create transactions from AI receipt image recognition results

2025-09-21 04:00:56 +08:00
parent 00f1d0418f
commit 5d88287ae2
50 changed files with 2356 additions and 22 deletions
--- a/README.md
+++ b/README.md
@@ -30,6 +30,7 @@ Live Demo: [https://ezbookkeeping-demo.mayswind.net](https://ezbookkeeping-demo.
    - PWA support for native-like mobile experience
    - Dark mode
 - **AI-Powered Features**
+    - Receipt image recognition
    - Supports MCP (Model Context Protocol) for AI integration
 - **Powerful Bookkeeping**
    - Two-level accounts and categories
--- a/cmd/initializer.go
+++ b/cmd/initializer.go
@@ -9,6 +9,7 @@ import (
 	"github.com/mayswind/ezbookkeeping/pkg/datastore"
 	"github.com/mayswind/ezbookkeeping/pkg/duplicatechecker"
 	"github.com/mayswind/ezbookkeeping/pkg/exchangerates"
+	"github.com/mayswind/ezbookkeeping/pkg/llm"
 	"github.com/mayswind/ezbookkeeping/pkg/log"
 	"github.com/mayswind/ezbookkeeping/pkg/mail"
 	"github.com/mayswind/ezbookkeeping/pkg/settings"
@@ -90,6 +91,15 @@ func initializeSystem(c *core.CliContext) (*settings.Config, error) {
 		return nil, err
 	}

+	err = llm.InitializeLargeLanguageModelProvider(config)
+
+	if err != nil {
+		if !isDisableBootLog {
+			log.BootErrorf(c, "[initializer.initializeSystem] initializes large language model provider failed, because %s", err.Error())
+		}
+		return nil, err
+	}
+
 	err = uuid.InitializeUuidGenerator(config)

 	if err != nil {
@@ -155,6 +165,9 @@ func getConfigWithoutSensitiveData(config *settings.Config) *settings.Config {
 	clonedConfig.DatabaseConfig.DatabasePassword = "****"
 	clonedConfig.SMTPConfig.SMTPPasswd = "****"
 	clonedConfig.MinIOConfig.SecretAccessKey = "****"
+	clonedConfig.OpenAIAPIKey = "****"
+	clonedConfig.OpenAICompatibleAPIKey = "****"
+	clonedConfig.OpenRouterAPIKey = "****"
 	clonedConfig.SecretKey = "****"
 	clonedConfig.AmapApplicationSecret = "****"

--- a/cmd/webserver.go
+++ b/cmd/webserver.go
@@ -396,6 +396,13 @@ func startWebServer(c *core.CliContext) error {
 			apiV1Route.POST("/transaction/templates/move.json", bindApi(api.TransactionTemplates.TemplateMoveHandler))
 			apiV1Route.POST("/transaction/templates/delete.json", bindApi(api.TransactionTemplates.TemplateDeleteHandler))

+			// Large Language Models
+			if config.LLMProvider != "" {
+				if config.TransactionFromAIImageRecognition {
+					apiV1Route.POST("/llm/transactions/recognize_receipt_image.json", bindApi(api.LargeLanguageModels.RecognizeReceiptImageHandler))
+				}
+			}
+
 			// Exchange Rates
 			apiV1Route.GET("/exchange_rates/latest.json", bindApi(api.ExchangeRates.LatestExchangeRateHandler))
 			apiV1Route.POST("/exchange_rates/user_custom/update.json", bindApi(api.ExchangeRates.UserCustomExchangeRateUpdateHandler))
--- a/conf/ezbookkeeping.ini
+++ b/conf/ezbookkeeping.ini
@@ -164,6 +164,53 @@ webdav_proxy = system
 # For "webdav" storage only, set to true to skip tls verification when connect webdav
 webdav_skip_tls_verify = false

+[llm]
+# Large Language Model (LLM) provider, supports the following types: "openai", "openai_compatible", "openrouter", "ollama"
+llm_provider =
+
+# For "openai" llm provider only, OpenAI API secret key, please visit https://platform.openai.com/api-keys for more information
+openai_api_key =
+
+# For "openai" llm provider only, receipt image recognition model for creating transactions from images
+openai_receipt_image_recognition_model_id =
+
+# For "openai_compatible" llm provider only, OpenAI compatible API base url, e.g. "https://api.openai.com/v1/"
+openai_compatible_base_url =
+
+# For "openai_compatible" llm provider only, OpenAI compatible API secret key
+openai_compatible_api_key =
+
+# For "openai_compatible" llm provider only, receipt image recognition model for creating transactions from images
+openai_compatible_receipt_image_recognition_model_id =
+
+# For "openrouter" llm provider only, OpenRouter API key, please visit https://openrouter.ai/settings/keys for more information
+openrouter_api_key =
+
+# For "openrouter" llm provider only, receipt image recognition model for creating transactions from images
+openrouter_receipt_image_recognition_model_id =
+
+# For "ollama" llm provider only, Ollama server url, e.g. "http://127.0.0.1:11434/"
+ollama_server_url =
+
+# For "ollama" llm provider only, receipt image recognition model for creating transactions from images
+ollama_receipt_image_recognition_model_id =
+
+# Set to true to enable creating transactions from AI image recognition results, requires llm_provider and its related receipt image recognition model to be configured properly
+transaction_from_ai_image_recognition = false
+
+# Maximum allowed AI recognition picture file size (1 - 4294967295 bytes)
+max_ai_recognition_picture_size = 10485760
+
+# Requesting large language model api timeout (0 - 4294967295 milliseconds)
+# Set to 0 to disable timeout for requesting large language model api, default is 60000 (60 seconds)
+request_timeout = 60000
+
+# Proxy for ezbookkeeping server requesting large language model api, supports "system" (use system proxy), "none" (do not use proxy), or proxy URL which starts with "http://", "https://" or "socks5://", default is "system"
+proxy = system
+
+# Set to true to skip tls verification when request large language model api
+skip_tls_verify = false
+
 [uuid]
 # Uuid generator type, supports "internal" currently
 generator_type = internal
--- a/pkg/api/large_language_models.go
+++ b/pkg/api/large_language_models.go
@@ -0,0 +1,345 @@
+package api
+
+import (
+	"bytes"
+	"encoding/json"
+	"io"
+	"strings"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/errs"
+	"github.com/mayswind/ezbookkeeping/pkg/llm"
+	"github.com/mayswind/ezbookkeeping/pkg/log"
+	"github.com/mayswind/ezbookkeeping/pkg/models"
+	"github.com/mayswind/ezbookkeeping/pkg/services"
+	"github.com/mayswind/ezbookkeeping/pkg/settings"
+	"github.com/mayswind/ezbookkeeping/pkg/templates"
+	"github.com/mayswind/ezbookkeeping/pkg/utils"
+)
+
+// LargeLanguageModelsApi represents large language models api
+type LargeLanguageModelsApi struct {
+	ApiUsingConfig
+	transactionCategories *services.TransactionCategoryService
+	transactionTags       *services.TransactionTagService
+	accounts              *services.AccountService
+	users                 *services.UserService
+}
+
+// Initialize a large language models api singleton instance
+var (
+	LargeLanguageModels = &LargeLanguageModelsApi{
+		ApiUsingConfig: ApiUsingConfig{
+			container: settings.Container,
+		},
+		transactionCategories: services.TransactionCategories,
+		transactionTags:       services.TransactionTags,
+		accounts:              services.Accounts,
+		users:                 services.Users,
+	}
+)
+
+// RecognizeReceiptImageHandler returns the recognized receipt image result
+func (a *LargeLanguageModelsApi) RecognizeReceiptImageHandler(c *core.WebContext) (any, *errs.Error) {
+	if a.CurrentConfig().LLMProvider == "" || !a.CurrentConfig().TransactionFromAIImageRecognition {
+		return nil, errs.ErrLargeLanguageModelProviderNotEnabled
+	}
+
+	utcOffset, err := c.GetClientTimezoneOffset()
+
+	if err != nil {
+		log.Warnf(c, "[large_language_models.RecognizeReceiptImageHandler] cannot get client timezone offset, because %s", err.Error())
+		return nil, errs.ErrClientTimezoneOffsetInvalid
+	}
+
+	uid := c.GetCurrentUid()
+	user, err := a.users.GetUserById(c, uid)
+
+	if err != nil {
+		if !errs.IsCustomError(err) {
+			log.Warnf(c, "[large_language_models.RecognizeReceiptImageHandler] failed to get user for user \"uid:%d\", because %s", uid, err.Error())
+		}
+
+		return false, errs.ErrUserNotFound
+	}
+
+	if user.FeatureRestriction.Contains(core.USER_FEATURE_RESTRICTION_TYPE_CREATE_TRANSACTION_FROM_AI_IMAGE_RECOGNITION) {
+		return false, errs.ErrNotPermittedToPerformThisAction
+	}
+
+	form, err := c.MultipartForm()
+
+	if err != nil {
+		log.Errorf(c, "[large_language_models.RecognizeReceiptImageHandler] failed to get multi-part form data for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrParameterInvalid
+	}
+
+	imageFiles := form.File["image"]
+
+	if len(imageFiles) < 1 {
+		log.Warnf(c, "[large_language_models.RecognizeReceiptImageHandler] there is no image in request for user \"uid:%d\"", uid)
+		return nil, errs.ErrNoAIRecognitionImage
+	}
+
+	if imageFiles[0].Size < 1 {
+		log.Warnf(c, "[large_language_models.RecognizeReceiptImageHandler] the size of image in request is zero for user \"uid:%d\"", uid)
+		return nil, errs.ErrAIRecognitionImageIsEmpty
+	}
+
+	if imageFiles[0].Size > int64(a.CurrentConfig().MaxAIRecognitionPictureFileSize) {
+		log.Warnf(c, "[large_language_models.RecognizeReceiptImageHandler] the upload file size \"%d\" exceeds the maximum size \"%d\" of image for user \"uid:%d\"", imageFiles[0].Size, a.CurrentConfig().MaxAIRecognitionPictureFileSize, uid)
+		return nil, errs.ErrExceedMaxAIRecognitionImageFileSize
+	}
+
+	fileExtension := utils.GetFileNameExtension(imageFiles[0].Filename)
+
+	if utils.GetImageContentType(fileExtension) == "" {
+		log.Warnf(c, "[large_language_models.RecognizeReceiptImageHandler] the file extension \"%s\" of image in request is not supported for user \"uid:%d\"", fileExtension, uid)
+		return nil, errs.ErrImageTypeNotSupported
+	}
+
+	imageFile, err := imageFiles[0].Open()
+
+	if err != nil {
+		log.Errorf(c, "[large_language_models.RecognizeReceiptImageHandler] failed to get image file from request for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrOperationFailed
+	}
+
+	defer imageFile.Close()
+
+	imageData, err := io.ReadAll(imageFile)
+
+	if err != nil {
+		log.Errorf(c, "[large_language_models.RecognizeReceiptImageHandler] failed to read image file from request for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrOperationFailed
+	}
+
+	accounts, err := a.accounts.GetAllAccountsByUid(c, uid)
+
+	if err != nil {
+		log.Errorf(c, "[large_language_models.RecognizeReceiptImageHandler] failed to get all accounts for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.Or(err, errs.ErrOperationFailed)
+	}
+
+	accountMap := a.accounts.GetVisibleAccountNameMapByList(accounts)
+	accountNames := make([]string, 0, len(accounts))
+
+	for i := 0; i < len(accounts); i++ {
+		if accounts[i].Hidden || accounts[i].Type == models.ACCOUNT_TYPE_MULTI_SUB_ACCOUNTS {
+			continue
+		}
+
+		accountNames = append(accountNames, accounts[i].Name)
+	}
+
+	categories, err := a.transactionCategories.GetAllCategoriesByUid(c, uid, 0, -1)
+
+	if err != nil {
+		log.Errorf(c, "[large_language_models.RecognizeReceiptImageHandler] failed to get categories for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.Or(err, errs.ErrOperationFailed)
+	}
+
+	incomeCategoryMap := make(map[string]*models.TransactionCategory)
+	incomeCategoryNames := make([]string, 0)
+
+	expenseCategoryMap := make(map[string]*models.TransactionCategory)
+	expenseCategoryNames := make([]string, 0)
+
+	transferCategoryMap := make(map[string]*models.TransactionCategory)
+	transferCategoryNames := make([]string, 0)
+
+	for i := 0; i < len(categories); i++ {
+		category := categories[i]
+
+		if category.Hidden || category.ParentCategoryId == models.LevelOneTransactionCategoryParentId {
+			continue
+		}
+
+		if category.Type == models.CATEGORY_TYPE_INCOME {
+			incomeCategoryMap[category.Name] = category
+			incomeCategoryNames = append(incomeCategoryNames, category.Name)
+		} else if category.Type == models.CATEGORY_TYPE_EXPENSE {
+			expenseCategoryMap[category.Name] = category
+			expenseCategoryNames = append(expenseCategoryNames, category.Name)
+		} else if category.Type == models.CATEGORY_TYPE_TRANSFER {
+			transferCategoryMap[category.Name] = category
+			transferCategoryNames = append(transferCategoryNames, category.Name)
+		}
+	}
+
+	tags, err := a.transactionTags.GetAllTagsByUid(c, uid)
+
+	if err != nil {
+		log.Errorf(c, "[large_language_models.RecognizeReceiptImageHandler] failed to get tags for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.Or(err, errs.ErrOperationFailed)
+	}
+
+	tagMap := a.transactionTags.GetVisibleTagNameMapByList(tags)
+	tagNames := make([]string, 0, len(tags))
+
+	for i := 0; i < len(tags); i++ {
+		if tags[i].Hidden {
+			continue
+		}
+
+		tagNames = append(tagNames, tags[i].Name)
+	}
+
+	systemPrompt, err := templates.GetTemplate(templates.SYSTEM_PROMPT_RECEIPT_IMAGE_RECOGNITION)
+
+	if err != nil {
+		return nil, errs.Or(err, errs.ErrOperationFailed)
+	}
+
+	systemPromptParams := map[string]any{
+		"AllExpenseCategoryNames":  strings.Join(expenseCategoryNames, "\n"),
+		"AllIncomeCategoryNames":   strings.Join(incomeCategoryNames, "\n"),
+		"AllTransferCategoryNames": strings.Join(transferCategoryNames, "\n"),
+		"AllAccountNames":          strings.Join(accountNames, "\n"),
+		"AllTagNames":              strings.Join(tagNames, "\n"),
+	}
+
+	var bodyBuffer bytes.Buffer
+	err = systemPrompt.Execute(&bodyBuffer, systemPromptParams)
+
+	if err != nil {
+		return nil, errs.Or(err, errs.ErrOperationFailed)
+	}
+
+	llmRequest := &llm.LargeLanguageModelRequest{
+		Stream:         false,
+		SystemPrompt:   strings.ReplaceAll(bodyBuffer.String(), "\r\n", "\n"),
+		UserPrompt:     imageData,
+		UserPromptType: llm.LARGE_LANGUAGE_MODEL_REQUEST_PROMPT_TYPE_IMAGE_URL,
+	}
+
+	llmResponse, err := llm.Container.GetJsonResponseByReceiptImageRecognitionModel(c, c.GetCurrentUid(), a.CurrentConfig(), llmRequest)
+
+	if err != nil {
+		return nil, errs.Or(err, errs.ErrOperationFailed)
+	}
+
+	var result *models.RecognizedReceiptImageResult
+
+	if err := json.Unmarshal([]byte(llmResponse.Content), &result); err != nil {
+		log.Errorf(c, "[large_language_models.RecognizeReceiptImageHandler] failed to unmarshal recognized receipt image result from llm response \"%s\" for user \"uid:%d\", because %s", llmResponse.Content, uid, err.Error())
+		return nil, errs.Or(err, errs.ErrOperationFailed)
+	}
+
+	return a.parseRecognizedReceiptImageResponse(c, uid, utcOffset, result, accountMap, expenseCategoryMap, incomeCategoryMap, transferCategoryMap, tagMap)
+}
+
+func (a *LargeLanguageModelsApi) parseRecognizedReceiptImageResponse(c *core.WebContext, uid int64, utcOffset int16, recognizedResult *models.RecognizedReceiptImageResult, accountMap map[string]*models.Account, expenseCategoryMap map[string]*models.TransactionCategory, incomeCategoryMap map[string]*models.TransactionCategory, transferCategoryMap map[string]*models.TransactionCategory, tagMap map[string]*models.TransactionTag) (*models.RecognizedReceiptImageResponse, *errs.Error) {
+	recognizedReceiptImageResponse := &models.RecognizedReceiptImageResponse{
+		Type: models.TRANSACTION_TYPE_EXPENSE,
+	}
+
+	if recognizedResult == nil {
+		log.Errorf(c, "[large_language_models.parseRecognizedReceiptImageResponse] recoginzed result is null")
+		return nil, errs.ErrOperationFailed
+	}
+
+	if recognizedResult.Type == "income" {
+		recognizedReceiptImageResponse.Type = models.TRANSACTION_TYPE_INCOME
+
+		if len(recognizedResult.CategoryName) > 0 {
+			category, exists := incomeCategoryMap[recognizedResult.CategoryName]
+
+			if exists {
+				recognizedReceiptImageResponse.CategoryId = category.CategoryId
+			}
+		}
+	} else if recognizedResult.Type == "expense" {
+		recognizedReceiptImageResponse.Type = models.TRANSACTION_TYPE_EXPENSE
+
+		if len(recognizedResult.CategoryName) > 0 {
+			category, exists := expenseCategoryMap[recognizedResult.CategoryName]
+
+			if exists {
+				recognizedReceiptImageResponse.CategoryId = category.CategoryId
+			}
+		}
+	} else if recognizedResult.Type == "transfer" {
+		recognizedReceiptImageResponse.Type = models.TRANSACTION_TYPE_TRANSFER
+
+		if len(recognizedResult.CategoryName) > 0 {
+			category, exists := transferCategoryMap[recognizedResult.CategoryName]
+
+			if exists {
+				recognizedReceiptImageResponse.CategoryId = category.CategoryId
+			}
+		}
+	} else {
+		log.Errorf(c, "[large_language_models.parseRecognizedReceiptImageResponse] recoginzed transaction type \"%s\" is invalid", recognizedResult.Type)
+		return nil, errs.ErrOperationFailed
+	}
+
+	if len(recognizedResult.Time) > 0 {
+		timestamp, err := utils.ParseFromLongDateTime(recognizedResult.Time, utcOffset)
+
+		if err != nil {
+			log.Warnf(c, "[large_language_models.parseRecognizedReceiptImageResponse] recoginzed time \"%s\" is invalid", recognizedResult.Time)
+		} else {
+			recognizedReceiptImageResponse.Time = timestamp.Unix()
+		}
+	}
+
+	if len(recognizedResult.Amount) > 0 {
+		amount, err := utils.ParseAmount(recognizedResult.Amount)
+
+		if err != nil {
+			log.Errorf(c, "[large_language_models.parseRecognizedReceiptImageResponse] recoginzed amount \"%s\" is invalid", recognizedResult.Amount)
+			return nil, errs.ErrOperationFailed
+		}
+
+		recognizedReceiptImageResponse.SourceAmount = amount
+
+		if recognizedReceiptImageResponse.Type == models.TRANSACTION_TYPE_TRANSFER && len(recognizedResult.DestinationAmount) > 0 {
+			destinationAmount, err := utils.ParseAmount(recognizedResult.DestinationAmount)
+
+			if err != nil {
+				log.Errorf(c, "[large_language_models.parseRecognizedReceiptImageResponse] recoginzed destination amount \"%s\" is invalid", recognizedResult.DestinationAmount)
+				return nil, errs.ErrOperationFailed
+			}
+
+			recognizedReceiptImageResponse.DestinationAmount = destinationAmount
+		}
+	}
+
+	if len(recognizedResult.AccountName) > 0 {
+		account, exists := accountMap[recognizedResult.AccountName]
+
+		if exists {
+			recognizedReceiptImageResponse.SourceAccountId = account.AccountId
+		}
+	}
+
+	if len(recognizedResult.DestinationAccountName) > 0 {
+		account, exists := accountMap[recognizedResult.DestinationAccountName]
+
+		if exists {
+			recognizedReceiptImageResponse.DestinationAccountId = account.AccountId
+		}
+	}
+
+	if len(recognizedResult.TagNames) > 0 {
+		tagIds := make([]string, 0, len(recognizedResult.TagNames))
+
+		for i := 0; i < len(recognizedResult.TagNames); i++ {
+			tagName := recognizedResult.TagNames[i]
+			tag, exists := tagMap[tagName]
+
+			if exists {
+				tagIds = append(tagIds, utils.Int64ToString(tag.TagId))
+			}
+		}
+
+		recognizedReceiptImageResponse.TagIds = tagIds
+	}
+
+	if len(recognizedResult.Description) > 0 {
+		recognizedReceiptImageResponse.Comment = recognizedResult.Description
+	}
+
+	return recognizedReceiptImageResponse, nil
+}
--- a/pkg/api/server_settings.go
+++ b/pkg/api/server_settings.go
@@ -47,6 +47,12 @@ func (a *ServerSettingsApi) ServerSettingsJavascriptHandler(c *core.WebContext)
 		a.appendBooleanSetting(builder, "mcp", config.EnableMCPServer)
 	}

