[Refactor]: Core/API Split (#1506)

Refactors api folder to core, creates firm split between backend code and api frontend.

pkg/schema/config.go

@@ -0,0 +1,400 @@
+package schema
+
+import (
+	"encoding/json"
+	"fmt"
+	"os"
+
+	"github.com/go-skynet/LocalAI/pkg/utils"
+	"gopkg.in/yaml.v3"
+)
+
+type Config struct {
+	PredictionOptions `yaml:"parameters"`
+	Name              string `yaml:"name"`
+
+	F16            bool              `yaml:"f16"`
+	Threads        int               `yaml:"threads"`
+	Debug          bool              `yaml:"debug"`
+	Roles          map[string]string `yaml:"roles"`
+	Embeddings     bool              `yaml:"embeddings"`
+	Backend        string            `yaml:"backend"`
+	TemplateConfig TemplateConfig    `yaml:"template"`
+
+	PromptStrings, InputStrings                []string `yaml:"-"`
+	InputToken                                 [][]int  `yaml:"-"`
+	functionCallString, functionCallNameString string   `yaml:"-"`
+
+	FunctionsConfig Functions `yaml:"function"`
+
+	FeatureFlag FeatureFlag `yaml:"feature_flags"` // Feature Flag registry. We move fast, and features may break on a per model/backend basis. Registry for (usually temporary) flags that indicate aborting something early.
+	// LLM configs (GPT4ALL, Llama.cpp, ...)
+	LLMConfig `yaml:",inline"`
+
+	// AutoGPTQ specifics
+	AutoGPTQ AutoGPTQ `yaml:"autogptq"`
+
+	// Diffusers
+	Diffusers Diffusers `yaml:"diffusers"`
+	Step      int       `yaml:"step"`
+
+	// GRPC Options
+	GRPC GRPC `yaml:"grpc"`
+
+	// Vall-e-x
+	VallE VallE `yaml:"vall-e"`
+
+	// CUDA
+	// Explicitly enable CUDA or not (some backends might need it)
+	CUDA bool `yaml:"cuda"`
+
+	DownloadFiles []File `yaml:"download_files"`
+}
+
+type File struct {
+	Filename string `yaml:"filename" json:"filename"`
+	SHA256   string `yaml:"sha256" json:"sha256"`
+	URI      string `yaml:"uri" json:"uri"`
+}
+
+type VallE struct {
+	AudioPath string `yaml:"audio_path"`
+}
+
+type FeatureFlag map[string]*bool
+
+func (ff FeatureFlag) Enabled(s string) bool {
+	v, exist := ff[s]
+	return exist && v != nil && *v
+}
+
+type GRPC struct {
+	Attempts          int `yaml:"attempts"`
+	AttemptsSleepTime int `yaml:"attempts_sleep_time"`
+}
+
+type Diffusers struct {
+	CUDA             bool    `yaml:"cuda"`
+	PipelineType     string  `yaml:"pipeline_type"`
+	SchedulerType    string  `yaml:"scheduler_type"`
+	EnableParameters string  `yaml:"enable_parameters"` // A list of comma separated parameters to specify
+	CFGScale         float32 `yaml:"cfg_scale"`         // Classifier-Free Guidance Scale
+	IMG2IMG          bool    `yaml:"img2img"`           // Image to Image Diffuser
+	ClipSkip         int     `yaml:"clip_skip"`         // Skip every N frames
+	ClipModel        string  `yaml:"clip_model"`        // Clip model to use
+	ClipSubFolder    string  `yaml:"clip_subfolder"`    // Subfolder to use for clip model
+	ControlNet       string  `yaml:"control_net"`
+}
+
+type LLMConfig struct {
+	SystemPrompt    string   `yaml:"system_prompt"`
+	TensorSplit     string   `yaml:"tensor_split"`
+	MainGPU         string   `yaml:"main_gpu"`
+	RMSNormEps      float32  `yaml:"rms_norm_eps"`
+	NGQA            int32    `yaml:"ngqa"`
+	PromptCachePath string   `yaml:"prompt_cache_path"`
+	PromptCacheAll  bool     `yaml:"prompt_cache_all"`
+	PromptCacheRO   bool     `yaml:"prompt_cache_ro"`
+	MirostatETA     float64  `yaml:"mirostat_eta"`
+	MirostatTAU     float64  `yaml:"mirostat_tau"`
+	Mirostat        int      `yaml:"mirostat"`
+	NGPULayers      int      `yaml:"gpu_layers"`
+	MMap            bool     `yaml:"mmap"`
+	MMlock          bool     `yaml:"mmlock"`
+	LowVRAM         bool     `yaml:"low_vram"`
+	Grammar         string   `yaml:"grammar"`
