1", len(listResp.Models))
+	}
+	if got := listResp.Models[0].Provider; got != "openrouter" {
+		t.Fatalf("provider = %q, want %q", got, "openrouter")
+	}
+	if got := listResp.Models[0].Model; got != "openai/gpt-5.4" {
+		t.Fatalf("model = %q, want %q", got, "openai/gpt-5.4")
+	}
+
+	recUpdate := httptest.NewRecorder()
+	reqUpdate := httptest.NewRequest(http.MethodPut, "/api/models/0", bytes.NewBufferString(`{
+		"model_name":"legacy-openrouter",
+		"model":"openai/gpt-5.4"
+	}`))
+	reqUpdate.Header.Set("Content-Type", "application/json")
+	mux.ServeHTTP(recUpdate, reqUpdate)
+
+	if recUpdate.Code != http.StatusOK {
+		t.Fatalf("update status = %d, want %d, body=%s", recUpdate.Code, http.StatusOK, recUpdate.Body.String())
+	}
+
+	updated, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	if got := updated.ModelList[0].Provider; got != "" {
+		t.Fatalf("provider = %q, want empty", got)
+	}
+	if got := updated.ModelList[0].Model; got != "openrouter/openai/gpt-5.4" {
+		t.Fatalf("model = %q, want %q", got, "openrouter/openai/gpt-5.4")
+	}
+}
+
+func TestHandleUpdateModel_PreservesLegacyModelPrefixWhenProviderOmittedAndModelChanges(t *testing.T) {
+	configPath, cleanup := setupOAuthTestEnv(t)
+	defer cleanup()
+
+	cfg, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	cfg.ModelList = []*config.ModelConfig{{
+		ModelName: "legacy-openrouter",
+		Model:     "openrouter/openai/gpt-5.4",
+	}}
+	if err := config.SaveConfig(configPath, cfg); err != nil {
+		t.Fatalf("SaveConfig() error = %v", err)
+	}
+
+	h := NewHandler(configPath)
+	mux := http.NewServeMux()
+	h.RegisterRoutes(mux)
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodPut, "/api/models/0", bytes.NewBufferString(`{
+		"model_name":"legacy-openrouter",
+		"model":"openai/gpt-5.5"
+	}`))
+	req.Header.Set("Content-Type", "application/json")
+	mux.ServeHTTP(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusOK, rec.Body.String())
+	}
+
+	updated, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	if got := updated.ModelList[0].Provider; got != "" {
+		t.Fatalf("provider = %q, want empty", got)
+	}
+	if got := updated.ModelList[0].Model; got != "openrouter/openai/gpt-5.5" {
+		t.Fatalf("model = %q, want %q", got, "openrouter/openai/gpt-5.5")
+	}
+}
+
+func TestHandleUpdateModel_OmittedProviderAllowsLegacyProviderChangeForSimpleVisibleModel(t *testing.T) {
+	configPath, cleanup := setupOAuthTestEnv(t)
+	defer cleanup()
+
+	cfg, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	cfg.ModelList = []*config.ModelConfig{{
+		ModelName: "legacy-openai",
+		Model:     "openai/gpt-5.4",
+	}}
+	if err := config.SaveConfig(configPath, cfg); err != nil {
+		t.Fatalf("SaveConfig() error = %v", err)
+	}
+
+	h := NewHandler(configPath)
+	mux := http.NewServeMux()
+	h.RegisterRoutes(mux)
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodPut, "/api/models/0", bytes.NewBufferString(`{
+		"model_name":"legacy-openai",
+		"model":"anthropic/claude-sonnet-4.6"
+	}`))
+	req.Header.Set("Content-Type", "application/json")
+	mux.ServeHTTP(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusOK, rec.Body.String())
+	}
+
+	updated, err := config.LoadConfig(configPath)
+	if err != nil {
+		t.Fatalf("LoadConfig() error = %v", err)
+	}
+	if got := updated.ModelList[0].Provider; got != "" {
+		t.Fatalf("provider = %q, want empty", got)
+	}
+	if got := updated.ModelList[0].Model; got != "anthropic/claude-sonnet-4.6" {
+		t.Fatalf("model = %q, want %q", got, "anthropic/claude-sonnet-4.6")
+	}
+}
+
 func TestHandleListModels_ReturnsProviderField(t *testing.T) {
 	configPath, cleanup := setupOAuthTestEnv(t)
 	defer cleanup()

web/frontend/src/components/models/edit-model-sheet.tsx

@@ -118,7 +118,7 @@ export function EditModelSheet({
     try {
       await updateModel(model.index, {
         model_name: model.model_name,
-        provider: form.provider.trim() || undefined,
+        provider: form.provider.trim(),
         model: form.modelId.trim(),
         api_base: form.apiBase || undefined,
         api_key: form.apiKey || undefined,