+	if config.LLMProvider != "" {
+		if config.TransactionFromAIImageRecognition {
+			a.appendBooleanSetting(builder, "llmt", config.TransactionFromAIImageRecognition)
+		}
+	}
+
 	if config.LoginPageTips.Enabled {
 		a.appendMultiLanguageTipSetting(builder, "lpt", config.LoginPageTips)
 	}
--- a/pkg/core/user_feature_restriction.go
+++ b/pkg/core/user_feature_restriction.go
@@ -76,19 +76,20 @@ type UserFeatureRestrictionType uint64

 // User Feature Restriction Type
 const (
-	USER_FEATURE_RESTRICTION_TYPE_UPDATE_PASSWORD           UserFeatureRestrictionType = 1
-	USER_FEATURE_RESTRICTION_TYPE_UPDATE_EMAIL              UserFeatureRestrictionType = 2
-	USER_FEATURE_RESTRICTION_TYPE_UPDATE_PROFILE_BASIC_INFO UserFeatureRestrictionType = 3
-	USER_FEATURE_RESTRICTION_TYPE_UPDATE_AVATAR             UserFeatureRestrictionType = 4
-	USER_FEATURE_RESTRICTION_TYPE_REVOKE_OTHER_SESSION      UserFeatureRestrictionType = 5
-	USER_FEATURE_RESTRICTION_TYPE_ENABLE_2FA                UserFeatureRestrictionType = 6
-	USER_FEATURE_RESTRICTION_TYPE_DISABLE_2FA               UserFeatureRestrictionType = 7
-	USER_FEATURE_RESTRICTION_TYPE_FORGET_PASSWORD           UserFeatureRestrictionType = 8
-	USER_FEATURE_RESTRICTION_TYPE_IMPORT_TRANSACTION        UserFeatureRestrictionType = 9
-	USER_FEATURE_RESTRICTION_TYPE_EXPORT_TRANSACTION        UserFeatureRestrictionType = 10
-	USER_FEATURE_RESTRICTION_TYPE_CLEAR_ALL_DATA            UserFeatureRestrictionType = 11
-	USER_FEATURE_RESTRICTION_TYPE_SYNC_APPLICATION_SETTINGS UserFeatureRestrictionType = 12
-	USER_FEATURE_RESTRICTION_TYPE_MCP_ACCESS                UserFeatureRestrictionType = 13
+	USER_FEATURE_RESTRICTION_TYPE_UPDATE_PASSWORD                              UserFeatureRestrictionType = 1
+	USER_FEATURE_RESTRICTION_TYPE_UPDATE_EMAIL                                 UserFeatureRestrictionType = 2
+	USER_FEATURE_RESTRICTION_TYPE_UPDATE_PROFILE_BASIC_INFO                    UserFeatureRestrictionType = 3
+	USER_FEATURE_RESTRICTION_TYPE_UPDATE_AVATAR                                UserFeatureRestrictionType = 4
+	USER_FEATURE_RESTRICTION_TYPE_REVOKE_OTHER_SESSION                         UserFeatureRestrictionType = 5
+	USER_FEATURE_RESTRICTION_TYPE_ENABLE_2FA                                   UserFeatureRestrictionType = 6
+	USER_FEATURE_RESTRICTION_TYPE_DISABLE_2FA                                  UserFeatureRestrictionType = 7
+	USER_FEATURE_RESTRICTION_TYPE_FORGET_PASSWORD                              UserFeatureRestrictionType = 8
+	USER_FEATURE_RESTRICTION_TYPE_IMPORT_TRANSACTION                           UserFeatureRestrictionType = 9
+	USER_FEATURE_RESTRICTION_TYPE_EXPORT_TRANSACTION                           UserFeatureRestrictionType = 10
+	USER_FEATURE_RESTRICTION_TYPE_CLEAR_ALL_DATA                               UserFeatureRestrictionType = 11
+	USER_FEATURE_RESTRICTION_TYPE_SYNC_APPLICATION_SETTINGS                    UserFeatureRestrictionType = 12
+	USER_FEATURE_RESTRICTION_TYPE_MCP_ACCESS                                   UserFeatureRestrictionType = 13
+	USER_FEATURE_RESTRICTION_TYPE_CREATE_TRANSACTION_FROM_AI_IMAGE_RECOGNITION UserFeatureRestrictionType = 14
 )

 const userFeatureRestrictionTypeMinValue UserFeatureRestrictionType = USER_FEATURE_RESTRICTION_TYPE_UPDATE_PASSWORD
--- a/pkg/errs/error.go
+++ b/pkg/errs/error.go
@@ -40,6 +40,7 @@ const (
 	NormalSubcategoryConverter              = 12
 	NormalSubcategoryUserCustomExchangeRate = 13
 	NormalSubcategoryModelContextProtocol   = 14
+	NormalSubcategoryLargeLanguageModel     = 15
 )