+	StopWords       []string `yaml:"stopwords"`
+	Cutstrings      []string `yaml:"cutstrings"`
+	TrimSpace       []string `yaml:"trimspace"`
+	TrimSuffix      []string `yaml:"trimsuffix"`
+
+	ContextSize  int     `yaml:"context_size"`
+	NUMA         bool    `yaml:"numa"`
+	LoraAdapter  string  `yaml:"lora_adapter"`
+	LoraBase     string  `yaml:"lora_base"`
+	LoraScale    float32 `yaml:"lora_scale"`
+	NoMulMatQ    bool    `yaml:"no_mulmatq"`
+	DraftModel   string  `yaml:"draft_model"`
+	NDraft       int32   `yaml:"n_draft"`
+	Quantization string  `yaml:"quantization"`
+	MMProj       string  `yaml:"mmproj"`
+
+	RopeScaling    string  `yaml:"rope_scaling"`
+	YarnExtFactor  float32 `yaml:"yarn_ext_factor"`
+	YarnAttnFactor float32 `yaml:"yarn_attn_factor"`
+	YarnBetaFast   float32 `yaml:"yarn_beta_fast"`
+	YarnBetaSlow   float32 `yaml:"yarn_beta_slow"`
+}
+
+type AutoGPTQ struct {
+	ModelBaseName    string `yaml:"model_base_name"`
+	Device           string `yaml:"device"`
+	Triton           bool   `yaml:"triton"`
+	UseFastTokenizer bool   `yaml:"use_fast_tokenizer"`
+}
+
+type Functions struct {
+	DisableNoAction         bool   `yaml:"disable_no_action"`
+	NoActionFunctionName    string `yaml:"no_action_function_name"`
+	NoActionDescriptionName string `yaml:"no_action_description_name"`
+}
+
+type TemplateConfig struct {
+	Chat        string `yaml:"chat"`
+	ChatMessage string `yaml:"chat_message"`
+	Completion  string `yaml:"completion"`
+	Edit        string `yaml:"edit"`
+	Functions   string `yaml:"function"`
+}
+
+func (c *Config) SetFunctionCallString(s string) {
+	c.functionCallString = s
+}
+
+func (c *Config) SetFunctionCallNameString(s string) {
+	c.functionCallNameString = s
+}
+
+func (c *Config) ShouldUseFunctions() bool {
+	return ((c.functionCallString != "none" || c.functionCallString == "") || c.ShouldCallSpecificFunction())
+}
+
+func (c *Config) ShouldCallSpecificFunction() bool {
+	return len(c.functionCallNameString) > 0
+}
+
+func (c *Config) FunctionToCall() string {
+	return c.functionCallNameString
+}
+
+func defaultPredictOptions(modelFile string) PredictionOptions {
+	return PredictionOptions{
+		TopP:        0.7,
+		TopK:        80,
+		Maxtokens:   512,
+		Temperature: 0.9,
+		Model:       modelFile,
+	}
+}
+
+func DefaultConfig(modelFile string) *Config {
+	return &Config{
+		PredictionOptions: defaultPredictOptions(modelFile),
+	}
+}
+
+func ReadConfigFile(file string) ([]*Config, error) {
+	c := &[]*Config{}
+	f, err := os.ReadFile(file)
+	if err != nil {
+		return nil, fmt.Errorf("cannot read config file: %w", err)
+	}
+	if err := yaml.Unmarshal(f, c); err != nil {
+		return nil, fmt.Errorf("cannot unmarshal config file: %w", err)
+	}
+
+	return *c, nil
+}
+
+func ReadSingleConfigFile(file string) (*Config, error) {
+	c := &Config{}
+	f, err := os.ReadFile(file)
+	if err != nil {
+		return nil, fmt.Errorf("cannot read config file: %w", err)
+	}
+	if err := yaml.Unmarshal(f, c); err != nil {
+		return nil, fmt.Errorf("cannot unmarshal config file: %w", err)
+	}
+
+	return c, nil
+}
+
+func UpdateConfigFromOpenAIRequest(config *Config, input *OpenAIRequest) {
+	if input.Echo {
+		config.Echo = input.Echo
+	}
+	if input.TopK != 0 {
+		config.TopK = input.TopK
+	}
+	if input.TopP != 0 {
+		config.TopP = input.TopP
+	}
+
+	if input.Backend != "" {
+		config.Backend = input.Backend
+	}
+
+	if input.ClipSkip != 0 {
+		config.Diffusers.ClipSkip = input.ClipSkip
+	}
+
+	if input.ModelBaseName != "" {
+		config.AutoGPTQ.ModelBaseName = input.ModelBaseName
+	}
+
+	if input.NegativePromptScale != 0 {
+		config.NegativePromptScale = input.NegativePromptScale
+	}
+
+	if input.UseFastTokenizer {
+		config.UseFastTokenizer = input.UseFastTokenizer
