[management] Say what went wrong when loading a provider's models fails

Pressing "Load models from provider" against a key the vendor refuses answered
"internal server error". Every outcome on that path is the operator's own key
or upstream, so a 500 was wrong twice over: it told them the server broke, and
it named nothing they could act on.

The failures are already typed and already have sentences written for them —
the save-time check translates the same set. Discovery now runs them through
the same classifier, so the button reports a refused credential or an
unreachable url in the words the form uses elsewhere.

ErrNoDiscovery and ErrInvalidRequest pass through untouched. The handler maps
both already, and a provider with no listing endpoint is a fact about the
catalog entry rather than a failure — the caller falls back to the catalog's
own models instead of showing an error at all.

management/internals/modules/agentnetwork/credentialcheck.go

@@ -49,6 +49,32 @@ func (m *managerImpl) checkProviderCredential(ctx context.Context, provider *typ
 	return status.Errorf(status.InvalidArgument, "%s", message)
 }
 
+// discoveryFailure renders a failed model listing for the operator who pressed
+// the button. Every outcome here is something they did or configured — a key
+// the vendor refused, an upstream that does not answer — so it owes them the
+// same sentence a refused save gives, not the generic 500 an unclassified
+// error turns into.
+//
+// ErrNoDiscovery and ErrInvalidRequest pass through untouched: the handler
+// already maps them, and "this provider has no listing endpoint" is a fact
+// about the catalog rather than a failure to report as one.
+func discoveryFailure(ctx context.Context, catalogID string, err error) error {
+	if errors.Is(err, modeldiscovery.ErrNoDiscovery) || errors.Is(err, modeldiscovery.ErrInvalidRequest) {
+		return err
+	}
+
+	message, _ := credentialCheckFailure(err)
+	if message == "" {
+		return err
+	}
+
+	// The operator's message carries no status code, so the vendor's number is
+	// recorded here or nowhere.
+	log.WithContext(ctx).Infof("agent network model discovery for %s failed: %v", catalogID, err)
+
+	return status.Errorf(status.InvalidArgument, "%s", message)
+}
+
 // credentialCheckFailure renders a discovery failure as the sentence the
 // provider form shows, and reports whether it should block the write.
 //

management/internals/modules/agentnetwork/credentialcheck_test.go

@@ -377,3 +377,82 @@ func TestCreateProvider_AProviderWeCannotCheckStillSaves(t *testing.T) {
 		})
 	}
 }
+
+// TestDiscoveryFailure_TellsTheOperatorWhatWentWrong covers the button, not the
+// save. Pressing "Load models from provider" against a bad key used to answer
+// "internal server error", which names neither the thing that failed nor
+// anything the operator could act on — every outcome here is their key or their
+// URL.
+func TestDiscoveryFailure_TellsTheOperatorWhatWentWrong(t *testing.T) {
+	cases := map[string]struct {
+		err  error
+		want string
+	}{
+		"refused credential": {
+			err:  &modeldiscovery.VendorStatusError{Provider: "Bedrock", Status: 403},
+			want: "the provider rejected the credential",
+		},
+		"upstream that is not the api": {
+			err:  &modeldiscovery.VendorStatusError{Provider: "OpenAI", Status: 404},
+			want: "the upstream url did not answer a model listing",
+		},
+		"upstream that does not resolve": {
+			err: &modeldiscovery.UnreachableError{
+				Provider: "OpenAI",
+				Err:      &net.DNSError{Err: "no such host", Name: "api.example.com", IsNotFound: true},
+			},
+			want: "the upstream url could not be reached: no such host",
+		},
+		"vendor having a bad day": {
+			err:  &modeldiscovery.VendorStatusError{Provider: "Anthropic", Status: 503},
+			want: "the provider returned an error",
+		},
+	}
+
+	for name, tc := range cases {
+		t.Run(name, func(t *testing.T) {
+			err := discoveryFailure(context.Background(), "openai_api", tc.err)
+			require.EqualError(t, err, tc.want)
+
+			var sErr *status.Error
+			require.ErrorAs(t, err, &sErr)
+			require.Equal(t, status.InvalidArgument, sErr.Type(),
+				"a failure the operator caused must not read as a server fault")
+		})
+	}
+}
+
+// TestDiscoveryFailure_LeavesTheCatalogFactsAlone keeps the two outcomes the
+// handler already maps. A provider with no listing endpoint is a fact about the
+// catalog entry, and the caller falls back to the catalog's own models rather
+// than showing an error at all — rewriting it as a refusal would turn a normal
+// path into one.
+func TestDiscoveryFailure_LeavesTheCatalogFactsAlone(t *testing.T) {
+	for name, err := range map[string]error{
+		"no listing endpoint": modeldiscovery.ErrNoDiscovery,
+		"bad request":         fmt.Errorf("%w: unknown catalog provider", modeldiscovery.ErrInvalidRequest),
+	} {
+		t.Run(name, func(t *testing.T) {
+			require.Equal(t, err, discoveryFailure(context.Background(), "openai_api", err),
+				"the handler's own mapping must still see the original error")
+		})
+	}
+}
+
+// TestDiscoverProviderModels_SurfacesTheVendorRefusal drives the manager rather
+// than the classifier, so a future refactor that stops translating on this path
+// fails here rather than silently going back to 500s.
+func TestDiscoverProviderModels_SurfacesTheVendorRefusal(t *testing.T) {
+	ctx := context.Background()
+	f := newBootstrapFixture(t)
+	f.vendor.err = &modeldiscovery.VendorStatusError{Provider: "OpenAI", Status: 401}
+	f.expectPermission("account1", "user1", modules.AgentNetworkProviders, operations.Create, true)
+
+	_, err := f.manager.DiscoverProviderModels(ctx, "account1", "user1", modeldiscovery.Request{
+		CatalogID:   "openai_api",
+		UpstreamURL: "https://api.openai.com",
+		APIKey:      "sk-wrong",
+	}, "")
+
+	require.EqualError(t, err, "the provider rejected the credential")
+}

management/internals/modules/agentnetwork/manager.go

@@ -231,7 +231,11 @@ func (m *managerImpl) DiscoverProviderModels(ctx context.Context, accountID, use
 		req.APIKey = record.APIKey
 	}
 
-	return m.modelDiscovery.Fetch(ctx, req)
+	models, err := m.modelDiscovery.Fetch(ctx, req)
+	if err != nil {
+		return nil, discoveryFailure(ctx, req.CatalogID, err)
+	}
+	return models, nil
 }
 
 // CreateProvider persists a new provider for the account. Providers have no