 // Error represents the specific error returned to user
--- a/pkg/errs/large_language_model.go
+++ b/pkg/errs/large_language_model.go
@@ -0,0 +1,11 @@
+package errs
+
+import "net/http"
+
+// Error codes related to large language model features
+var (
+	ErrLargeLanguageModelProviderNotEnabled = NewNormalError(NormalSubcategoryLargeLanguageModel, 0, http.StatusBadRequest, "llm provider is not enabled")
+	ErrNoAIRecognitionImage                 = NewNormalError(NormalSubcategoryLargeLanguageModel, 1, http.StatusBadRequest, "no image for AI recognition")
+	ErrAIRecognitionImageIsEmpty            = NewNormalError(NormalSubcategoryLargeLanguageModel, 2, http.StatusBadRequest, "image for AI recognition is empty")
+	ErrExceedMaxAIRecognitionImageFileSize  = NewNormalError(NormalSubcategoryLargeLanguageModel, 3, http.StatusBadRequest, "exceed the maximum size of image file for AI recognition")
+)
--- a/pkg/errs/setting.go
+++ b/pkg/errs/setting.go
@@ -24,4 +24,6 @@ var (
 	ErrInvalidPasswordResetTokenExpiredTime           = NewSystemError(SystemSubcategorySetting, 17, http.StatusInternalServerError, "invalid password reset token expired time")
 	ErrInvalidExchangeRatesDataSource                 = NewSystemError(SystemSubcategorySetting, 18, http.StatusInternalServerError, "invalid exchange rates data source")
 	ErrInvalidIpAddressPattern                        = NewSystemError(SystemSubcategorySetting, 19, http.StatusInternalServerError, "invalid ip address pattern")
+	ErrInvalidLLMProvider                             = NewSystemError(SystemSubcategorySetting, 20, http.StatusInternalServerError, "invalid llm provider")
+	ErrInvalidLLMModelId                              = NewSystemError(SystemSubcategorySetting, 21, http.StatusInternalServerError, "invalid llm model id")
 )
--- a/pkg/llm/http_large_language_model_provider.go
+++ b/pkg/llm/http_large_language_model_provider.go
@@ -0,0 +1,91 @@
+package llm
+
+import (
+	"crypto/tls"
+	"io"
+	"net/http"
+	"time"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/errs"
+	"github.com/mayswind/ezbookkeeping/pkg/log"
+	"github.com/mayswind/ezbookkeeping/pkg/settings"
+	"github.com/mayswind/ezbookkeeping/pkg/utils"
+)
+
+// HttpLargeLanguageModelProvider defines the structure of http large language model provider
+type HttpLargeLanguageModelProvider interface {
+	// BuildTextualRequest returns the http request by the provider api definition
+	BuildTextualRequest(c core.Context, uid int64, request *LargeLanguageModelRequest, modelId string, responseType LargeLanguageModelResponseFormat) (*http.Request, error)
+
+	// ParseTextualResponse returns the textual response entity by the provider api definition
+	ParseTextualResponse(c core.Context, uid int64, body []byte, responseType LargeLanguageModelResponseFormat) (*LargeLanguageModelTextualResponse, error)
+
+	// GetReceiptImageRecognitionModelID returns the receipt image recognition model id if supported, otherwise returns empty string
+	GetReceiptImageRecognitionModelID() string
+}
+
+// CommonHttpLargeLanguageModelProvider defines the structure of common http large language model provider
+type CommonHttpLargeLanguageModelProvider struct {
+	LargeLanguageModelProvider
+	provider HttpLargeLanguageModelProvider
+}
+
+// GetJsonResponseByReceiptImageRecognitionModel returns the json response from the OpenAI common compatible large language model provider
+func (p *CommonHttpLargeLanguageModelProvider) GetJsonResponseByReceiptImageRecognitionModel(c core.Context, uid int64, currentConfig *settings.Config, request *LargeLanguageModelRequest) (*LargeLanguageModelTextualResponse, error) {
+	return p.getTextualResponse(c, uid, currentConfig, request, p.provider.GetReceiptImageRecognitionModelID(), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+}
+
+func (p *CommonHttpLargeLanguageModelProvider) getTextualResponse(c core.Context, uid int64, currentConfig *settings.Config, request *LargeLanguageModelRequest, modelId string, responseType LargeLanguageModelResponseFormat) (*LargeLanguageModelTextualResponse, error) {
+	if modelId == "" {
+		return nil, errs.ErrInvalidLLMModelId
+	}
+
+	transport := http.DefaultTransport.(*http.Transport).Clone()
+	utils.SetProxyUrl(transport, currentConfig.LargeLanguageModelAPIProxy)
+
+	if currentConfig.LargeLanguageModelAPISkipTLSVerify {
+		transport.TLSClientConfig = &tls.Config{
+			InsecureSkipVerify: true,
+		}
+	}
+
+	client := &http.Client{
+		Transport: transport,
+		Timeout:   time.Duration(currentConfig.LargeLanguageModelAPIRequestTimeout) * time.Millisecond,
+	}
+
+	httpRequest, err := p.provider.BuildTextualRequest(c, uid, request, modelId, responseType)
+
+	if err != nil {
+		log.Errorf(c, "[http_large_language_model_provider.getTextualResponse] failed to build requests for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	httpRequest.Header.Set("User-Agent", settings.GetUserAgent())
+
+	resp, err := client.Do(httpRequest)
+
+	if err != nil {
+		log.Errorf(c, "[http_large_language_model_provider.getTextualResponse] failed to request large language model api for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	defer resp.Body.Close()
+	body, err := io.ReadAll(resp.Body)
+
+	log.Debugf(c, "[http_large_language_model_provider.getTextualResponse] response is %s", body)
+
+	if resp.StatusCode != 200 {
+		log.Errorf(c, "[http_large_language_model_provider.getTextualResponse] failed to get large language model api response for user \"uid:%d\", because response code is %d", uid, resp.StatusCode)
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	return p.provider.ParseTextualResponse(c, uid, body, responseType)
+}
+
+func newCommonHttpLargeLanguageModelProvider(provider HttpLargeLanguageModelProvider) *CommonHttpLargeLanguageModelProvider {
+	return &CommonHttpLargeLanguageModelProvider{
+		provider: provider,
+	}
+}
--- a/pkg/llm/large_language_model_data.go
+++ b/pkg/llm/large_language_model_data.go
@@ -0,0 +1,33 @@
+package llm
+
+import "reflect"
+
+type LargeLanguageModelRequestPromptType byte
+
+// Large Language Model Request Prompt Type
+const (
+	LARGE_LANGUAGE_MODEL_REQUEST_PROMPT_TYPE_TEXT      LargeLanguageModelRequestPromptType = 0
+	LARGE_LANGUAGE_MODEL_REQUEST_PROMPT_TYPE_IMAGE_URL LargeLanguageModelRequestPromptType = 1
+)
+
+type LargeLanguageModelResponseFormat byte
+
+// Large Language Model Response Format
+const (
+	LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_TEXT LargeLanguageModelResponseFormat = 0
+	LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON LargeLanguageModelResponseFormat = 1
+)
+
+// LargeLanguageModelRequest represents a request to a large language model
+type LargeLanguageModelRequest struct {
+	Stream                 bool
+	SystemPrompt           string
+	UserPrompt             []byte
+	UserPromptType         LargeLanguageModelRequestPromptType
+	ResponseJsonObjectType reflect.Type
+}
+
+// LargeLanguageModelTextualResponse represents a textual response from a large language model
+type LargeLanguageModelTextualResponse struct {
+	Content string
+}
--- a/pkg/llm/large_language_model_provider.go
+++ b/pkg/llm/large_language_model_provider.go
@@ -0,0 +1,12 @@
+package llm
+
+import (
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/settings"
+)
+
+// LargeLanguageModelProvider defines the structure of large language model provider
+type LargeLanguageModelProvider interface {
+	// GetJsonResponseByReceiptImageRecognitionModel returns the json response from the large language model provider by receipt image recognition model
+	GetJsonResponseByReceiptImageRecognitionModel(c core.Context, uid int64, currentConfig *settings.Config, request *LargeLanguageModelRequest) (*LargeLanguageModelTextualResponse, error)
+}
--- a/pkg/llm/large_language_model_provider_container.go
+++ b/pkg/llm/large_language_model_provider_container.go
@@ -0,0 +1,45 @@
+package llm
+
+import (
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/errs"
+	"github.com/mayswind/ezbookkeeping/pkg/settings"
+)
+
+// LargeLanguageModelProviderContainer contains the current large language model provider
+type LargeLanguageModelProviderContainer struct {
+	current LargeLanguageModelProvider
+}
+
+// Initialize a large language model provider container singleton instance
+var (
+	Container = &LargeLanguageModelProviderContainer{}
+)
+
+// InitializeLargeLanguageModelProvider initializes the current large language model provider according to the config
+func InitializeLargeLanguageModelProvider(config *settings.Config) error {
+	if config.LLMProvider == settings.OpenAILLMProvider {
+		Container.current = NewOpenAILargeLanguageModelProvider(config)
+		return nil
+	} else if config.LLMProvider == settings.OpenAICompatibleLLMProvider {
+		Container.current = NewOpenAICompatibleLargeLanguageModelProvider(config)
+		return nil
+	} else if config.LLMProvider == settings.OpenRouterLLMProvider {
+		Container.current = NewOpenRouterLargeLanguageModelProvider(config)
+		return nil
+	} else if config.LLMProvider == settings.OllamaLLMProvider {
+		Container.current = NewOllamaLargeLanguageModelProvider(config)
+		return nil
+	}
+
+	return errs.ErrInvalidLLMProvider
+}
+
+// GetJsonResponseByReceiptImageRecognitionModel returns the json response from the current large language model provider by receipt image recognition model
+func (l *LargeLanguageModelProviderContainer) GetJsonResponseByReceiptImageRecognitionModel(c core.Context, uid int64, currentConfig *settings.Config, request *LargeLanguageModelRequest) (*LargeLanguageModelTextualResponse, error) {
+	if Container.current == nil {
+		return nil, errs.ErrInvalidLLMProvider
+	}
+
+	return l.current.GetJsonResponseByReceiptImageRecognitionModel(c, uid, currentConfig, request)
+}
--- a/pkg/llm/ollama_large_language_model_provider.go
+++ b/pkg/llm/ollama_large_language_model_provider.go
@@ -0,0 +1,153 @@
+package llm
+
+import (
+	"bytes"
+	"encoding/base64"
+	"encoding/json"
+	"net/http"
+	"strings"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/errs"
+	"github.com/mayswind/ezbookkeeping/pkg/log"
+	"github.com/mayswind/ezbookkeeping/pkg/settings"
+)
+
+const ollamaChatCompletionsPath = "api/chat"
+
+// OllamaLargeLanguageModelProvider defines the structure of Ollama large language model provider
+type OllamaLargeLanguageModelProvider struct {
+	CommonHttpLargeLanguageModelProvider
+	OllamaServerURL                string
+	ReceiptImageRecognitionModelID string
+}
+
+// BuildTextualRequest returns the http request by Ollama provider
+func (p *OllamaLargeLanguageModelProvider) BuildTextualRequest(c core.Context, uid int64, request *LargeLanguageModelRequest, modelId string, responseType LargeLanguageModelResponseFormat) (*http.Request, error) {
+	requestBody, err := p.buildJsonRequestBody(c, uid, request, modelId, responseType)
+
+	if err != nil {
+		return nil, err
+	}
+
+	httpRequest, err := http.NewRequest("POST", p.getOllamaRequestUrl(), bytes.NewReader(requestBody))
+
+	if err != nil {
+		return nil, err
+	}
+
+	httpRequest.Header.Set("Content-Type", "application/json")
+
+	return httpRequest, nil
+}
+
+// ParseTextualResponse returns the textual response by Ollama provider
+func (p *OllamaLargeLanguageModelProvider) ParseTextualResponse(c core.Context, uid int64, body []byte, responseType LargeLanguageModelResponseFormat) (*LargeLanguageModelTextualResponse, error) {
+	responseBody := make(map[string]any)
+	err := json.Unmarshal(body, &responseBody)
+
+	if err != nil {
+		log.Errorf(c, "[ollama_large_language_model_provider.ParseTextualResponse] failed to parse response for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	message, ok := responseBody["message"].(map[string]any)
+
+	if !ok {
+		log.Errorf(c, "[ollama_large_language_model_provider.ParseTextualResponse] no message found in response for user \"uid:%d\"", uid)
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	content, ok := message["content"].(string)
+
+	if !ok {
+		log.Errorf(c, "[ollama_large_language_model_provider.ParseTextualResponse] no content found in message for user \"uid:%d\"", uid)
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	if responseType == LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON {
+		if strings.HasPrefix(content, "```json") && strings.HasSuffix(content, "```") {
+			content = strings.TrimPrefix(content, "```json")
+			content = strings.TrimSuffix(content, "```")
+		} else if strings.HasPrefix(content, "```") && strings.HasSuffix(content, "```") {
+			content = strings.TrimPrefix(content, "```")
+			content = strings.TrimSuffix(content, "```")
+		}
+	}
+
+	textualResponse := &LargeLanguageModelTextualResponse{
+		Content: content,
+	}
+
+	return textualResponse, nil
+}
+
+// GetReceiptImageRecognitionModelID returns the receipt image recognition model id of Ollama provider
+func (p *OllamaLargeLanguageModelProvider) GetReceiptImageRecognitionModelID() string {
+	return p.ReceiptImageRecognitionModelID
+}
+
+func (p *OllamaLargeLanguageModelProvider) buildJsonRequestBody(c core.Context, uid int64, request *LargeLanguageModelRequest, modelId string, responseType LargeLanguageModelResponseFormat) ([]byte, error) {
+	requestMessages := make([]any, 0)
+
+	if request.SystemPrompt != "" {
+		requestMessages = append(requestMessages, map[string]string{
+			"role":    "system",
+			"content": request.SystemPrompt,
+		})
+	}
+
+	if len(request.UserPrompt) > 0 {
+		imageBase64Data := base64.StdEncoding.EncodeToString(request.UserPrompt)
+		if request.UserPromptType == LARGE_LANGUAGE_MODEL_REQUEST_PROMPT_TYPE_IMAGE_URL {
+			requestMessages = append(requestMessages, map[string]any{
+				"role":    "user",
+				"content": "",
+				"images":  []string{imageBase64Data},
+			})
+		} else {
+			requestMessages = append(requestMessages, map[string]string{
+				"role":    "user",
+				"content": string(request.UserPrompt),
+			})
+		}
+	}
+
+	requestBody := make(map[string]any)
+	requestBody["model"] = modelId
+	requestBody["stream"] = request.Stream
+	requestBody["messages"] = requestMessages
+
+	if responseType == LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON {
+		requestBody["format"] = "json"
+	}
+
+	requestBodyBytes, err := json.Marshal(requestBody)
+
+	if err != nil {
+		log.Errorf(c, "[ollama_large_language_model_provider.buildJsonRequestBody] failed to marshal request body for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrOperationFailed
+	}
+
+	log.Debugf(c, "[ollama_large_language_model_provider.buildJsonRequestBody] request body is %s", requestBodyBytes)
+	return requestBodyBytes, nil
+}
+
+func (p *OllamaLargeLanguageModelProvider) getOllamaRequestUrl() string {
+	url := p.OllamaServerURL
+
+	if url[len(url)-1] != '/' {
+		url += "/"
+	}
+
+	url += ollamaChatCompletionsPath
+	return url
+}
+
+// NewOllamaLargeLanguageModelProvider creates a new Ollama large language model provider instance
+func NewOllamaLargeLanguageModelProvider(config *settings.Config) LargeLanguageModelProvider {
+	return newCommonHttpLargeLanguageModelProvider(&OllamaLargeLanguageModelProvider{
+		OllamaServerURL:                config.OllamaServerURL,
+		ReceiptImageRecognitionModelID: config.OllamaReceiptImageRecognitionModelID,
+	})
+}
--- a/pkg/llm/ollama_large_language_model_provider_test.go
+++ b/pkg/llm/ollama_large_language_model_provider_test.go
@@ -0,0 +1,138 @@
+package llm
+
+import (
+	"encoding/json"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+)
+
+func TestOllamaLargeLanguageModelProvider_buildJsonRequestBody_TextualUserPrompt(t *testing.T) {
+	provider := &OllamaLargeLanguageModelProvider{}
+
+	request := &LargeLanguageModelRequest{
+		SystemPrompt: "You are a helpful assistant.",
+		UserPrompt:   []byte("Hello, how are you?"),
+	}
+
+	bodyBytes, err := provider.buildJsonRequestBody(core.NewNullContext(), 0, request, "test", LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.Nil(t, err)
+
+	var body map[string]interface{}
+	err = json.Unmarshal(bodyBytes, &body)
+	assert.Nil(t, err)
+
+	assert.Equal(t, "{\"format\":\"json\",\"messages\":[{\"content\":\"You are a helpful assistant.\",\"role\":\"system\"},{\"content\":\"Hello, how are you?\",\"role\":\"user\"}],\"model\":\"test\",\"stream\":false}", string(bodyBytes))
+}
+
+func TestOllamaLargeLanguageModelProvider_buildJsonRequestBody_ImageUserPrompt(t *testing.T) {
+	provider := &OllamaLargeLanguageModelProvider{}
+
+	request := &LargeLanguageModelRequest{
+		SystemPrompt:   "What's in this image?",
+		UserPrompt:     []byte("fakedata"),
+		UserPromptType: LARGE_LANGUAGE_MODEL_REQUEST_PROMPT_TYPE_IMAGE_URL,
+	}
+
+	bodyBytes, err := provider.buildJsonRequestBody(core.NewNullContext(), 0, request, "test", LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.Nil(t, err)
+
+	var body map[string]interface{}
+	err = json.Unmarshal(bodyBytes, &body)
+	assert.Nil(t, err)
+
+	assert.Equal(t, "{\"format\":\"json\",\"messages\":[{\"content\":\"What's in this image?\",\"role\":\"system\"},{\"content\":\"\",\"images\":[\"ZmFrZWRhdGE=\"],\"role\":\"user\"}],\"model\":\"test\",\"stream\":false}", string(bodyBytes))
+}
+
+func TestOllamaLargeLanguageModelProvider_ParseTextualResponse_ValidJsonResponse(t *testing.T) {
+	provider := &OllamaLargeLanguageModelProvider{}
+
+	response := `{
+		"model": "test",
+		"created_at": "2025-09-01T01:02:03.456789Z",
+		"message": {
+			"role": "assistant",
+			"content": "This is a test response"
+		}
+	}`
+
+	result, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.Nil(t, err)
+	assert.Equal(t, "This is a test response", result.Content)
+}
+
+func TestOllamaLargeLanguageModelProvider_ParseTextualResponse_EmptyResponse(t *testing.T) {
+	provider := &OllamaLargeLanguageModelProvider{}
+
+	response := `{
+		"model": "test",
+		"created_at": "2025-09-01T01:02:03.456789Z",
+		"message": {
+			"role": "assistant",
+			"content": ""
+		}
+	}`
+
+	result, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.Nil(t, err)
+	assert.Equal(t, "", result.Content)
+}
+
+func TestOllamaLargeLanguageModelProvider_ParseTextualResponse_EmptyChoices(t *testing.T) {
+	provider := &OllamaLargeLanguageModelProvider{}
+
+	response := `{
+		"model": "test",
+		"created_at": "2025-09-01T01:02:03.456789Z",
+		"message": {}
+	}`
+
+	_, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.EqualError(t, err, "failed to request third party api")
+}
+
+func TestOllamaLargeLanguageModelProvider_ParseTextualResponse_NoChoiceContent(t *testing.T) {
+	provider := &OllamaLargeLanguageModelProvider{}
+
+	response := `{
+		"model": "test",
+		"created_at": "2025-09-01T01:02:03.456789Z",
+		"message": {
+			"role": "assistant"
+		}
+	}`
+
+	_, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.EqualError(t, err, "failed to request third party api")
+}
+
+func TestOllamaLargeLanguageModelProvider_ParseTextualResponse_InvalidJson(t *testing.T) {
+	provider := &OllamaLargeLanguageModelProvider{}
+
+	response := "error"
+
+	_, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.EqualError(t, err, "failed to request third party api")
+}
+
+func TestOllamaLargeLanguageModelProvider_GetOllamaRequestUrl(t *testing.T) {
+	provider := &OllamaLargeLanguageModelProvider{
+		OllamaServerURL: "http://localhost:11434/",
+	}
+	url := provider.getOllamaRequestUrl()
+	assert.Equal(t, "http://localhost:11434/api/chat", url)
+
+	provider = &OllamaLargeLanguageModelProvider{
+		OllamaServerURL: "http://localhost:11434",
+	}
+	url = provider.getOllamaRequestUrl()
+	assert.Equal(t, "http://localhost:11434/api/chat", url)
+
+	provider = &OllamaLargeLanguageModelProvider{
+		OllamaServerURL: "http://example.com/ollama/",
+	}
+	url = provider.getOllamaRequestUrl()
+	assert.Equal(t, "http://example.com/ollama/api/chat", url)
+}
--- a/pkg/llm/openai_common_compatible_large_language_model_provider.go
+++ b/pkg/llm/openai_common_compatible_large_language_model_provider.go
@@ -0,0 +1,187 @@
+package llm
+
+import (
+	"bytes"
+	"encoding/base64"
+	"encoding/json"
+	"io"
+	"net/http"
+	"strings"
+
+	"github.com/invopop/jsonschema"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/errs"
+	"github.com/mayswind/ezbookkeeping/pkg/log"
+)
+
+// OpenAIChatCompletionsLargeLanguageModelProvider defines the structure of OpenAI chat completions compatible large language model provider
+type OpenAIChatCompletionsLargeLanguageModelProvider interface {
+	// BuildChatCompletionsHttpRequest returns the chat completions http request
+	BuildChatCompletionsHttpRequest(c core.Context, uid int64) (*http.Request, error)
+
+	// GetReceiptImageRecognitionModelID returns the receipt image recognition model id if supported, otherwise returns empty string
+	GetReceiptImageRecognitionModelID() string
+}
+