+	}
+
+	if input.NegativePrompt != "" {
+		config.NegativePrompt = input.NegativePrompt
+	}
+
+	if input.RopeFreqBase != 0 {
+		config.RopeFreqBase = input.RopeFreqBase
+	}
+
+	if input.RopeFreqScale != 0 {
+		config.RopeFreqScale = input.RopeFreqScale
+	}
+
+	if input.Grammar != "" {
+		config.Grammar = input.Grammar
+	}
+
+	if input.Temperature != 0 {
+		config.Temperature = input.Temperature
+	}
+
+	if input.Maxtokens != 0 {
+		config.Maxtokens = input.Maxtokens
+	}
+
+	if input.RepeatPenalty != 0 {
+		config.RepeatPenalty = input.RepeatPenalty
+	}
+
+	if input.Keep != 0 {
+		config.Keep = input.Keep
+	}
+
+	if input.Batch != 0 {
+		config.Batch = input.Batch
+	}
+
+	if input.F16 {
+		config.F16 = input.F16
+	}
+
+	if input.IgnoreEOS {
+		config.IgnoreEOS = input.IgnoreEOS
+	}
+
+	if input.Seed != 0 {
+		config.Seed = input.Seed
+	}
+
+	if input.Mirostat != 0 {
+		config.LLMConfig.Mirostat = input.Mirostat
+	}
+
+	if input.MirostatETA != 0 {
+		config.LLMConfig.MirostatETA = input.MirostatETA
+	}
+
+	if input.MirostatTAU != 0 {
+		config.LLMConfig.MirostatTAU = input.MirostatTAU
+	}
+
+	if input.TypicalP != 0 {
+		config.TypicalP = input.TypicalP
+	}
+
+	switch stop := input.Stop.(type) {
+	case string:
+		if stop != "" {
+			config.StopWords = append(config.StopWords, stop)
+		}
+	case []interface{}:
+		for _, pp := range stop {
+			if s, ok := pp.(string); ok {
+				config.StopWords = append(config.StopWords, s)
+			}
+		}
+	}
+
+	// Decode each request's message content
+	index := 0
+	for i, m := range input.Messages {
+		switch content := m.Content.(type) {
+		case string:
+			input.Messages[i].StringContent = content
+		case []interface{}:
+			dat, _ := json.Marshal(content)
+			c := []Content{}
+			json.Unmarshal(dat, &c)
+			for _, pp := range c {
+				if pp.Type == "text" {
+					input.Messages[i].StringContent = pp.Text
+				} else if pp.Type == "image_url" {
+					// Detect if pp.ImageURL is an URL, if it is download the image and encode it in base64:
+					base64, err := utils.GetBase64Image(pp.ImageURL.URL)
+					if err == nil {
+						input.Messages[i].StringImages = append(input.Messages[i].StringImages, base64) // TODO: make sure that we only return base64 stuff
+						// set a placeholder for each image
+						input.Messages[i].StringContent = fmt.Sprintf("[img-%d]", index) + input.Messages[i].StringContent
+						index++
+					} else {
+						fmt.Print("Failed encoding image", err)
+					}
+
+				}
+			}
+		}
+	}
+
+	// TODO: check that this was merged correctly? I _think_ it is?
+	switch inputs := input.Input.(type) {
+	case string:
+		if inputs != "" {
+			config.InputStrings = append(config.InputStrings, inputs)
+		}
+	case []interface{}:
+		for _, pp := range inputs {
+			switch i := pp.(type) {
+			case string:
+				config.InputStrings = append(config.InputStrings, i)
+			case []interface{}:
+				tokens := []int{}
+				for _, ii := range i {
+					tokens = append(tokens, int(ii.(float64)))
+				}
+				config.InputToken = append(config.InputToken, tokens)
+			}
+		}
+	}
+
+	// Can be either a string or an object
+	switch fnc := input.FunctionCall.(type) {
+	case string:
+		if fnc != "" {
+			config.SetFunctionCallString(fnc)
+		}
+	case map[string]interface{}:
+		var name string
+		n, exists := fnc["name"]
+		if exists {
+			nn, e := n.(string)
+			if e {
+				name = nn
+			}
+		}
+		config.SetFunctionCallNameString(name)
+	}
+
+	switch p := input.Prompt.(type) {
+	case string:
+		config.PromptStrings = append(config.PromptStrings, p)
+	case []interface{}:
+		for _, pp := range p {
+			if s, ok := pp.(string); ok {
+				config.PromptStrings = append(config.PromptStrings, s)
+			}
+		}
+	}
+
+}

