feat(conversion): add fallback for unavailable model (#156)

Because

- docling model might not be deployed under circumstances

This commit

- add fallback mechanism for unavailable model

Author: heiruwu

pkg/handler/knowledgebase.go

@@ -10,7 +10,6 @@ import (
 	"go.uber.org/zap"
 	"google.golang.org/grpc/metadata"
 
-	"github.com/instill-ai/artifact-backend/config"
 	"github.com/instill-ai/artifact-backend/pkg/constant"
 	"github.com/instill-ai/artifact-backend/pkg/customerror"
 	"github.com/instill-ai/artifact-backend/pkg/logger"
@@ -152,7 +151,8 @@ func (ph *PublicHandler) CreateCatalog(ctx context.Context, req *artifactpb.Crea
 			CreateTime:  dbData.CreateTime.String(),
 			UpdateTime:  dbData.UpdateTime.String(),
 			ConvertingPipelines: []string{
-				config.Config.ModelBackend.Namespace + "/" + service.ChunkMdPipelineID + "@" + service.ConvertDocToMDModelVersion,
+				"instill/" + service.ConvertDocToMDModelID + "@" + service.ConvertDocToMDModelVersion,
+				service.NamespaceID + "/" + service.ConvertDocToMDPipelineID + "@" + service.DocToMDVersion,
 			},
 			SummarizingPipelines: []string{
 				service.NamespaceID + "/" + service.GenerateSummaryPipelineID + "@" + service.GenerateSummaryVersion,
@@ -235,7 +235,8 @@ func (ph *PublicHandler) ListCatalogs(ctx context.Context, req *artifactpb.ListC
 			UpdateTime:  kb.UpdateTime.String(),
 			OwnerName:   kb.Owner,
 			ConvertingPipelines: []string{
-				config.Config.ModelBackend.Namespace + "/" + service.ChunkMdPipelineID + "@" + service.ConvertDocToMDModelVersion,
+				"instill/" + service.ConvertDocToMDModelID + "@" + service.ConvertDocToMDModelVersion,
+				service.NamespaceID + "/" + service.ConvertDocToMDPipelineID + "@" + service.DocToMDVersion,
 			},
 			SummarizingPipelines: []string{
 				service.NamespaceID + "/" + service.GenerateSummaryPipelineID + "@" + service.GenerateSummaryVersion,
@@ -332,7 +333,8 @@ func (ph *PublicHandler) UpdateCatalog(ctx context.Context, req *artifactpb.Upda
 			UpdateTime:  kb.UpdateTime.String(),
 			OwnerName:   kb.Owner,
 			ConvertingPipelines: []string{
-				config.Config.ModelBackend.Namespace + "/" + service.ChunkMdPipelineID + "@" + service.ConvertDocToMDModelVersion,
+				"instill/" + service.ConvertDocToMDModelID + "@" + service.ConvertDocToMDModelVersion,
+				service.NamespaceID + "/" + service.ConvertDocToMDPipelineID + "@" + service.DocToMDVersion,
 			},
 			SummarizingPipelines: []string{
 				service.NamespaceID + "/" + service.GenerateSummaryPipelineID + "@" + service.GenerateSummaryVersion,

pkg/service/model.go

@@ -10,7 +10,6 @@ import (
 	"google.golang.org/grpc/metadata"
 	"google.golang.org/protobuf/types/known/structpb"
 
-	"github.com/instill-ai/artifact-backend/config"
 	"github.com/instill-ai/artifact-backend/pkg/constant"
 	"github.com/instill-ai/artifact-backend/pkg/logger"
 
@@ -42,8 +41,9 @@ func (s *Service) ConvertToMDModel(ctx context.Context, fileUID uuid.UUID, calle
 	// Get the appropriate prefix for the file type
 	prefix := getFileTypePrefix(fileType)
 
+	namespaceID := "admin"
 	req := &modelpb.TriggerNamespaceModelRequest{
-		NamespaceId: config.Config.ModelBackend.Namespace,
+		NamespaceId: namespaceID,
 		ModelId:     ConvertDocToMDModelID,
 		Version:     ConvertDocToMDModelVersion,
 		TaskInputs: []*structpb.Struct{
@@ -63,8 +63,20 @@ func (s *Service) ConvertToMDModel(ctx context.Context, fileUID uuid.UUID, calle
 
 	resp, err := s.ModelPub.TriggerNamespaceModel(ctx, req)
 	if err != nil {
-		logger.Error("failed to trigger model", zap.Error(err))
-		return "", fmt.Errorf("failed to trigger %s model: %w", ConvertDocToMDModelID, err)
+		namespaceID = "instill"
+		req.NamespaceId = namespaceID
+		resp, err = s.ModelPub.TriggerNamespaceModel(ctx, req)
+		if err != nil {
+			logger.Error(fmt.Sprintf("failed to trigger %s model", ConvertDocToMDModelID), zap.Error(err))
+			return "", fmt.Errorf("failed to trigger %s model: %w", ConvertDocToMDModelID, err)
+		}
+	}
+
+	convertingModelMetadata := namespaceID + "/" + ConvertDocToMDModelID + "@" + ConvertDocToMDModelVersion
+	err = s.Repository.UpdateKbFileExtraMetaData(ctx, fileUID, "", convertingModelMetadata, "", "", "", nil, nil, nil, nil, nil)
+	if err != nil {
+		logger.Error("Failed to save converting pipeline metadata.", zap.String("File uid:", fileUID.String()))
+		return "", fmt.Errorf("failed to save converting model metadata: %w", err)
 	}
 