+// OpenAICommonChatCompletionsHttpLargeLanguageModelProvider defines the structure of OpenAI common compatible large language model provider based on chat completions api
+type OpenAICommonChatCompletionsHttpLargeLanguageModelProvider struct {
+	CommonHttpLargeLanguageModelProvider
+	provider OpenAIChatCompletionsLargeLanguageModelProvider
+}
+
+// BuildTextualRequest returns the http request by OpenAI common compatible provider
+func (p *OpenAICommonChatCompletionsHttpLargeLanguageModelProvider) BuildTextualRequest(c core.Context, uid int64, request *LargeLanguageModelRequest, modelId string, responseType LargeLanguageModelResponseFormat) (*http.Request, error) {
+	requestBody, err := p.buildJsonRequestBody(c, uid, request, modelId, responseType)
+
+	if err != nil {
+		return nil, err
+	}
+
+	httpRequest, err := p.provider.BuildChatCompletionsHttpRequest(c, uid)
+
+	if err != nil {
+		return nil, err
+	}
+
+	httpRequest.Body = io.NopCloser(bytes.NewReader(requestBody))
+	httpRequest.Header.Set("Content-Type", "application/json")
+
+	return httpRequest, nil
+}
+
+// ParseTextualResponse returns the textual response by OpenAI common compatible provider
+func (p *OpenAICommonChatCompletionsHttpLargeLanguageModelProvider) ParseTextualResponse(c core.Context, uid int64, body []byte, responseType LargeLanguageModelResponseFormat) (*LargeLanguageModelTextualResponse, error) {
+	responseBody := make(map[string]any)
+	err := json.Unmarshal(body, &responseBody)
+
+	if err != nil {
+		log.Errorf(c, "[openai_common_compatible_large_language_model_provider.ParseTextualResponse] failed to parse response for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	choices, ok := responseBody["choices"].([]any)
+
+	if !ok || len(choices) < 1 {
+		log.Errorf(c, "[openai_common_compatible_large_language_model_provider.ParseTextualResponse] no choices found in response for user \"uid:%d\"", uid)
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	firstChoice, ok := choices[0].(map[string]any)
+
+	if !ok {
+		log.Errorf(c, "[openai_common_compatible_large_language_model_provider.ParseTextualResponse] invalid choice format in response for user \"uid:%d\"", uid)
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	message, ok := firstChoice["message"].(map[string]any)
+
+	if !ok {
+		log.Errorf(c, "[openai_common_compatible_large_language_model_provider.ParseTextualResponse] no message found in choice for user \"uid:%d\"", uid)
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	content, ok := message["content"].(string)
+
+	if !ok {
+		log.Errorf(c, "[openai_common_compatible_large_language_model_provider.ParseTextualResponse] no content found in message for user \"uid:%d\"", uid)
+		return nil, errs.ErrFailedToRequestRemoteApi
+	}
+
+	if responseType == LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON {
+		if strings.HasPrefix(content, "```json") && strings.HasSuffix(content, "```") {
+			content = strings.TrimPrefix(content, "```json")
+			content = strings.TrimSuffix(content, "```")
+		} else if strings.HasPrefix(content, "```") && strings.HasSuffix(content, "```") {
+			content = strings.TrimPrefix(content, "```")
+			content = strings.TrimSuffix(content, "```")
+		}
+	}
+
+	textualResponse := &LargeLanguageModelTextualResponse{
+		Content: content,
+	}
+
+	return textualResponse, nil
+}
+
+// GetReceiptImageRecognitionModelID returns the receipt image recognition model id of OpenAI common compatible provider
+func (p *OpenAICommonChatCompletionsHttpLargeLanguageModelProvider) GetReceiptImageRecognitionModelID() string {
+	return p.provider.GetReceiptImageRecognitionModelID()
+}
+
+func (p *OpenAICommonChatCompletionsHttpLargeLanguageModelProvider) buildJsonRequestBody(c core.Context, uid int64, request *LargeLanguageModelRequest, modelId string, responseType LargeLanguageModelResponseFormat) ([]byte, error) {
+	requestMessages := make([]any, 0)
+
+	if request.SystemPrompt != "" {
+		requestMessages = append(requestMessages, map[string]string{
+			"role":    "system",
+			"content": request.SystemPrompt,
+		})
+	}
+
+	if len(request.UserPrompt) > 0 {
+		if request.UserPromptType == LARGE_LANGUAGE_MODEL_REQUEST_PROMPT_TYPE_IMAGE_URL {
+			imageBase64Data := "data:image/png;base64," + base64.StdEncoding.EncodeToString(request.UserPrompt)
+			requestMessages = append(requestMessages, map[string]any{
+				"role": "user",
+				"content": []any{
+					core.O{
+						"type": "image_url",
+						"image_url": core.O{
+							"url": imageBase64Data,
+						},
+					},
+				},
+			})
+		} else {
+			requestMessages = append(requestMessages, map[string]string{
+				"role":    "user",
+				"content": string(request.UserPrompt),
+			})
+		}
+	}
+
+	requestBody := make(map[string]any)
+	requestBody["model"] = modelId
+	requestBody["stream"] = request.Stream
+	requestBody["messages"] = requestMessages
+
+	if responseType == LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON {
+		if request.ResponseJsonObjectType != nil {
+			schemeGenerator := jsonschema.Reflector{
+				Anonymous:      true,
+				DoNotReference: true,
+				ExpandedStruct: true,
+			}
+
+			schema := schemeGenerator.ReflectFromType(request.ResponseJsonObjectType)
+			schema.Version = ""
+
+			requestBody["response_format"] = core.O{
+				"type":        "json_schema",
+				"json_schema": schema,
+			}
+		} else {
+			requestBody["response_format"] = core.O{
+				"type": "json_object",
+			}
+		}
+	}
+
+	requestBodyBytes, err := json.Marshal(requestBody)
+
+	if err != nil {
+		log.Errorf(c, "[openai_common_compatible_large_language_model_provider.buildJsonRequestBody] failed to marshal request body for user \"uid:%d\", because %s", uid, err.Error())
+		return nil, errs.ErrOperationFailed
+	}
+
+	log.Debugf(c, "[openai_common_compatible_large_language_model_provider.buildJsonRequestBody] request body is %s", requestBodyBytes)
+	return requestBodyBytes, nil
+}
+
+func newOpenAICommonChatCompletionsHttpLargeLanguageModelProvider(provider OpenAIChatCompletionsLargeLanguageModelProvider) LargeLanguageModelProvider {
+	return newCommonHttpLargeLanguageModelProvider(&OpenAICommonChatCompletionsHttpLargeLanguageModelProvider{
+		provider: provider,
+	})
+}
--- a/pkg/llm/openai_common_compatible_large_language_model_provider_test.go
+++ b/pkg/llm/openai_common_compatible_large_language_model_provider_test.go
@@ -0,0 +1,157 @@
+package llm
+
+import (
+	"encoding/json"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+)
+
+func TestOpenAICommonChatCompletionsHttpLargeLanguageModelProvider_buildJsonRequestBody_TextualUserPrompt(t *testing.T) {
+	provider := &OpenAICommonChatCompletionsHttpLargeLanguageModelProvider{
+		provider: &OpenAILargeLanguageModelProvider{},
+	}
+
+	request := &LargeLanguageModelRequest{
+		SystemPrompt: "You are a helpful assistant.",
+		UserPrompt:   []byte("Hello, how are you?"),
+	}
+
+	bodyBytes, err := provider.buildJsonRequestBody(core.NewNullContext(), 0, request, "test", LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.Nil(t, err)
+
+	var body map[string]interface{}
+	err = json.Unmarshal(bodyBytes, &body)
+	assert.Nil(t, err)
+
+	assert.Equal(t, "{\"messages\":[{\"content\":\"You are a helpful assistant.\",\"role\":\"system\"},{\"content\":\"Hello, how are you?\",\"role\":\"user\"}],\"model\":\"test\",\"response_format\":{\"type\":\"json_object\"},\"stream\":false}", string(bodyBytes))
+}
+
+func TestOpenAICommonChatCompletionsHttpLargeLanguageModelProvider_buildJsonRequestBody_ImageUserPrompt(t *testing.T) {
+	provider := &OpenAICommonChatCompletionsHttpLargeLanguageModelProvider{
+		provider: &OpenAILargeLanguageModelProvider{},
+	}
+
+	request := &LargeLanguageModelRequest{
+		SystemPrompt:   "What's in this image?",
+		UserPrompt:     []byte("fakedata"),
+		UserPromptType: LARGE_LANGUAGE_MODEL_REQUEST_PROMPT_TYPE_IMAGE_URL,
+	}
+
+	bodyBytes, err := provider.buildJsonRequestBody(core.NewNullContext(), 0, request, "test", LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.Nil(t, err)
+
+	var body map[string]interface{}
+	err = json.Unmarshal(bodyBytes, &body)
+	assert.Nil(t, err)
+
+	assert.Equal(t, "{\"messages\":[{\"content\":\"What's in this image?\",\"role\":\"system\"},{\"content\":[{\"image_url\":{\"url\":\"data:image/png;base64,ZmFrZWRhdGE=\"},\"type\":\"image_url\"}],\"role\":\"user\"}],\"model\":\"test\",\"response_format\":{\"type\":\"json_object\"},\"stream\":false}", string(bodyBytes))
+}
+
+func TestOpenAICommonChatCompletionsHttpLargeLanguageModelProvider_ParseTextualResponse_ValidJsonResponse(t *testing.T) {
+	provider := &OpenAICommonChatCompletionsHttpLargeLanguageModelProvider{
+		provider: &OpenAILargeLanguageModelProvider{},
+	}
+
+	response := `{
+		"id": "test-123",
+		"object": "chat.completion",
+		"created": 1234567890,
+		"model": "test",
+		"usage": {
+			"prompt_tokens": 13,
+			"completion_tokens": 7,
+			"total_tokens": 20
+		},
+		"choices": [
+			{
+				"finish_reason": "stop",
+				"index": 0,
+				"message": {
+					"role": "assistant",
+					"content": "This is a test response"
+				}
+			}
+		]
+	}`
+
+	result, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.Nil(t, err)
+	assert.Equal(t, "This is a test response", result.Content)
+}
+
+func TestOpenAICommonChatCompletionsHttpLargeLanguageModelProvider_ParseTextualResponse_EmptyResponse(t *testing.T) {
+	provider := &OpenAICommonChatCompletionsHttpLargeLanguageModelProvider{
+		provider: &OpenAILargeLanguageModelProvider{},
+	}
+
+	response := `{
+		"id": "test-123",
+		"object": "chat.completion",
+		"choices": [
+			{
+				"finish_reason": "stop",
+				"index": 0,
+				"message": {
+					"role": "assistant",
+					"content": ""
+				}
+			}
+		]
+	}`
+
+	result, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.Nil(t, err)
+	assert.Equal(t, "", result.Content)
+}
+
+func TestOpenAICommonChatCompletionsHttpLargeLanguageModelProvider_ParseTextualResponse_EmptyChoices(t *testing.T) {
+	provider := &OpenAICommonChatCompletionsHttpLargeLanguageModelProvider{
+		provider: &OpenAILargeLanguageModelProvider{},
+	}
+
+	response := `{
+		"id": "test-123",
+		"object": "chat.completion",
+		"choices": []
+	}`
+
+	_, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.EqualError(t, err, "failed to request third party api")
+}
+
+func TestOpenAICommonChatCompletionsHttpLargeLanguageModelProvider_ParseTextualResponse_NoChoiceContent(t *testing.T) {
+	provider := &OpenAICommonChatCompletionsHttpLargeLanguageModelProvider{
+		provider: &OpenAILargeLanguageModelProvider{},
+	}
+
+	response := `{
+		"id": "chatcmpl-123",
+		"object": "chat.completion",
+		"choices": [
+			{
+				"finish_reason": "stop",
+				"index": 0,
+				"message": {
+					"role": "assistant"
+				}
+			}
+		]
+	}`
+
+	_, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.EqualError(t, err, "failed to request third party api")
+}
+
+func TestOpenAICommonChatCompletionsHttpLargeLanguageModelProvider_ParseTextualResponse_InvalidJson(t *testing.T) {
+	provider := &OpenAICommonChatCompletionsHttpLargeLanguageModelProvider{
+		provider: &OpenAILargeLanguageModelProvider{},
+	}
+
+	response := "error"
+
+	_, err := provider.ParseTextualResponse(core.NewNullContext(), 0, []byte(response), LARGE_LANGUAGE_MODEL_RESPONSE_FORMAT_JSON)
+	assert.EqualError(t, err, "failed to request third party api")
+}
--- a/pkg/llm/openai_compatible_large_language_model_provider.go
+++ b/pkg/llm/openai_compatible_large_language_model_provider.go
@@ -0,0 +1,58 @@
+package llm
+
+import (
+	"net/http"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/settings"
+)
+
+const openAICompatibleChatCompletionsPath = "chat/completions"
+
+// OpenAICompatibleLargeLanguageModelProvider defines the structure of OpenAI compatible large language model provider
+type OpenAICompatibleLargeLanguageModelProvider struct {
+	OpenAIChatCompletionsLargeLanguageModelProvider
+	OpenAICompatibleBaseURL        string
+	OpenAICompatibleAPIKey         string
+	ReceiptImageRecognitionModelID string
+}
+
+// BuildChatCompletionsHttpRequest returns the chat completions http request by OpenAI compatible provider
+func (p *OpenAICompatibleLargeLanguageModelProvider) BuildChatCompletionsHttpRequest(c core.Context, uid int64) (*http.Request, error) {
+	req, err := http.NewRequest("POST", p.getFinalChatCompletionsRequestUrl(), nil)
+
+	if err != nil {
+		return nil, err
+	}
+
+	if p.OpenAICompatibleAPIKey != "" {
+		req.Header.Set("Authorization", "Bearer "+p.OpenAICompatibleAPIKey)
+	}
+
+	return req, nil
+}
+
+// GetReceiptImageRecognitionModelID returns the receipt image recognition model id of OpenAI compatible provider
+func (p *OpenAICompatibleLargeLanguageModelProvider) GetReceiptImageRecognitionModelID() string {
+	return p.ReceiptImageRecognitionModelID
+}
+
+func (p *OpenAICompatibleLargeLanguageModelProvider) getFinalChatCompletionsRequestUrl() string {
+	url := p.OpenAICompatibleBaseURL
+
+	if url[len(url)-1] != '/' {
+		url += "/"
+	}
+
+	url += openAICompatibleChatCompletionsPath
+	return url
+}
+
+// NewOpenAICompatibleLargeLanguageModelProvider creates a new OpenAI compatible large language model provider instance
+func NewOpenAICompatibleLargeLanguageModelProvider(config *settings.Config) LargeLanguageModelProvider {
+	return newOpenAICommonChatCompletionsHttpLargeLanguageModelProvider(&OpenAICompatibleLargeLanguageModelProvider{
+		OpenAICompatibleBaseURL:        config.OpenAICompatibleBaseURL,
+		OpenAICompatibleAPIKey:         config.OpenAICompatibleAPIKey,
+		ReceiptImageRecognitionModelID: config.OpenAICompatibleReceiptImageRecognitionModelID,
+	})
+}
--- a/pkg/llm/openai_compatible_large_language_model_provider_test.go
+++ b/pkg/llm/openai_compatible_large_language_model_provider_test.go
@@ -0,0 +1,27 @@
+package llm
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestOpenAICompatibleLargeLanguageModelProvider_GetFinalRequestUrl(t *testing.T) {
+	provider := &OpenAICompatibleLargeLanguageModelProvider{
+		OpenAICompatibleBaseURL: "https://api.example.com/v1/",
+	}
+	url := provider.getFinalChatCompletionsRequestUrl()
+	assert.Equal(t, "https://api.example.com/v1/chat/completions", url)
+
+	provider = &OpenAICompatibleLargeLanguageModelProvider{
+		OpenAICompatibleBaseURL: "https://api.example.com/v1",
+	}
+	url = provider.getFinalChatCompletionsRequestUrl()
+	assert.Equal(t, "https://api.example.com/v1/chat/completions", url)
+
+	provider = &OpenAICompatibleLargeLanguageModelProvider{
+		OpenAICompatibleBaseURL: "https://example.com/api",
+	}
+	url = provider.getFinalChatCompletionsRequestUrl()
+	assert.Equal(t, "https://example.com/api/chat/completions", url)
+}
--- a/pkg/llm/openai_large_language_model_provider.go
+++ b/pkg/llm/openai_large_language_model_provider.go
@@ -0,0 +1,43 @@
+package llm
+
+import (
+	"net/http"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/settings"
+)
+
+// OpenAILargeLanguageModelProvider defines the structure of OpenAI large language model provider
+type OpenAILargeLanguageModelProvider struct {
+	OpenAIChatCompletionsLargeLanguageModelProvider
+	OpenAIAPIKey                   string
+	ReceiptImageRecognitionModelID string
+}
+
+const openAIChatCompletionsUrl = "https://api.openai.com/v1/chat/completions"
+
+// BuildChatCompletionsHttpRequest returns the chat completions http request by OpenAI provider
+func (p *OpenAILargeLanguageModelProvider) BuildChatCompletionsHttpRequest(c core.Context, uid int64) (*http.Request, error) {
+	req, err := http.NewRequest("POST", openAIChatCompletionsUrl, nil)
+
+	if err != nil {
+		return nil, err
+	}
+
+	req.Header.Set("Authorization", "Bearer "+p.OpenAIAPIKey)
+
+	return req, nil
+}
+
+// GetReceiptImageRecognitionModelID returns the receipt image recognition model id of OpenAI provider
+func (p *OpenAILargeLanguageModelProvider) GetReceiptImageRecognitionModelID() string {
+	return p.ReceiptImageRecognitionModelID
+}
+
+// NewOpenAILargeLanguageModelProvider creates a new OpenAI large language model provider instance
+func NewOpenAILargeLanguageModelProvider(config *settings.Config) LargeLanguageModelProvider {
+	return newOpenAICommonChatCompletionsHttpLargeLanguageModelProvider(&OpenAILargeLanguageModelProvider{
+		OpenAIAPIKey:                   config.OpenAIAPIKey,
+		ReceiptImageRecognitionModelID: config.OpenAIReceiptImageRecognitionModelID,
+	})
+}
--- a/pkg/llm/openrouter_large_language_model_provider.go
+++ b/pkg/llm/openrouter_large_language_model_provider.go
@@ -0,0 +1,45 @@
+package llm
+
+import (
+	"net/http"
+
+	"github.com/mayswind/ezbookkeeping/pkg/core"
+	"github.com/mayswind/ezbookkeeping/pkg/settings"
+)
+
+// OpenRouterLargeLanguageModelProvider defines the structure of OpenRouter large language model provider
+type OpenRouterLargeLanguageModelProvider struct {
+	OpenAIChatCompletionsLargeLanguageModelProvider
+	OpenRouterAPIKey               string
+	ReceiptImageRecognitionModelID string
+}
+
+const openRouterChatCompletionsUrl = "https://openrouter.ai/api/v1/chat/completions"
+
+// BuildChatCompletionsHttpRequest returns the chat completions http request by OpenRouter provider
+func (p *OpenRouterLargeLanguageModelProvider) BuildChatCompletionsHttpRequest(c core.Context, uid int64) (*http.Request, error) {
+	req, err := http.NewRequest("POST", openRouterChatCompletionsUrl, nil)
+
+	if err != nil {
+		return nil, err
+	}
+
+	req.Header.Set("Authorization", "Bearer "+p.OpenRouterAPIKey)
+	req.Header.Set("HTTP-Referer", "https://ezbookkeeping.mayswind.net/")
+	req.Header.Set("X-Title", "ezBookkeeping")
+
+	return req, nil
+}
+
+// GetReceiptImageRecognitionModelID returns the receipt image recognition model id of OpenRouter provider
+func (p *OpenRouterLargeLanguageModelProvider) GetReceiptImageRecognitionModelID() string {
+	return p.ReceiptImageRecognitionModelID
+}
+
+// NewOpenRouterLargeLanguageModelProvider creates a new OpenRouter large language model provider instance
+func NewOpenRouterLargeLanguageModelProvider(config *settings.Config) LargeLanguageModelProvider {
+	return newOpenAICommonChatCompletionsHttpLargeLanguageModelProvider(&OpenRouterLargeLanguageModelProvider{
+		OpenRouterAPIKey:               config.OpenRouterAPIKey,
+		ReceiptImageRecognitionModelID: config.OpenRouterReceiptImageRecognitionModelID,
+	})
+}
--- a/pkg/models/large_language_model.go
+++ b/pkg/models/large_language_model.go
@@ -0,0 +1,27 @@
+package models
+
+// RecognizedReceiptImageResponse represents a view-object of recognized receipt image response
+type RecognizedReceiptImageResponse struct {
+	Type                 TransactionType `json:"type"`
+	Time                 int64           `json:"time,omitempty"`
+	CategoryId           int64           `json:"categoryId,string,omitempty"`
+	SourceAccountId      int64           `json:"sourceAccountId,string,omitempty"`
+	DestinationAccountId int64           `json:"destinationAccountId,string,omitempty"`
+	SourceAmount         int64           `json:"sourceAmount,omitempty"`
+	DestinationAmount    int64           `json:"destinationAmount,omitempty"`
+	TagIds               []string        `json:"tagIds,omitempty"`
+	Comment              string          `json:"comment,omitempty"`
+}
+
+// RecognizedReceiptImageResult represents the result of recognized receipt image
+type RecognizedReceiptImageResult struct {
+	Type                   string   `json:"type,omitempty" jsonschema:"enum=income,enum=expense,enum=transfer" jsonschema_description:"Transaction type (income, expense, transfer)"`
+	Time                   string   `json:"time" jsonschema:"format=date-time" jsonschema_description:"Transaction time in long date time format (YYYY-MM-DD HH:mm:ss, e.g. 2023-01-01 12:00:00)"`
+	Amount                 string   `json:"amount,omitempty" jsonschema_description:"Transaction amount"`
+	AccountName            string   `json:"account,omitempty" jsonschema_description:"Account name for the transaction"`
+	CategoryName           string   `json:"category,omitempty" jsonschema_description:"Category name for the transaction"`
+	TagNames               []string `json:"tags,omitempty" jsonschema_description:"List of tags associated with the transaction (maximum 10 tags allowed)"`
+	Description            string   `json:"description,omitempty" jsonschema_description:"Transaction description"`
+	DestinationAmount      string   `json:"destination_amount,omitempty" jsonschema_description:"Destination amount for transfer transactions"`
+	DestinationAccountName string   `json:"destination_account,omitempty" jsonschema_description:"Destination account name for transfer transactions"`
+}
--- a/pkg/settings/setting.go
+++ b/pkg/settings/setting.go
@@ -66,6 +66,13 @@ const (
 	WebDAVStorageType                string = "webdav"
 )