pkg/schema/config_test.go

@@ -0,0 +1,51 @@
+package schema_test
+
+import (
+	"os"
+
+	"github.com/go-skynet/LocalAI/core/services"
+	"github.com/go-skynet/LocalAI/pkg/schema"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("Test cases for config related functions", func() {
+
+	var (
+		configFile string
+	)
+
+	Context("Test Read configuration functions", func() {
+		configFile = os.Getenv("CONFIG_FILE")
+		It("Test ReadConfigFile", func() {
+			config, err := schema.ReadConfigFile(configFile)
+			Expect(err).To(BeNil())
+			Expect(config).ToNot(BeNil())
+			// two configs in config.yaml
+			Expect(config[0].Name).To(Equal("list1"))
+			Expect(config[1].Name).To(Equal("list2"))
+		})
+
+		It("Test LoadConfigs", func() {
+			cm := services.NewConfigLoader()
+			err := cm.LoadConfigs(os.Getenv("MODELS_PATH"))
+			Expect(err).To(BeNil())
+			Expect(cm.ListConfigs()).ToNot(BeNil())
+
+			// config should includes gpt4all models's api.config
+			Expect(cm.ListConfigs()).To(ContainElements("gpt4all"))
+
+			// config should includes gpt2 models's api.config
+			Expect(cm.ListConfigs()).To(ContainElements("gpt4all-2"))
+
+			// config should includes text-embedding-ada-002 models's api.config
+			Expect(cm.ListConfigs()).To(ContainElements("text-embedding-ada-002"))
+
+			// config should includes rwkv_test models's api.config
+			Expect(cm.ListConfigs()).To(ContainElements("rwkv_test"))
+
+			// config should includes whisper-1 models's api.config
+			Expect(cm.ListConfigs()).To(ContainElements("whisper-1"))
+		})
+	})
+})