 	result, err := getModelConvertResult(resp)

pkg/service/pipeline.go

@@ -61,8 +61,8 @@ var PresetPipelinesList = []struct {
 	{ID: QAPipelineID, Version: QAVersion},
 }
 
-// ConvertToMDPipeForFiles using converting pipeline to convert some file type to MD and consume caller's credits
-func (s *Service) ConvertToMDPipeForFiles(ctx context.Context, fileUID uuid.UUID, caller uuid.UUID, requester uuid.UUID, fileBase64 string, fileType artifactpb.FileType) (string, error) {
+// ConvertToMDPipe using converting pipeline to convert some file type to MD and consume caller's credits
+func (s *Service) ConvertToMDPipe(ctx context.Context, fileUID uuid.UUID, caller uuid.UUID, requester uuid.UUID, fileBase64 string, fileType artifactpb.FileType) (string, error) {
 	logger, _ := logger.GetZapLogger(ctx)
 	var md metadata.MD
 	if requester != uuid.Nil {
@@ -136,6 +136,13 @@ func (s *Service) ConvertToMDPipeForFiles(ctx context.Context, fileUID uuid.UUID
 		return "", fmt.Errorf("failed to trigger %s pipeline: %w", pipelineID, err)
 	}
 
+	convertingModelMetadata := NamespaceID + "/" + ConvertDocToMDModelID + "@" + ConvertDocToMDModelVersion
+	err = s.Repository.UpdateKbFileExtraMetaData(ctx, fileUID, "", convertingModelMetadata, "", "", "", nil, nil, nil, nil, nil)
+	if err != nil {
+		logger.Error("Failed to save converting pipeline metadata.", zap.String("File uid:", fileUID.String()))
+		return "", fmt.Errorf("failed to save converting model metadata: %w", err)
+	}
+
 	result, err := getConvertResult(resp)
 	if err != nil {
 		logger.Error("failed to get convert result", zap.Error(err))

pkg/worker/persistentcatalogworker.go

@@ -11,7 +11,6 @@ import (
 	"go.uber.org/zap"
 
 	"github.com/gofrs/uuid"
-	"github.com/instill-ai/artifact-backend/config"
 	"github.com/instill-ai/artifact-backend/pkg/constant"
 	"github.com/instill-ai/artifact-backend/pkg/logger"
 	"github.com/instill-ai/artifact-backend/pkg/minio"
@@ -379,15 +378,11 @@ func (wp *persistentCatalogFileToEmbWorkerPool) processConvertingFile(ctx contex
 	requesterUID := file.RequesterUID
 	convertedMD, err := wp.svc.ConvertToMDModel(ctx, file.UID, file.CreatorUID, requesterUID, base64Data, artifactpb.FileType(artifactpb.FileType_value[file.Type]))
 	if err != nil {
-		logger.Error("Failed to convert pdf to md using pdf-to-md pipeline.", zap.String("File path", fileInMinIOPath))
-		return nil, artifactpb.FileProcessStatus_FILE_PROCESS_STATUS_UNSPECIFIED, err
-	}
-
-	convertingModelMetadata := config.Config.ModelBackend.Namespace + "/" + service.ConvertDocToMDModelID + "@" + service.ConvertDocToMDModelVersion
-	err = wp.svc.Repository.UpdateKbFileExtraMetaData(ctx, file.UID, "", convertingModelMetadata, "", "", "", nil, nil, nil, nil, nil)
-	if err != nil {
-		logger.Error("Failed to save converting pipeline metadata.", zap.String("File uid:", file.UID.String()))
-		return nil, artifactpb.FileProcessStatus_FILE_PROCESS_STATUS_UNSPECIFIED, fmt.Errorf("failed to save converting model metadata: %w", err)
+		logger.Error("Failed to convert pdf to md using docling model, fallback to pipeline.")
+		if convertedMD, err = wp.svc.ConvertToMDPipe(ctx, file.UID, file.CreatorUID, requesterUID, base64Data, artifactpb.FileType(artifactpb.FileType_value[file.Type])); err != nil {
+			logger.Error("Failed to convert pdf to md using pdf-to-md pipeline.", zap.String("File path", fileInMinIOPath))
+			return nil, artifactpb.FileProcessStatus_FILE_PROCESS_STATUS_UNSPECIFIED, err
+		}
 	}
 
 	// save the converted file into object storage and metadata into database