+const (
+	OpenAILLMProvider           string = "openai"
+	OpenAICompatibleLLMProvider string = "openai_compatible"
+	OpenRouterLLMProvider       string = "openrouter"
+	OllamaLLMProvider           string = "ollama"
+)
+
 // Uuid generator types
 const (
 	InternalUuidGeneratorType string = "internal"
@@ -140,6 +147,9 @@ const (

 	defaultWebDAVRequestTimeout uint32 = 10000 // 10 seconds

+	defaultAIRecognitionPictureMaxSize         uint32 = 10485760 // 10MB
+	defaultLargeLanguageModelAPIRequestTimeout uint32 = 60000    // 60 seconds
+
 	defaultInMemoryDuplicateCheckerCleanupInterval uint32 = 60  // 1 minutes
 	defaultDuplicateSubmissionsInterval            uint32 = 300 // 5 minutes

@@ -281,6 +291,23 @@ type Config struct {
 	MinIOConfig         *MinIOConfig
 	WebDAVConfig        *WebDAVConfig

+	// Large Language Model
+	LLMProvider                                    string
+	OpenAIAPIKey                                   string
+	OpenAIReceiptImageRecognitionModelID           string
+	OpenAICompatibleBaseURL                        string
+	OpenAICompatibleAPIKey                         string
+	OpenAICompatibleReceiptImageRecognitionModelID string
+	OpenRouterAPIKey                               string
+	OpenRouterReceiptImageRecognitionModelID       string
+	OllamaServerURL                                string
+	OllamaReceiptImageRecognitionModelID           string
+	TransactionFromAIImageRecognition              bool
+	MaxAIRecognitionPictureFileSize                uint32
+	LargeLanguageModelAPIRequestTimeout            uint32
+	LargeLanguageModelAPIProxy                     string
+	LargeLanguageModelAPISkipTLSVerify             bool
+
 	// Uuid
 	UuidGeneratorType string
 	UuidServerId      uint8
@@ -426,6 +453,12 @@ func LoadConfiguration(configFilePath string) (*Config, error) {
 		return nil, err
 	}

+	err = loadLLMConfiguration(config, cfgFile, "llm")
+
+	if err != nil {
+		return nil, err
+	}
+
 	err = loadUuidConfiguration(config, cfgFile, "uuid")

 	if err != nil {
@@ -751,6 +784,46 @@ func loadStorageConfiguration(config *Config, configFile *ini.File, sectionName
 	return nil
 }

+func loadLLMConfiguration(config *Config, configFile *ini.File, sectionName string) error {
+	llmProvider := getConfigItemStringValue(configFile, sectionName, "llm_provider")
+
+	if llmProvider == "" {
+		config.LLMProvider = ""
+	} else if llmProvider == OpenAILLMProvider {
+		config.LLMProvider = OpenAILLMProvider
+	} else if llmProvider == OpenAICompatibleLLMProvider {
+		config.LLMProvider = OpenAICompatibleLLMProvider
+	} else if llmProvider == OpenRouterLLMProvider {
+		config.LLMProvider = OpenRouterLLMProvider
+	} else if llmProvider == OllamaLLMProvider {
+		config.LLMProvider = OllamaLLMProvider
+	} else {
+		return errs.ErrInvalidLLMProvider
+	}
+
+	config.OpenAIAPIKey = getConfigItemStringValue(configFile, sectionName, "openai_api_key")
+	config.OpenAIReceiptImageRecognitionModelID = getConfigItemStringValue(configFile, sectionName, "openai_receipt_image_recognition_model_id")
+
+	config.OpenAICompatibleBaseURL = getConfigItemStringValue(configFile, sectionName, "openai_compatible_base_url")
+	config.OpenAICompatibleAPIKey = getConfigItemStringValue(configFile, sectionName, "openai_compatible_api_key")
+	config.OpenAICompatibleReceiptImageRecognitionModelID = getConfigItemStringValue(configFile, sectionName, "openai_compatible_receipt_image_recognition_model_id")
+
+	config.OpenRouterAPIKey = getConfigItemStringValue(configFile, sectionName, "openrouter_api_key")
+	config.OpenRouterReceiptImageRecognitionModelID = getConfigItemStringValue(configFile, sectionName, "openrouter_receipt_image_recognition_model_id")
+
+	config.OllamaServerURL = getConfigItemStringValue(configFile, sectionName, "ollama_server_url")
+	config.OllamaReceiptImageRecognitionModelID = getConfigItemStringValue(configFile, sectionName, "ollama_receipt_image_recognition_model_id")
+
+	config.TransactionFromAIImageRecognition = getConfigItemBoolValue(configFile, sectionName, "transaction_from_ai_image_recognition", false)
+	config.MaxAIRecognitionPictureFileSize = getConfigItemUint32Value(configFile, sectionName, "max_ai_recognition_picture_size", defaultAIRecognitionPictureMaxSize)
+
+	config.LargeLanguageModelAPIProxy = getConfigItemStringValue(configFile, sectionName, "proxy", "system")
+	config.LargeLanguageModelAPIRequestTimeout = getConfigItemUint32Value(configFile, sectionName, "request_timeout", defaultLargeLanguageModelAPIRequestTimeout)
+	config.LargeLanguageModelAPISkipTLSVerify = getConfigItemBoolValue(configFile, sectionName, "skip_tls_verify", false)
+
+	return nil
+}
+
 func loadUuidConfiguration(config *Config, configFile *ini.File, sectionName string) error {
 	if getConfigItemStringValue(configFile, sectionName, "generator_type") == InternalUuidGeneratorType {
 		config.UuidGeneratorType = InternalUuidGeneratorType
--- a/pkg/templates/known_template.go
+++ b/pkg/templates/known_template.go
@@ -4,6 +4,7 @@ type KnownTemplate string