pkg/schema/localai.go

@@ -0,0 +1,39 @@
+package schema
+
+import (
+	"context"
+
+	gopsutil "github.com/shirou/gopsutil/v3/process"
+
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/metric"
+)
+
+type BackendMonitorRequest struct {
+	Model string `json:"model" yaml:"model"`
+}
+
+type BackendMonitorResponse struct {
+	MemoryInfo    *gopsutil.MemoryInfoStat
+	MemoryPercent float32
+	CPUPercent    float64
+}
+
+type TTSRequest struct {
+	Model   string `json:"model" yaml:"model"`
+	Input   string `json:"input" yaml:"input"`
+	Backend string `json:"backend" yaml:"backend"`
+}
+
+type LocalAIMetrics struct {
+	Meter         metric.Meter
+	ApiTimeMetric metric.Float64Histogram
+}
+
+func (m *LocalAIMetrics) ObserveAPICall(method string, path string, duration float64) {
+	opts := metric.WithAttributes(
+		attribute.String("method", method),
+		attribute.String("path", path),
+	)
+	m.ApiTimeMetric.Record(context.Background(), duration, opts)
+}

pkg/schema/openai.go

@@ -0,0 +1,133 @@
+package schema
+
+import (
+	"context"
+
+	"github.com/go-skynet/LocalAI/pkg/grammar"
+)
+
+// APIError provides error information returned by the OpenAI API.
+type APIError struct {
+	Code    any     `json:"code,omitempty"`
+	Message string  `json:"message"`
+	Param   *string `json:"param,omitempty"`
+	Type    string  `json:"type"`
+}
+
+type ErrorResponse struct {
+	Error *APIError `json:"error,omitempty"`
+}
+
+type OpenAIUsage struct {
+	PromptTokens     int `json:"prompt_tokens"`
+	CompletionTokens int `json:"completion_tokens"`
+	TotalTokens      int `json:"total_tokens"`
+}
+
+type Item struct {
+	Embedding []float32 `json:"embedding"`
+	Index     int       `json:"index"`
+	Object    string    `json:"object,omitempty"`
+
+	// Images
+	URL     string `json:"url,omitempty"`
+	B64JSON string `json:"b64_json,omitempty"`
+}
+
+type OpenAIResponse struct {
+	Created int      `json:"created,omitempty"`
+	Object  string   `json:"object,omitempty"`
+	ID      string   `json:"id,omitempty"`
+	Model   string   `json:"model,omitempty"`
+	Choices []Choice `json:"choices,omitempty"`
+	Data    []Item   `json:"data,omitempty"`
+
+	Usage OpenAIUsage `json:"usage"`
+}
+
+type Choice struct {
+	Index        int      `json:"index"`
+	FinishReason string   `json:"finish_reason,omitempty"`
+	Message      *Message `json:"message,omitempty"`
+	Delta        *Message `json:"delta,omitempty"`
+	Text         string   `json:"text,omitempty"`
+}
+
+type Content struct {
+	Type     string     `json:"type" yaml:"type"`
+	Text     string     `json:"text" yaml:"text"`
+	ImageURL ContentURL `json:"image_url" yaml:"image_url"`
+}
+
+type ContentURL struct {
+	URL string `json:"url" yaml:"url"`
+}
+
+type Message struct {
+	// The message role
+	Role string `json:"role,omitempty" yaml:"role"`
+	// The message content
+	Content interface{} `json:"content" yaml:"content"`
+
+	StringContent string   `json:"string_content,omitempty" yaml:"string_content,omitempty"`
+	StringImages  []string `json:"string_images,omitempty" yaml:"string_images,omitempty"`
+
+	// A result of a function call
+	FunctionCall interface{} `json:"function_call,omitempty" yaml:"function_call,omitempty"`
+}
+
+type OpenAIModel struct {
+	ID     string `json:"id"`
+	Object string `json:"object"`
+}
+
+type ChatCompletionResponseFormatType string
+
+type ChatCompletionResponseFormat struct {
+	Type ChatCompletionResponseFormatType `json:"type,omitempty"`
+}
+
+type OpenAIRequest struct {
+	PredictionOptions
+
+	Context context.Context
+	Cancel  context.CancelFunc
+
+	// whisper
+	File string `json:"file" validate:"required"`
+	//whisper/image
+	ResponseFormat ChatCompletionResponseFormat `json:"response_format"`
+	// image
+	Size string `json:"size"`
+	// Prompt is read only by completion/image API calls
+	Prompt interface{} `json:"prompt" yaml:"prompt"`
+
+	// Edit endpoint
+	Instruction string      `json:"instruction" yaml:"instruction"`
+	Input       interface{} `json:"input" yaml:"input"`
+
+	Stop interface{} `json:"stop" yaml:"stop"`
+
+	// Messages is read only by chat/completion API calls
+	Messages []Message `json:"messages" yaml:"messages"`
+
+	// A list of available functions to call
+	Functions    []grammar.Function `json:"functions" yaml:"functions"`
+	FunctionCall interface{}        `json:"function_call" yaml:"function_call"` // might be a string or an object
+
+	Stream bool `json:"stream"`
+
+	// Image (not supported by OpenAI)
+	Mode int `json:"mode"`
+	Step int `json:"step"`
+
+	// A grammar to constrain the LLM output
+	Grammar string `json:"grammar" yaml:"grammar"`
+
+	JSONFunctionGrammarObject *grammar.JSONFunctionStructure `json:"grammar_json_functions" yaml:"grammar_json_functions"`
+
+	Backend string `json:"backend" yaml:"backend"`
+
+	// AutoGPTQ
+	ModelBaseName string `json:"model_base_name" yaml:"model_base_name"`
+}

pkg/schema/prediction.go

@@ -0,0 +1,50 @@
+package schema
+
+type PredictionOptions struct {
+
+	// Also part of the OpenAI official spec
+	Model string `json:"model" yaml:"model"`
+
+	// Also part of the OpenAI official spec
+	Language string `json:"language"`
+
+	// Also part of the OpenAI official spec. use it for returning multiple results
+	N int `json:"n"`
+
+	// Common options between all the API calls, part of the OpenAI spec
+	TopP        float64 `json:"top_p" yaml:"top_p"`
+	TopK        int     `json:"top_k" yaml:"top_k"`
+	Temperature float64 `json:"temperature" yaml:"temperature"`
+	Maxtokens   int     `json:"max_tokens" yaml:"max_tokens"`
+	Echo        bool    `json:"echo"`
+
+	// Custom parameters - not present in the OpenAI API
+	Batch         int     `json:"batch" yaml:"batch"`
+	F16           bool    `json:"f16" yaml:"f16"`
+	IgnoreEOS     bool    `json:"ignore_eos" yaml:"ignore_eos"`
+	RepeatPenalty float64 `json:"repeat_penalty" yaml:"repeat_penalty"`
+	Keep          int     `json:"n_keep" yaml:"n_keep"`
+
+	MirostatETA float64 `json:"mirostat_eta" yaml:"mirostat_eta"`
+	MirostatTAU float64 `json:"mirostat_tau" yaml:"mirostat_tau"`
+	Mirostat    int     `json:"mirostat" yaml:"mirostat"`
+
+	FrequencyPenalty float64 `json:"frequency_penalty" yaml:"frequency_penalty"`
+	TFZ              float64 `json:"tfz" yaml:"tfz"`
+
+	TypicalP float64 `json:"typical_p" yaml:"typical_p"`
+	Seed     int     `json:"seed" yaml:"seed"`
+
+	NegativePrompt      string  `json:"negative_prompt" yaml:"negative_prompt"`
+	RopeFreqBase        float32 `json:"rope_freq_base" yaml:"rope_freq_base"`
+	RopeFreqScale       float32 `json:"rope_freq_scale" yaml:"rope_freq_scale"`
+	NegativePromptScale float32 `json:"negative_prompt_scale" yaml:"negative_prompt_scale"`
+	// AutoGPTQ
+	UseFastTokenizer bool `json:"use_fast_tokenizer" yaml:"use_fast_tokenizer"`
+
+	// Diffusers
+	ClipSkip int `json:"clip_skip" yaml:"clip_skip"`
+
+	// RWKV (?)
+	Tokenizer string `json:"tokenizer" yaml:"tokenizer"`
+}