 // Known templates
 const (
-	TEMPLATE_VERIFY_EMAIL   KnownTemplate = "email/verify_email"
-	TEMPLATE_PASSWORD_RESET KnownTemplate = "email/password_reset"
+	TEMPLATE_VERIFY_EMAIL                   KnownTemplate = "email/verify_email"
+	TEMPLATE_PASSWORD_RESET                 KnownTemplate = "email/password_reset"
+	SYSTEM_PROMPT_RECEIPT_IMAGE_RECOGNITION KnownTemplate = "prompt/receipt_image_recognition"
 )
--- a/src/components/mobile/AIImageRecognitionSheet.vue
+++ b/src/components/mobile/AIImageRecognitionSheet.vue
@@ -0,0 +1,181 @@
+<template>
+    <f7-sheet swipe-to-close swipe-handler=".swipe-handler" style="height:auto"
+              :opened="show" @sheet:open="onSheetOpen" @sheet:closed="onSheetClosed">
+        <f7-toolbar>
+            <div class="swipe-handler"></div>
+            <div class="left">
+                <f7-link :class="{ 'disabled': loading || recognizing }" :text="tt('Choose from Library')" @click="showOpenImage"></f7-link>
+            </div>
+            <div class="right">
+                <f7-link :class="{ 'disabled': loading || recognizing }" :text="tt('Take Photo')" @click="showCamera"></f7-link>
+            </div>
+        </f7-toolbar>
+        <f7-page-content class="margin-top no-padding-top">
+            <div class="padding-horizontal padding-bottom">
+                <div class="image-container display-flex justify-content-center width-100 margin-bottom" style="height: 240px">
+                    <img height="240px" :src="imageSrc" v-if="imageSrc" />
+                    <div class="image-container-background display-flex justify-content-center align-items-center" v-if="!imageSrc">
+                        <span>{{ tt('Please select a receipt or transaction image first') }}</span>
+                    </div>
+                </div>
+                <f7-button large fill color="primary"
+                           :class="{ 'disabled': loading || recognizing || !imageFile }"
+                           :text="tt('Recognize')"
+                           @click="confirm">
+                </f7-button>
+                <div class="margin-top text-align-center">
+                    <f7-link :class="{ 'disabled': loading || recognizing }" @click="cancel" :text="tt('Cancel')"></f7-link>
+                </div>
+            </div>
+        </f7-page-content>
+
+        <input ref="imageInput" type="file" style="display: none" :accept="SUPPORTED_IMAGE_EXTENSIONS" @change="openImage($event)" />
+        <input ref="cameraInput" type="file" style="display: none" :accept="SUPPORTED_IMAGE_EXTENSIONS" capture="environment" @change="openImage($event)" />
+    </f7-sheet>
+</template>
+
+
+<script setup lang="ts">
+import { ref, useTemplateRef } from 'vue';
+
+import { useI18n } from '@/locales/helpers.ts';
+import { useI18nUIComponents, showLoading, hideLoading } from '@/lib/ui/mobile.ts';
+
+import { useTransactionsStore } from '@/stores/transaction.ts';
+
+import { KnownFileType } from '@/core/file.ts';
+import { SUPPORTED_IMAGE_EXTENSIONS } from '@/consts/file.ts';
+
+import type { RecognizedReceiptImageResponse } from '@/models/large_language_model.ts';
+
+import { compressJpgImage } from '@/lib/ui/common.ts';
+import logger from '@/lib/logger.ts';
+
+defineProps<{
+    show: boolean;
+}>();
+
+const emit = defineEmits<{
+    (e: 'update:show', value: boolean): void;
+    (e: 'recognition:change', value: RecognizedReceiptImageResponse): void;
+}>();
+
+const { tt } = useI18n();
+const { showToast } = useI18nUIComponents();
+
+const transactionsStore = useTransactionsStore();
+
+const imageInput = useTemplateRef<HTMLInputElement>('imageInput');
+const cameraInput = useTemplateRef<HTMLInputElement>('cameraInput');
+
+const loading = ref<boolean>(false);
+const recognizing = ref<boolean>(false);
+const imageFile = ref<File | null>(null);
+const imageSrc = ref<string | undefined>(undefined);
+
+function loadImage(file: File): void {
+    compressJpgImage(file, 1280, 1280, 0.8).then(blob => {
+        imageFile.value = KnownFileType.JPG.createFileFromBlob(blob, "image");
+        imageSrc.value = URL.createObjectURL(blob);
+    }).catch(error => {
+        imageFile.value = null;
+        imageSrc.value = undefined;
+        logger.error('failed to compress image', error);
+        showToast('Unable to load image');
+    });
+}
+
+function showOpenImage(): void {
+    if (loading.value || recognizing.value) {
+        return;
+    }
+
+    imageInput.value?.click();
+}
+
+function showCamera(): void {
+    if (loading.value || recognizing.value) {
+        return;
+    }
+
+    cameraInput.value?.click();
+}
+
+function openImage(event: Event): void {
+    if (!event || !event.target) {
+        return;
+    }
+
+    const el = event.target as HTMLInputElement;
+
+    if (!el.files || !el.files.length || !el.files[0]) {
+        return;
+    }
+
+    const image = el.files[0] as File;
+
+    el.value = '';
+
+    loadImage(image);
+}
+
+function confirm(): void {
+    if (loading.value || recognizing.value || !imageFile.value) {
+        return;
+    }
+
+    recognizing.value = true;
+    showLoading(() => recognizing.value);
+
+    transactionsStore.recognizeReceiptImage({
+        imageFile: imageFile.value
+    }).then(response => {
+        recognizing.value = false;
+        hideLoading();
+        emit('update:show', false);
+        emit('recognition:change', response);
+    }).catch(error => {
+        recognizing.value = false;
+        hideLoading();
+
+        if (!error.processed) {
+            showToast(error.message || error);
+        }
+    });
+}
+
+function cancel(): void {
+    close();
+}
+
+function close(): void {
+    emit('update:show', false);
+    loading.value = false;
+    recognizing.value = false;
+    imageFile.value = null;
+    imageSrc.value = undefined;
+}
+
+function onSheetOpen(): void {
+    loading.value = false;
+    recognizing.value = false;
+    imageFile.value = null;
+    imageSrc.value = undefined;
+}
+
+function onSheetClosed(): void {
+    close();
+}
+</script>
+
+<style>
+.image-container {
+    border: 1px solid var(--f7-page-master-border-color);
+}
+
+.image-container-background {
+    width: 100%;
+    height: 100%;
+    background-color: var(--f7-page-bg-color);
+}
+</style>
--- a/src/consts/api.ts
+++ b/src/consts/api.ts
@@ -7,6 +7,7 @@ export const DEFAULT_API_TIMEOUT: number = 10000; // 10s
 export const DEFAULT_UPLOAD_API_TIMEOUT: number = 30000; // 30s
 export const DEFAULT_EXPORT_API_TIMEOUT: number = 180000; // 180s
 export const DEFAULT_IMPORT_API_TIMEOUT: number = 1800000; // 1800s
+export const DEFAULT_LLM_API_TIMEOUT: number = 600000; // 600s

 export const GOOGLE_MAP_JAVASCRIPT_URL: string = 'https://maps.googleapis.com/maps/api/js';
 export const BAIDU_MAP_JAVASCRIPT_URL: string = 'https://api.map.baidu.com/api?v=3.0';
--- a/src/core/file.ts
+++ b/src/core/file.ts
@@ -6,6 +6,7 @@ export class KnownFileType {
    public static readonly TSV = new KnownFileType('tsv', 'text/tab-separated-values');
    public static readonly MARKDOWN = new KnownFileType('md', 'text/markdown');
    public static readonly JS = new KnownFileType('js', 'application/javascript');
+    public static readonly JPG = new KnownFileType('jpg', 'image/jpeg');

    public readonly extension: string;
    public readonly contentType: string;
@@ -37,6 +38,12 @@ export class KnownFileType {
        });
    }

+    public createFileFromBlob(blob: Blob, fileName: string): File {
+        return new File([blob], this.formatFileName(fileName), {
+            type: this.contentType,
+        });
+    }
+
    public static parse(extension: string): KnownFileType | undefined {
        return KnownFileType.allInstancesByExtension[extension];
    }
--- a/src/lib/server_settings.ts
+++ b/src/lib/server_settings.ts
@@ -35,6 +35,10 @@ export function isMCPServerEnabled(): boolean {
    return getServerSetting('mcp') === 1;
 }

+export function isTransactionFromAIImageRecognitionEnabled(): boolean {
+    return getServerSetting('llmt') === 1;
+}
+
 export function getLoginPageTips(): Record<string, string>{
    return getServerSetting('lpt') as Record<string, string>;
 }
--- a/src/lib/services.ts
+++ b/src/lib/services.ts
@@ -21,6 +21,7 @@ import {
    DEFAULT_UPLOAD_API_TIMEOUT,
    DEFAULT_EXPORT_API_TIMEOUT,
    DEFAULT_IMPORT_API_TIMEOUT,
+    DEFAULT_LLM_API_TIMEOUT,
    GOOGLE_MAP_JAVASCRIPT_URL,
    BAIDU_MAP_JAVASCRIPT_URL,
    AMAP_JAVASCRIPT_URL
@@ -134,6 +135,9 @@ import type {
 import type {
    UserApplicationCloudSettingsUpdateRequest
 } from '@/models/user_app_cloud_setting.ts';
+import type {
+    RecognizedReceiptImageResponse
+} from '@/models/large_language_model.ts';

 import {
    getCurrentToken,
@@ -635,6 +639,13 @@ export default {
    deleteTransactionTemplate: (req: TransactionTemplateDeleteRequest): ApiResponsePromise<boolean> => {
        return axios.post<ApiResponse<boolean>>('v1/transaction/templates/delete.json', req);
    },
+    recognizeReceiptImage: ({ imageFile }: { imageFile: File }): ApiResponsePromise<RecognizedReceiptImageResponse> => {
+        return axios.postForm<ApiResponse<RecognizedReceiptImageResponse>>('v1/llm/transactions/recognize_receipt_image.json', {
+            image: imageFile
+        }, {
+            timeout: DEFAULT_LLM_API_TIMEOUT
+        });
+    },
    getLatestExchangeRates: (param: { ignoreError?: boolean }): ApiResponsePromise<LatestExchangeRateResponse> => {
        return axios.get<ApiResponse<LatestExchangeRateResponse>>('v1/exchange_rates/latest.json', {
            ignoreError: !!param.ignoreError,
--- a/src/lib/ui/common.ts
+++ b/src/lib/ui/common.ts
@@ -3,6 +3,7 @@ import Clipboard from 'clipboard';
 import { ThemeType } from '@/core/theme.ts';

 import { type AmountColor, PresetAmountColor } from '@/core/color.ts';
+import { KnownFileType } from '@/core/file.ts';

 import logger from '../logger.ts';

@@ -134,6 +135,64 @@ export function startDownloadFile(fileName: string, fileData: Blob): void {
    dataLink.click();
 }