pkg/schema/startup_options.go

@@ -0,0 +1,260 @@
+package schema
+
+import (
+	"context"
+	"embed"
+	"encoding/json"
+	"time"
+
+	"github.com/go-skynet/LocalAI/pkg/gallery"
+	"github.com/rs/zerolog/log"
+)
+
+type StartupOptions struct {
+	Context                             context.Context
+	ConfigFile                          string
+	ModelPath                           string
+	UploadLimitMB, Threads, ContextSize int
+	F16                                 bool
+	Debug, DisableMessage               bool
+	ImageDir                            string
+	AudioDir                            string
+	CORS                                bool
+	PreloadJSONModels                   string
+	PreloadModelsFromPath               string
+	CORSAllowOrigins                    string
+	ApiKeys                             []string
+	Metrics                             *LocalAIMetrics
+
+	Galleries []gallery.Gallery
+
+	BackendAssets     embed.FS
+	AssetsDestination string
+
+	ExternalGRPCBackends map[string]string
+
+	AutoloadGalleries bool
+
+	SingleBackend           bool
+	ParallelBackendRequests bool
+
+	WatchDogIdle bool
+	WatchDogBusy bool
+	WatchDog     bool
+
+	ModelsURL []string
+
+	WatchDogBusyTimeout, WatchDogIdleTimeout time.Duration
+
+	LocalAIConfigDir string
+}
+
+type AppOption func(*StartupOptions)
+
+func NewStartupOptions(o ...AppOption) *StartupOptions {
+	opt := &StartupOptions{
+		Context:        context.Background(),
+		UploadLimitMB:  15,
+		Threads:        1,
+		ContextSize:    512,
+		Debug:          true,
+		DisableMessage: true,
+	}
+	for _, oo := range o {
+		oo(opt)
+	}
+	return opt
+}
+
+func WithModelsURL(urls ...string) AppOption {
+	return func(o *StartupOptions) {
+		o.ModelsURL = urls
+	}
+}
+
+func WithCors(b bool) AppOption {
+	return func(o *StartupOptions) {
+		o.CORS = b
+	}
+}
+
+var EnableWatchDog = func(o *StartupOptions) {
+	o.WatchDog = true
+}
+
+var EnableWatchDogIdleCheck = func(o *StartupOptions) {
+	o.WatchDog = true
+	o.WatchDogIdle = true
+}
+
+var EnableWatchDogBusyCheck = func(o *StartupOptions) {
+	o.WatchDog = true
+	o.WatchDogBusy = true
+}
+
+func SetWatchDogBusyTimeout(t time.Duration) AppOption {
+	return func(o *StartupOptions) {
+		o.WatchDogBusyTimeout = t
+	}
+}
+
+func SetWatchDogIdleTimeout(t time.Duration) AppOption {
+	return func(o *StartupOptions) {
+		o.WatchDogIdleTimeout = t
+	}
+}
+
+var EnableSingleBackend = func(o *StartupOptions) {
+	o.SingleBackend = true
+}
+
+var EnableParallelBackendRequests = func(o *StartupOptions) {
+	o.ParallelBackendRequests = true
+}
+
+var EnableGalleriesAutoload = func(o *StartupOptions) {
+	o.AutoloadGalleries = true
+}
+
+func WithExternalBackend(name string, uri string) AppOption {
+	return func(o *StartupOptions) {
+		if o.ExternalGRPCBackends == nil {
+			o.ExternalGRPCBackends = make(map[string]string)
+		}
+		o.ExternalGRPCBackends[name] = uri
+	}
+}
+
+func WithCorsAllowOrigins(b string) AppOption {
+	return func(o *StartupOptions) {
+		o.CORSAllowOrigins = b
+	}
+}
+
+func WithBackendAssetsOutput(out string) AppOption {
+	return func(o *StartupOptions) {
+		o.AssetsDestination = out
+	}
+}
+
+func WithBackendAssets(f embed.FS) AppOption {
+	return func(o *StartupOptions) {
+		o.BackendAssets = f
+	}
+}
+
+func WithStringGalleries(galls string) AppOption {
+	return func(o *StartupOptions) {
+		if galls == "" {
+			log.Debug().Msgf("no galleries to load")
+			o.Galleries = []gallery.Gallery{}
+			return
+		}
+		var galleries []gallery.Gallery
+		if err := json.Unmarshal([]byte(galls), &galleries); err != nil {
+			log.Error().Msgf("failed loading galleries: %s", err.Error())
+		}
+		o.Galleries = append(o.Galleries, galleries...)
+	}
+}
+
+func WithGalleries(galleries []gallery.Gallery) AppOption {
+	return func(o *StartupOptions) {
+		o.Galleries = append(o.Galleries, galleries...)
+	}
+}
+
+func WithContext(ctx context.Context) AppOption {
+	return func(o *StartupOptions) {
+		o.Context = ctx
+	}
+}
+
+func WithYAMLConfigPreload(configFile string) AppOption {
+	return func(o *StartupOptions) {
+		o.PreloadModelsFromPath = configFile
+	}
+}
+
+func WithJSONStringPreload(configFile string) AppOption {
+	return func(o *StartupOptions) {
+		o.PreloadJSONModels = configFile
+	}
+}
+func WithConfigFile(configFile string) AppOption {
+	return func(o *StartupOptions) {
+		o.ConfigFile = configFile
+	}
+}
+
+func WithModelPath(path string) AppOption {
+	return func(o *StartupOptions) {
+		o.ModelPath = path
+	}
+}
+
+func WithUploadLimitMB(limit int) AppOption {
+	return func(o *StartupOptions) {
+		o.UploadLimitMB = limit
+	}
+}
+
+func WithThreads(threads int) AppOption {
+	return func(o *StartupOptions) {
+		o.Threads = threads
+	}
+}
+
+func WithContextSize(ctxSize int) AppOption {
+	return func(o *StartupOptions) {
+		o.ContextSize = ctxSize
+	}
+}
+
+func WithF16(f16 bool) AppOption {
+	return func(o *StartupOptions) {
+		o.F16 = f16
+	}
+}
+
+func WithDebug(debug bool) AppOption {
+	return func(o *StartupOptions) {
+		o.Debug = debug
+	}
+}
+
+func WithDisableMessage(disableMessage bool) AppOption {
+	return func(o *StartupOptions) {
+		o.DisableMessage = disableMessage
+	}
+}
+
+func WithAudioDir(audioDir string) AppOption {
+	return func(o *StartupOptions) {
+		o.AudioDir = audioDir
+	}
+}
+
+func WithImageDir(imageDir string) AppOption {
+	return func(o *StartupOptions) {
+		o.ImageDir = imageDir
+	}
+}
+
+func WithApiKeys(apiKeys []string) AppOption {
+	return func(o *StartupOptions) {
+		o.ApiKeys = apiKeys
+	}
+}
+
+func WithMetrics(metrics *LocalAIMetrics) AppOption {
+	return func(o *StartupOptions) {
+		o.Metrics = metrics
+	}
+}
+
+func WithLocalAIConfigDir(configDir string) AppOption {
+	return func(o *StartupOptions) {
+		o.LocalAIConfigDir = configDir
+	}
+}

pkg/schema/whisper.go

@@ -0,0 +1,16 @@
+package schema
+
+import "time"
+
+type WhisperSegment struct {
+	Id     int           `json:"id"`
+	Start  time.Duration `json:"start"`
+	End    time.Duration `json:"end"`
+	Text   string        `json:"text"`
+	Tokens []int         `json:"tokens"`
+}
+
+type WhisperResult struct {
+	Segments []WhisperSegment `json:"segments"`
+	Text     string           `json:"text"`
+}