+export function compressJpgImage(file: File, maxWidth: number, maxHeight: number, quality: number): Promise<Blob> {
+    return new Promise((resolve, reject) => {
+        const reader = new FileReader();
+
+        reader.onload = (event) => {
+            const img = new Image();
+
+            img.onload = () => {
+                let width = img.width;
+                let height = img.height;
+
+                if (width > maxWidth || height > maxHeight) {
+                    const scale = Math.min(maxWidth / width, maxHeight / height);
+                    width = Math.floor(width * scale);
+                    height = Math.floor(height * scale);
+                }
+
+                const canvas = document.createElement('canvas');
+                const ctx = canvas.getContext('2d');
+
+                if (!ctx) {
+                    reject(new Error('failed to get canvas context'));
+                    return;
+                }
+
+                canvas.width = width;
+                canvas.height = height;
+
+                ctx.drawImage(img, 0, 0, width, height);
+
+                canvas.toBlob((blob) => {
+                    if (blob) {
+                        resolve(blob);
+                    } else {
+                        reject(new Error('failed to compress image'));
+                    }
+                }, KnownFileType.JPG.contentType, quality);
+            };
+
+            img.onerror = (error) => {
+                reject(error);
+            };
+
+            if (event.target && event.target.result) {
+                img.src = event.target.result as string;
+            } else {
+                reject(new Error('failed to read file'));
+            }
+        };
+
+        reader.onerror = (error) => {
+            reject(error);
+        };
+
+        reader.readAsDataURL(file);
+    });
+}
+
 export function clearBrowserCaches(): Promise<void> {
    if (!window.caches) {
        logger.error('caches API is not supported in this browser');
--- a/src/locales/de.json
+++ b/src/locales/de.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Cannot update exchange rate data for base currency",
        "cannot delete exchange rate data for base currency": "Cannot delete exchange rate data for base currency",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "Abfrageelemente dürfen nicht leer sein",
        "query items too much": "Zu viele Abfrageelemente",
        "query items have invalid item": "Ungültiges Element in Abfrageelementen",
@@ -1389,6 +1393,7 @@
    "Refresh": "Aktualisieren",
    "Clear": "Löschen",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "Keine",
    "Unspecified": "Nicht angegeben",
    "Not set": "Nicht festgelegt",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Duplicate (With Time)",
    "Duplicate (With Geographic Location)": "Duplicate (With Geographic Location)",
    "Duplicate (With Time and Geographic Location)": "Duplicate (With Time and Geographic Location)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Kategorie",
    "Secondary Category": "Secondary Category",
    "Expense Category": "Expense Category",
--- a/src/locales/en.json
+++ b/src/locales/en.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Cannot update exchange rate data for base currency",
        "cannot delete exchange rate data for base currency": "Cannot delete exchange rate data for base currency",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "There are no query items",
        "query items too much": "There are too many query items",
        "query items have invalid item": "There is invalid item in query items",
@@ -1389,6 +1393,7 @@
    "Refresh": "Refresh",
    "Clear": "Clear",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "None",
    "Unspecified": "Unspecified",
    "Not set": "Not set",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Duplicate (With Time)",
    "Duplicate (With Geographic Location)": "Duplicate (With Geographic Location)",
    "Duplicate (With Time and Geographic Location)": "Duplicate (With Time and Geographic Location)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Category",
    "Secondary Category": "Secondary Category",
    "Expense Category": "Expense Category",
--- a/src/locales/es.json
+++ b/src/locales/es.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Cannot update exchange rate data for base currency",
        "cannot delete exchange rate data for base currency": "Cannot delete exchange rate data for base currency",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "--",
        "query items too much": "--",
        "query items have invalid item": "Hay un elemento no válido en los elementos de consulta",
@@ -1389,6 +1393,7 @@
    "Refresh": "Refrescar",
    "Clear": "Claro",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "Ninguno",
    "Unspecified": "No especificado",
    "Not set": "No establecido",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Duplicate (With Time)",
    "Duplicate (With Geographic Location)": "Duplicate (With Geographic Location)",
    "Duplicate (With Time and Geographic Location)": "Duplicate (With Time and Geographic Location)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Categoría",
    "Secondary Category": "Secondary Category",
    "Expense Category": "Expense Category",
--- a/src/locales/it.json
+++ b/src/locales/it.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Cannot update exchange rate data for base currency",
        "cannot delete exchange rate data for base currency": "Cannot delete exchange rate data for base currency",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "Non ci sono elementi di query",
        "query items too much": "Ci sono troppi elementi di query",
        "query items have invalid item": "C'è un elemento non valido negli elementi di query",
@@ -1389,6 +1393,7 @@
    "Refresh": "Aggiorna",
    "Clear": "Pulisci",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "Nessuno",
    "Unspecified": "Non specificato",
    "Not set": "Non impostato",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Duplica (con ora)",
    "Duplicate (With Geographic Location)": "Duplica (con posizione geografica)",
    "Duplicate (With Time and Geographic Location)": "Duplica (con ora e posizione geografica)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Categoria",
    "Secondary Category": "Categoria secondaria",
    "Expense Category": "Expense Category",
--- a/src/locales/ja.json
+++ b/src/locales/ja.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Cannot update exchange rate data for base currency",
        "cannot delete exchange rate data for base currency": "Cannot delete exchange rate data for base currency",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "クエリ項目がありません",
        "query items too much": "クエリ項目が多すぎます",
        "query items have invalid item": "クエリ項目に無効な項目があります",
@@ -1389,6 +1393,7 @@
    "Refresh": "リフレッシュ",
    "Clear": "消去",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "なし",
    "Unspecified": "不特定",
    "Not set": "セットしていない",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "複製（時間含む）",
    "Duplicate (With Geographic Location)": "複製（地理座標を含む）",
    "Duplicate (With Time and Geographic Location)": "複製（時間と地理座標を含む）",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "カテゴリ",
    "Secondary Category": "二次カテゴリ",
    "Expense Category": "Expense Category",
--- a/src/locales/nl.json
+++ b/src/locales/nl.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Wisselkoersgegevens voor basisvaluta kunnen niet worden bijgewerkt",
        "cannot delete exchange rate data for base currency": "Wisselkoersgegevens voor basisvaluta kunnen niet worden verwijderd",
        "mcp server is not enabled": "MCP-server is niet ingeschakeld",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "Geen zoekitems opgegeven",
        "query items too much": "Te veel zoekitems",
        "query items have invalid item": "Ongeldig item in zoekitems",
@@ -1389,6 +1393,7 @@
    "Refresh": "Vernieuwen",
    "Clear": "Wissen",
    "Generate": "Genereren",
+    "Recognize": "Recognize",
    "None": "Geen",
    "Unspecified": "Niet gespecificeerd",
    "Not set": "Niet ingesteld",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Dupliceren (met tijd)",
    "Duplicate (With Geographic Location)": "Dupliceren (met geografische locatie)",
    "Duplicate (With Time and Geographic Location)": "Dupliceren (met tijd en locatie)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Categorie",
    "Secondary Category": "Secundaire categorie",
    "Expense Category": "Uitgavecategorie",
--- a/src/locales/pt_BR.json
+++ b/src/locales/pt_BR.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Não é possível atualizar dados de taxa de câmbio para a moeda base",
        "cannot delete exchange rate data for base currency": "Não é possível excluir dados de taxa de câmbio para a moeda base",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "Não há itens de consulta",
        "query items too much": "Há muitos itens de consulta",
        "query items have invalid item": "Há item inválido nos itens de consulta",
@@ -1389,6 +1393,7 @@
    "Refresh": "Atualizar",
    "Clear": "Limpar",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "Nenhum",
    "Unspecified": "Não especificado",
    "Not set": "Não definido",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Duplicar (Com Tempo)",
    "Duplicate (With Geographic Location)": "Duplicar (Com Localização Geográfica)",
    "Duplicate (With Time and Geographic Location)": "Duplicar (Com Tempo e Localização Geográfica)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Categoria",
    "Secondary Category": "Categoria Secundária",
    "Expense Category": "Expense Category",
--- a/src/locales/ru.json
+++ b/src/locales/ru.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Cannot update exchange rate data for base currency",
        "cannot delete exchange rate data for base currency": "Cannot delete exchange rate data for base currency",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "Нет элементов запроса",
        "query items too much": "Слишком много элементов запроса",
        "query items have invalid item": "В элементах запроса присутствует недопустимый элемент",
@@ -1389,6 +1393,7 @@
    "Refresh": "Обновить",
    "Clear": "Очистить",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "Нет",
    "Unspecified": "Не указано",
    "Not set": "Не установлено",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Duplicate (With Time)",
    "Duplicate (With Geographic Location)": "Duplicate (With Geographic Location)",
    "Duplicate (With Time and Geographic Location)": "Duplicate (With Time and Geographic Location)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Категория",
    "Secondary Category": "Secondary Category",
    "Expense Category": "Expense Category",
--- a/src/locales/uk.json
+++ b/src/locales/uk.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Cannot update exchange rate data for base currency",
        "cannot delete exchange rate data for base currency": "Cannot delete exchange rate data for base currency",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "Елементи запиту не можуть бути порожніми",
        "query items too much": "Занадто багато елементів запиту",
        "query items have invalid item": "Запит містить недійсний елемент",
@@ -1389,6 +1393,7 @@
    "Refresh": "Оновити",
    "Clear": "Очистити",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "Немає",
    "Unspecified": "Не вказано",
    "Not set": "Не встановлено",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Дублювати (з часом)",
    "Duplicate (With Geographic Location)": "Дублювати (з геолокацією)",
    "Duplicate (With Time and Geographic Location)": "Дублювати (з часом і геолокацією)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Категорія",
    "Secondary Category": "Вторинна категорія",
    "Expense Category": "Expense Category",
--- a/src/locales/vi.json
+++ b/src/locales/vi.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "Cannot update exchange rate data for base currency",
        "cannot delete exchange rate data for base currency": "Cannot delete exchange rate data for base currency",
        "mcp server is not enabled": "MCP Server is not enabled",
+        "llm provider is not enabled": "Large Language Model provider is not enabled",
+        "no image for AI recognition": "There is no image for AI recognition",
+        "image for AI recognition is empty": "Image for AI recognition file is empty",
+        "exceed the maximum size of image file for AI recognition": "The uploaded image for AI recognition exceeds the maximum allowed file size",
        "query items cannot be blank": "Không có mục truy vấn",
        "query items too much": "Có quá nhiều mục truy vấn",
        "query items have invalid item": "Có mục không hợp lệ trong các mục truy vấn",
@@ -1389,6 +1393,7 @@
    "Refresh": "Làm mới",
    "Clear": "Xóa",
    "Generate": "Generate",
+    "Recognize": "Recognize",
    "None": "Không có",
    "Unspecified": "Không xác định",
    "Not set": "Not set",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "Duplicate (With Time)",
    "Duplicate (With Geographic Location)": "Duplicate (With Geographic Location)",
    "Duplicate (With Time and Geographic Location)": "Duplicate (With Time and Geographic Location)",
+    "AI Image Recognition": "AI Image Recognition",
+    "Choose from Library": "Choose from Library",
+    "Take Photo": "Take Photo",
+    "Unable to load image": "Unable to load image",
+    "Unable to recognize image": "Unable to recognize image",
+    "Drag and drop a receipt or transaction image here, or click to select one": "Drag and drop a receipt or transaction image here, or click to select one",
+    "Release to load image": "Release to load image",
+    "Please select a receipt or transaction image first": "Please select a receipt or transaction image first",
    "Category": "Danh mục",
    "Secondary Category": "Secondary Category",
    "Expense Category": "Expense Category",
--- a/src/locales/zh_Hans.json
+++ b/src/locales/zh_Hans.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "不能更新默认货币的汇率数据",
        "cannot delete exchange rate data for base currency": "不能删除默认货币的汇率数据",
        "mcp server is not enabled": "MCP 服务器没有启用",
+        "llm provider is not enabled": "大语言模型服务提供者没有启用",
+        "no image for AI recognition": "没有用于AI识别的图片",
+        "image for AI recognition is empty": "用于AI识别的图片为空",
+        "exceed the maximum size of image file for AI recognition": "用于AI识别的图片超出了允许的最大文件大小",
        "query items cannot be blank": "请求项目不能为空",
        "query items too much": "请求项目过多",
        "query items have invalid item": "请求项目中有非法项目",
@@ -1389,6 +1393,7 @@
    "Refresh": "刷新",
    "Clear": "清除",
    "Generate": "生成",
+    "Recognize": "识别",
    "None": "无",
    "Unspecified": "未指定",
    "Not set": "未设置",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "复制 (含时间)",
    "Duplicate (With Geographic Location)": "复制 (含地理位置)",
    "Duplicate (With Time and Geographic Location)": "复制 (含时间和地理位置)",
+    "AI Image Recognition": "AI识图",
+    "Choose from Library": "从图库选择",
+    "Take Photo": "拍照",
+    "Unable to load image": "无法加载图片",
+    "Unable to recognize image": "无法识别图片",
+    "Drag and drop a receipt or transaction image here, or click to select one": "拖拽收据或交易图片到此处，或点击选择图片",
+    "Release to load image": "释放以加载图片",
+    "Please select a receipt or transaction image first": "请先选择收据或交易图片",
    "Category": "分类",
    "Secondary Category": "二级分类",
    "Expense Category": "支出分类",
--- a/src/locales/zh_Hant.json
+++ b/src/locales/zh_Hant.json
@@ -1224,6 +1224,10 @@
        "cannot update exchange rate data for base currency": "不能更新基準貨幣的匯率資料",
        "cannot delete exchange rate data for base currency": "不能刪除基準貨幣的匯率資料",
        "mcp server is not enabled": "MCP 伺服器未啟用",
+        "llm provider is not enabled": "大型語言模型服務提供者未啟用",
+        "no image for AI recognition": "沒有用於AI識別的圖片檔案",
+        "image for AI recognition is empty": "用於AI識別的圖片檔案為空",
+        "exceed the maximum size of image file for AI recognition": "用於AI識別的圖片超出了允許的最大檔案大小",
        "query items cannot be blank": "查詢項目不能為空",
        "query items too much": "查詢項目過多",
        "query items have invalid item": "查詢項目中有非法項目",
@@ -1389,6 +1393,7 @@
    "Refresh": "重新載入",
    "Clear": "清除",
    "Generate": "產生",
+    "Recognize": "識別",
    "None": "無",
    "Unspecified": "未指定",
    "Not set": "未設置",
@@ -1719,6 +1724,14 @@
    "Duplicate (With Time)": "複製 (含時間)",
    "Duplicate (With Geographic Location)": "複製 (含地理位置)",
    "Duplicate (With Time and Geographic Location)": "複製 (含時間和地理位置)",
+    "AI Image Recognition": "AI識圖",
+    "Choose from Library": "從相簿選擇",
+    "Take Photo": "拍照",
+    "Unable to load image": "無法載入圖片",
+    "Unable to recognize image": "無法識別圖片",
+    "Drag and drop a receipt or transaction image here, or click to select one": "將收據或交易圖片拖放到此處，或點擊以選擇圖片",
+    "Release to load image": "放開以載入圖片",
+    "Please select a receipt or transaction image first": "請先選擇收據或交易圖片",
    "Category": "分類",
    "Secondary Category": "次分類",
    "Expense Category": "支出分類",
--- a/src/mobile-main.ts
+++ b/src/mobile-main.ts
@@ -79,6 +79,7 @@ import MapSheet from '@/components/mobile/MapSheet.vue';
 import TransactionTagSelectionSheet from '@/components/mobile/TransactionTagSelectionSheet.vue';
 import ScheduleFrequencySheet from '@/components/mobile/ScheduleFrequencySheet.vue';
 import AccountBalanceTrendsBarChart from '@/components/mobile/AccountBalanceTrendsBarChart.vue';
+import AIImageRecognitionSheet from '@/components/mobile/AIImageRecognitionSheet.vue';

 import TextareaAutoSize from '@/directives/mobile/textareaAutoSize.ts';

@@ -170,8 +171,9 @@ app.component('InformationSheet', InformationSheet);
 app.component('NumberPadSheet', NumberPadSheet);
 app.component('MapSheet', MapSheet);
 app.component('TransactionTagSelectionSheet', TransactionTagSelectionSheet);
-app.component('AccountBalanceTrendsBarChart', AccountBalanceTrendsBarChart);
 app.component('ScheduleFrequencySheet', ScheduleFrequencySheet);
+app.component('AccountBalanceTrendsBarChart', AccountBalanceTrendsBarChart);
+app.component('AIImageRecognitionSheet', AIImageRecognitionSheet);

 app.directive('TextareaAutoSize', TextareaAutoSize);

--- a/src/models/large_language_model.ts
+++ b/src/models/large_language_model.ts
@@ -0,0 +1,11 @@
+export interface RecognizedReceiptImageResponse {
+    readonly type: number;
+    readonly time?: number;
+    readonly categoryId?: string;
+    readonly sourceAccountId?: string;
+    readonly destinationAccountId?: string;
+    readonly sourceAmount?: number;
+    readonly destinationAmount?: number;
+    readonly tagIds?: string[];
+    readonly comment?: string;
+}
--- a/src/stores/transaction.ts
+++ b/src/stores/transaction.ts
@@ -33,6 +33,9 @@ import {
 import {
    type ExportTransactionDataRequest
 } from '@/models/data_management.ts';
+import type {
+    RecognizedReceiptImageResponse
+} from '@/models/large_language_model.ts';

 import {
    getUserTransactionDraft,
@@ -1157,6 +1160,31 @@ export const useTransactionsStore = defineStore('transactions', () => {
        });
    }

+    function recognizeReceiptImage({ imageFile }: { imageFile: File }): Promise<RecognizedReceiptImageResponse> {
+        return new Promise((resolve, reject) => {
+            services.recognizeReceiptImage({ imageFile }).then(response => {
+                const data = response.data;
+
+                if (!data || !data.success || !data.result) {
+                    reject({ message: 'Unable to recognize image' });
+                    return;
+                }
+
+                resolve(data.result);
+            }).catch(error => {
+                logger.error('failed to recognize image', error);
+
+                if (error.response && error.response.data && error.response.data.errorMessage) {
+                    reject({ error: error.response.data });
+                } else if (!error.processed) {
+                    reject({ message: 'Unable to recognize image' });
+                } else {
+                    reject(error);
+                }
+            });
+        });
+    }
+
    function parseImportDsvFile({ fileType, fileEncoding, importFile }: { fileType: string, fileEncoding?: string, importFile: File }): Promise<string[][]> {
        return new Promise((resolve, reject) => {
            services.parseImportDsvFile({ fileType, fileEncoding, importFile }).then(response => {
@@ -1370,6 +1398,7 @@ export const useTransactionsStore = defineStore('transactions', () => {
        getTransaction,
        saveTransaction,
        deleteTransaction,
+        recognizeReceiptImage,
        parseImportDsvFile,
        parseImportTransaction,
        importTransactions,
--- a/src/views/desktop/transactions/ListPage.vue
+++ b/src/views/desktop/transactions/ListPage.vue
@@ -63,11 +63,16 @@
                                            <v-btn class="ms-3" color="default" variant="outlined"
                                                   :disabled="loading || !canAddTransaction" @click="add()">
                                                {{ tt('Add') }}
-                                                <v-menu activator="parent" :open-on-hover="true" v-if="allTransactionTemplates && allTransactionTemplates.length">
+                                                <v-menu activator="parent" :open-on-hover="true" v-if="isTransactionFromAIImageRecognitionEnabled() || (allTransactionTemplates && allTransactionTemplates.length)">
                                                    <v-list>
-                                                        <v-list-item :title="template.name"
+                                                        <v-list-item key="AIImageRecognition"
+                                                                     :title="tt('AI Image Recognition')"
+                                                                     :prepend-icon="mdiMagicStaff"
+                                                                     v-if="isTransactionFromAIImageRecognitionEnabled()"
+                                                                     @click="addByRecognizingImage"></v-list-item>
+                                                        <v-list-item :key="template.id"
+                                                                     :title="template.name"
                                                                     :prepend-icon="mdiTextBoxOutline"
-                                                                     :key="template.id"
                                                                     v-for="template in allTransactionTemplates"
                                                                     @click="add(template)"></v-list-item>
                                                    </v-list>
@@ -620,6 +625,7 @@
                            @error="onShowDateRangeError" />

    <edit-dialog ref="editDialog" :type="TransactionEditPageType.Transaction" />
+    <a-i-image-recognition-dialog ref="aiImageRecognitionDialog" />
    <import-dialog ref="importDialog" :persistent="true" />

    <v-dialog width="800" v-model="showFilterAccountDialog">
@@ -647,6 +653,7 @@ import PaginationButtons from '@/components/desktop/PaginationButtons.vue';
 import ConfirmDialog from '@/components/desktop/ConfirmDialog.vue';
 import SnackBar from '@/components/desktop/SnackBar.vue';
 import EditDialog from './list/dialogs/EditDialog.vue';
+import AIImageRecognitionDialog from './list/dialogs/AIImageRecognitionDialog.vue';
 import ImportDialog from './import/ImportDialog.vue';
 import AccountFilterSettingsCard from '@/views/desktop/common/cards/AccountFilterSettingsCard.vue';
 import CategoryFilterSettingsCard from '@/views/desktop/common/cards/CategoryFilterSettingsCard.vue';
@@ -716,7 +723,7 @@ import {
    categoryTypeToTransactionType,
    transactionTypeToCategoryType
 } from '@/lib/category.ts';
-import { isDataExportingEnabled, isDataImportingEnabled } from '@/lib/server_settings.ts';
+import { isDataExportingEnabled, isDataImportingEnabled, isTransactionFromAIImageRecognitionEnabled } from '@/lib/server_settings.ts';
 import { startDownloadFile } from '@/lib/ui/common.ts';
 import { scrollToSelectedItem } from '@/lib/ui/desktop.ts';
 import logger from '@/lib/logger.ts';
@@ -738,6 +745,7 @@ import {
    mdiMinusBoxMultipleOutline,
    mdiCloseBoxMultipleOutline,
    mdiPound,
+    mdiMagicStaff,
    mdiTextBoxOutline
 } from '@mdi/js';

@@ -760,6 +768,7 @@ const props = defineProps<TransactionListProps>();
 type ConfirmDialogType = InstanceType<typeof ConfirmDialog>;
 type SnackBarType = InstanceType<typeof SnackBar>;
 type EditDialogType = InstanceType<typeof EditDialog>;
+type AIImageRecognitionDialogType = InstanceType<typeof AIImageRecognitionDialog>;
 type ImportDialogType = InstanceType<typeof ImportDialog>;

 interface TransactionTemplateWithIcon {
@@ -859,6 +868,7 @@ const tagFilterMenu = useTemplateRef<VMenu>('tagFilterMenu');
 const confirmDialog = useTemplateRef<ConfirmDialogType>('confirmDialog');
 const snackbar = useTemplateRef<SnackBarType>('snackbar');
 const editDialog = useTemplateRef<EditDialogType>('editDialog');
+const aiImageRecognitionDialog = useTemplateRef<AIImageRecognitionDialogType>('aiImageRecognitionDialog');
 const importDialog = useTemplateRef<ImportDialogType>('importDialog');

 const activeTab = ref<string>('transactionPage');
@@ -1597,6 +1607,33 @@ function add(template?: TransactionTemplate): void {
    });
 }

+function addByRecognizingImage(): void {
+    aiImageRecognitionDialog.value?.open().then(result => {
+        editDialog.value?.open({
+            time: result.time,
+            type: result.type,
+            categoryId: result.categoryId,
+            accountId: result.sourceAccountId,
+            destinationAccountId: result.destinationAccountId,
+            amount: result.sourceAmount,
+            destinationAmount: result.destinationAmount,
+            tagIds: result.tagIds ? result.tagIds.join(',') : undefined,
+            comment: result.comment,
+            noTransactionDraft: true
+        }).then(result => {
+            if (result && result.message) {
+                snackbar.value?.showMessage(result.message);
+            }
+
+            reload(false, false);
+        }).catch(error => {
+            if (error) {
+                snackbar.value?.showError(error);
+            }
+        });
+    });
+}
+
 function importTransaction(): void {
    importDialog.value?.open().then(() => {
        reload(false, false);
--- a/src/views/desktop/transactions/list/dialogs/AIImageRecognitionDialog.vue
+++ b/src/views/desktop/transactions/list/dialogs/AIImageRecognitionDialog.vue
@@ -0,0 +1,208 @@
+<template>
+    <v-dialog width="800" :persistent="loading || recognizing || !!imageFile" v-model="showState">
+        <v-card class="pa-2 pa-sm-4 pa-md-4">
+            <template #title>
+                <div class="d-flex align-center justify-center">
+                    <h4 class="text-h4">{{ tt('AI Image Recognition') }}</h4>
+                </div>
+            </template>
+
+            <v-card-text class="d-flex justify-center w-100 my-md-4 pt-0">
+                <div class="w-100 border position-relative"
+                     @dragenter.prevent="onDragEnter"
+                     @dragover.prevent
+                     @dragleave.prevent="onDragLeave"
+                     @drop.prevent="onDrop">
+                    <div class="d-flex w-100 fill-height justify-center align-center justify-content-center"
+                         :class="{ 'dropzone': true, 'dropzone-dragover': isDragOver }" style="height: 480px">
+                        <h3 v-if="!imageFile && !isDragOver">{{ tt('Drag and drop a receipt or transaction image here, or click to select one') }}</h3>
+                        <h3 v-if="isDragOver">{{ tt('Release to load image') }}</h3>
+                    </div>
+                    <v-img height="480px" :class="{ 'cursor-pointer': !loading || !recognizing || !isDragOver }"
+                           :src="imageSrc" @click="showOpenImageDialog">
+                        <template #placeholder>
+                            <div class="w-100 fill-height bg-grey-200"></div>
+                        </template>
+                    </v-img>
+                </div>
+            </v-card-text>
+
+            <v-card-text class="overflow-y-visible">
+                <div ref="buttonContainer" class="w-100 d-flex justify-center gap-4">
+                    <v-btn :disabled="loading || recognizing || !imageFile" @click="recognize">
+                        {{ tt('Recognize') }}
+                        <v-progress-circular indeterminate size="22" class="ms-2" v-if="recognizing"></v-progress-circular>
+                    </v-btn>
+                    <v-btn color="secondary" variant="tonal" :disabled="loading || recognizing"
+                           @click="cancel">{{ tt('Cancel') }}</v-btn>
+                </div>
+            </v-card-text>
+        </v-card>
+    </v-dialog>
+
+    <snack-bar ref="snackbar" />
+    <input ref="imageInput" type="file" style="display: none" :accept="SUPPORTED_IMAGE_EXTENSIONS" @change="openImage($event)" />
+</template>
+
+<script setup lang="ts">
+import SnackBar from '@/components/desktop/SnackBar.vue';
+
+import { ref, useTemplateRef } from 'vue';
+
+import { useI18n } from '@/locales/helpers.ts';
+
+import { useTransactionsStore } from '@/stores/transaction.ts';
+
+import { KnownFileType } from '@/core/file.ts';
+import { SUPPORTED_IMAGE_EXTENSIONS } from '@/consts/file.ts';
+
+import type { RecognizedReceiptImageResponse } from '@/models/large_language_model.ts';
+
+import { compressJpgImage } from '@/lib/ui/common.ts';
+import logger from '@/lib/logger.ts';
+
+type SnackBarType = InstanceType<typeof SnackBar>;
+
+const { tt } = useI18n();
+
+const transactionsStore = useTransactionsStore();
+
+const snackbar = useTemplateRef<SnackBarType>('snackbar');
+const imageInput = useTemplateRef<HTMLInputElement>('imageInput');
+
+let resolveFunc: ((response: RecognizedReceiptImageResponse) => void) | null = null;
+let rejectFunc: ((reason?: unknown) => void) | null = null;
+
+const showState = ref<boolean>(false);
+const loading = ref<boolean>(false);
+const recognizing = ref<boolean>(false);
+const imageFile = ref<File | null>(null);
+const imageSrc = ref<string | undefined>(undefined);
+const isDragOver = ref<boolean>(false);
+
+function loadImage(file: File): void {
+    compressJpgImage(file, 1280, 1280, 0.8).then(blob => {
+        imageFile.value = KnownFileType.JPG.createFileFromBlob(blob, "image");
+        imageSrc.value = URL.createObjectURL(blob);
+    }).catch(error => {
+        imageFile.value = null;
+        imageSrc.value = undefined;
+        logger.error('failed to compress image', error);
+        snackbar.value?.showError('Unable to load image');
+    });
+}
+
+function open(): Promise<RecognizedReceiptImageResponse> {
+    showState.value = true;
+    loading.value = false;
+    recognizing.value = false;
+    imageFile.value = null;
+    imageSrc.value = undefined;
+
+    return new Promise((resolve, reject) => {
+        resolveFunc = resolve;
+        rejectFunc = reject;
+    });
+}
+
+function showOpenImageDialog(): void {
+    if (loading.value || recognizing.value || isDragOver.value) {
+        return;
+    }
+
+    imageInput.value?.click();
+}
+
+function openImage(event: Event): void {
+    if (!event || !event.target) {
+        return;
+    }
+
+    const el = event.target as HTMLInputElement;
+
+    if (!el.files || !el.files.length || !el.files[0]) {
+        return;
+    }
+
+    const image = el.files[0] as File;
+
+    el.value = '';
+
+    loadImage(image);
+}
+
+function recognize(): void {
+    if (loading.value || recognizing.value || !imageFile.value) {
+        return;
+    }
+
+    recognizing.value = true;
+
+    transactionsStore.recognizeReceiptImage({
+        imageFile: imageFile.value
+    }).then(response => {
+        resolveFunc?.(response);
+        showState.value = false;
+        recognizing.value = false;
+    }).catch(error => {
+        recognizing.value = false;
+
+        if (!error.processed) {
+            snackbar.value?.showError(error);
+        }
+    });
+}
+
+function cancel(): void {
+    rejectFunc?.();
+    showState.value = false;
+    loading.value = false;
+    recognizing.value = false;
+    imageFile.value = null;
+    imageSrc.value = undefined;
+}
+
+function onDragEnter(): void {
+    if (loading.value || recognizing.value) {
+        return;
+    }
+
+    isDragOver.value = true;
+}
+
+function onDragLeave(): void {
+    isDragOver.value = false;
+}
+
+function onDrop(event: DragEvent): void {
+    if (loading.value || recognizing.value) {
+        return;
+    }
+
+    isDragOver.value = false;
+
+    if (event.dataTransfer && event.dataTransfer.files && event.dataTransfer.files.length && event.dataTransfer.files[0]) {
+        loadImage(event.dataTransfer.files[0] as File);
+    }
+}
+
+defineExpose({
+    open
+});
+</script>
+
+<style>
+.dropzone {
+    position: absolute;
+    top: 0;
+    left: 0;
+    width: 100%;
+    pointer-events: none;
+    border-radius: 8px;
+    z-index: 10;
+}
+
+.dropzone-dragover {
+    border: 6px dashed rgba(var(--v-border-color),var(--v-border-opacity));
+}
+</style>
--- a/src/views/mobile/HomePage.vue
+++ b/src/views/mobile/HomePage.vue
@@ -188,7 +188,14 @@
        <f7-popover class="template-popover-menu" target-el="#homepage-add-button"
                    v-model:opened="showTransactionTemplatePopover">
            <f7-list dividers v-if="allTransactionTemplates">
-                <f7-list-item :title="template.name" :key="template.id"
+                <f7-list-item key="AIImageRecognition" :title="tt('AI Image Recognition')"
+                              @click="showAIReceiptImageRecognitionSheet = true; showTransactionTemplatePopover = false"
+                              v-if="isTransactionFromAIImageRecognitionEnabled()">
+                    <template #media>
+                        <f7-icon f7="wand_stars"></f7-icon>
+                    </template>
+                </f7-list-item>
+                <f7-list-item :key="template.id" :title="template.name"
                              :link="'/transaction/add?templateId=' + template.id"
                              v-for="template in allTransactionTemplates">
                    <template #media>
@@ -197,11 +204,15 @@
                </f7-list-item>
            </f7-list>
        </f7-popover>
+
+        <a-i-image-recognition-sheet v-model:show="showAIReceiptImageRecognitionSheet"
+                                     @recognition:change="onReceiptRecognitionChanged"/>
    </f7-page>
 </template>

 <script setup lang="ts">
 import { ref, computed } from 'vue';
+import type { Router } from 'framework7/types';

 import { useI18n } from '@/locales/helpers.ts';
 import { useI18nUIComponents } from '@/lib/ui/mobile.ts';
@@ -215,8 +226,14 @@ import { useOverviewStore } from '@/stores/overview.ts';
 import { DateRange } from '@/core/datetime.ts';
 import { TemplateType } from '@/core/template.ts';
 import { TransactionTemplate } from '@/models/transaction_template.ts';
+import type { RecognizedReceiptImageResponse } from '@/models/large_language_model.ts';

 import { isUserLogined, isUserUnlocked } from '@/lib/userstate.ts';
+import { isTransactionFromAIImageRecognitionEnabled } from '@/lib/server_settings.ts';
+
+const props = defineProps<{
+    f7router: Router.Router;
+}>();

 const { tt } = useI18n();
 const { showToast } = useI18nUIComponents();
@@ -236,6 +253,7 @@ const overviewStore = useOverviewStore();

 const loading = ref<boolean>(true);
 const showTransactionTemplatePopover = ref<boolean>(false);
+const showAIReceiptImageRecognitionSheet = ref<boolean>(false);

 const allTransactionTemplates = computed<TransactionTemplate[]>(() => {
    const allTemplates = transactionTemplatesStore.allVisibleTemplates;
@@ -243,7 +261,7 @@ const allTransactionTemplates = computed<TransactionTemplate[]>(() => {
 });

 function openTransactionTemplatePopover(): void {
-    if (allTransactionTemplates.value && allTransactionTemplates.value.length) {
+    if (isTransactionFromAIImageRecognitionEnabled() || (allTransactionTemplates.value && allTransactionTemplates.value.length)) {
        showTransactionTemplatePopover.value = true;
    }
 }
@@ -291,6 +309,48 @@ function reload(done?: () => void): void {
    });
 }

+function onReceiptRecognitionChanged(result: RecognizedReceiptImageResponse): void {
+    const params: string[] = [];
+
+    if (result.type) {
+        params.push(`type=${result.type}`);
+    }
+
+    if (result.time) {
+        params.push(`time=${result.time}`);
+    }
+
+    if (result.categoryId) {
+        params.push(`categoryId=${result.categoryId}`);
+    }
+
+    if (result.sourceAccountId) {
+        params.push(`accountId=${result.sourceAccountId}`);
+    }
+
+    if (result.destinationAccountId) {
+        params.push(`destinationAccountId=${result.destinationAccountId}`);
+    }
+
+    if (result.sourceAmount) {
+        params.push(`amount=${result.sourceAmount}`);
+    }
+
+    if (result.destinationAmount) {
+        params.push(`destinationAmount=${result.destinationAmount}`);
+    }
+
+    if (result.tagIds) {
+        params.push(`tagIds=${result.tagIds.join(',')}`);
+    }
+
+    if (result.comment) {
+        params.push(`comment=${encodeURIComponent(result.comment)}`);
+    }
+
+    props.f7router.navigate(`/transaction/add?${params.join('&')}`);
+}
+
 function onPageAfterIn(): void {
    if (!loading.value) {
        reload();
--- a/templates/prompt/receipt_image_recognition.tmpl
+++ b/templates/prompt/receipt_image_recognition.tmpl
@@ -0,0 +1,44 @@
+## Role
+You are a financial assistant.
+Your task is to extract structured transaction data from images provided by the user (such as receipts, transaction records, or vouchers).
+
+## Output
+1. Format: JSON only
+2. No explanations, comments, or extra text outside JSON
+
+## JSON Schema (with field descriptions)
+```
+{
+  "type": "string (transaction type: expense | income | transfer)",
+  "time": "string (transaction time, format: YYYY-MM-DD HH:mm:ss)",
+  "amount": "string (transaction amount, numeric, up to 2 decimals)",
+  "account": "string (source account name)",
+  "category": "string (transaction category)",
+  "tags": ["string (tag name, max 10 allowed)"],
+  "description": "string (transaction description)",
+  "destination_amount": "string (destination amount, numeric, up to 2 decimals, only for transfer)",
+  "destination_account": "string (destination account name, only for transfer)"
+}
+```
+
+## Important rules
+1. Only include fields you can confidently identify.
+2. If unsure about a value, omit the field (do not guess).
+3. If the image contains multiple items, please combine them into a single transaction.
+4. Always return valid JSON.
+
+## Options
+### Expense categories:
+{{.AllExpenseCategoryNames}}
+
+### Income categories:
+{{.AllIncomeCategoryNames}}
+
+### Transfer categories:
+{{.AllTransferCategoryNames}}
+
+### Account names:
+{{.AllAccountNames}}
+
+### Tags:
+{{.AllTagNames}}