package backend

Get desktop application:
View/edit binary Protocol Buffers messages

• rpc AudioTranscription (TranscriptRequest, TranscriptResult)

  backend.proto:18

  message TranscriptRequest

  backend.proto:274

  • string dst = 2

  • string language = 3

  • uint32 threads = 4

  • bool translate = 5

  • bool diarize = 6

  message TranscriptResult

  backend.proto:282

  • repeated TranscriptSegment segments = 1

  • string text = 2

• rpc Detect (DetectOptions, DetectResponse)

  backend.proto:23

  message DetectOptions

  backend.proto:387

  • string src = 1

  message DetectResponse

  backend.proto:400

  • repeated Detection Detections = 1

• rpc Embedding (PredictOptions, EmbeddingResult)

  backend.proto:15

  message EmbeddingResult

  backend.proto:270

  • repeated float embeddings = 1

• rpc GenerateImage (GenerateImageRequest, Result)

  backend.proto:16

  message GenerateImageRequest

  backend.proto:295

  • int32 height = 1

  • int32 width = 2

  • int32 mode = 3

  • int32 step = 4

  • int32 seed = 5

  • string positive_prompt = 6

  • string negative_prompt = 7

  • string dst = 8

  • string src = 9

  • string EnableParameters = 10

    Diffusers

  • int32 CLIPSkip = 11

  • repeated string ref_images = 12

    Reference images for models that support them (e.g., Flux Kontext)

• rpc GenerateVideo (GenerateVideoRequest, Result)

  backend.proto:17

  message GenerateVideoRequest

  backend.proto:314

  • string prompt = 1

  • string negative_prompt = 2

    Negative prompt for video generation

  • string start_image = 3

    Path or base64 encoded image for the start frame

  • string end_image = 4

    Path or base64 encoded image for the end frame

  • int32 width = 5

  • int32 height = 6

  • int32 num_frames = 7

    Number of frames to generate

  • int32 fps = 8

    Frames per second

  • int32 seed = 9

  • float cfg_scale = 10

    Classifier-free guidance scale

  • int32 step = 11

    Number of inference steps

  • string dst = 12

    Output path for the generated video

• rpc GetMetrics (MetricsRequest, MetricsResponse)

  backend.proto:32

  message MetricsRequest

  backend.proto:38

  Define the empty request

  (message has no fields)

  message MetricsResponse

  backend.proto:40

  • int32 slot_id = 1

  • string prompt_json_for_slot = 2

    Stores the prompt as a JSON string.

  • float tokens_per_second = 3

  • int32 tokens_generated = 4

  • int32 prompt_tokens_processed = 5

• rpc Health (HealthMessage, Reply)

  backend.proto:11

• rpc LoadModel (ModelOptions, Result)

  backend.proto:13

  message ModelOptions

  backend.proto:173

  • string Model = 1

  • int32 ContextSize = 2

  • int32 Seed = 3

  • int32 NBatch = 4

  • bool F16Memory = 5

  • bool MLock = 6

  • bool MMap = 7

  • bool VocabOnly = 8

  • bool LowVRAM = 9

  • bool Embeddings = 10

  • bool NUMA = 11

  • int32 NGPULayers = 12

  • string MainGPU = 13

  • string TensorSplit = 14

  • int32 Threads = 15

  • float RopeFreqBase = 17

  • float RopeFreqScale = 18

  • float RMSNormEps = 19

  • int32 NGQA = 20

  • string ModelFile = 21

  • string PipelineType = 26

    Diffusers

  • string SchedulerType = 27

  • bool CUDA = 28

  • float CFGScale = 29

  • bool IMG2IMG = 30

  • string CLIPModel = 31

  • string CLIPSubfolder = 32

  • int32 CLIPSkip = 33

  • string ControlNet = 48

  • string Tokenizer = 34

  • string LoraBase = 35

    LLM (llama.cpp)

  • string LoraAdapter = 36

  • float LoraScale = 42

  • bool NoMulMatQ = 37

  • string DraftModel = 39

  • string AudioPath = 38

  • string Quantization = 40

    vllm

  • float GPUMemoryUtilization = 50

  • bool TrustRemoteCode = 51

  • bool EnforceEager = 52

  • int32 SwapSpace = 53

  • int32 MaxModelLen = 54

  • int32 TensorParallelSize = 55

  • string LoadFormat = 58

  • bool DisableLogStatus = 66

  • string DType = 67

  • int32 LimitImagePerPrompt = 68

  • int32 LimitVideoPerPrompt = 69

  • int32 LimitAudioPerPrompt = 70

  • string MMProj = 41

  • string RopeScaling = 43

  • float YarnExtFactor = 44

  • float YarnAttnFactor = 45

  • float YarnBetaFast = 46

  • float YarnBetaSlow = 47

  • string Type = 49

  • string FlashAttention = 56

  • bool NoKVOffload = 57

  • string ModelPath = 59

  • repeated string LoraAdapters = 60

  • repeated float LoraScales = 61

  • repeated string Options = 62

  • string CacheTypeKey = 63

  • string CacheTypeValue = 64

  • repeated GrammarTrigger GrammarTriggers = 65

  • bool Reranking = 71

  • repeated string Overrides = 72

• rpc Predict (PredictOptions, Reply)

  backend.proto:12

• rpc PredictStream (PredictOptions, stream Reply)

  backend.proto:14

• rpc Rerank (RerankRequest, RerankResult)

  backend.proto:30

  message RerankRequest

  backend.proto:48

  • string query = 1

  • repeated string documents = 2

  • int32 top_n = 3

  message RerankResult

  backend.proto:54

  • optional Usage usage = 1

  • repeated DocumentResult results = 2

• rpc SoundGeneration (SoundGenerationRequest, Result)

  backend.proto:20

  message SoundGenerationRequest

  backend.proto:350

  • string text = 1

  • string model = 2

  • string dst = 3

  • optional float duration = 4

  • optional float temperature = 5

  • optional bool sample = 6

  • optional string src = 7

  • optional int32 src_divisor = 8

• rpc Status (HealthMessage, StatusResponse)

  backend.proto:22

  message StatusResponse

  backend.proto:371

  • StatusResponse.State state = 1

  • optional MemoryUsageData memory = 2

• rpc StoresDelete (StoresDeleteOptions, Result)

  backend.proto:26

  message StoresDeleteOptions

  backend.proto:83

  • repeated StoresKey Keys = 1

• rpc StoresFind (StoresFindOptions, StoresFindResult)

  backend.proto:28

  message StoresFindOptions

  backend.proto:96

  • optional StoresKey Key = 1

  • int32 TopK = 2

  message StoresFindResult

  backend.proto:101

  • repeated StoresKey Keys = 1

  • repeated StoresValue Values = 2

  • repeated float Similarities = 3

• rpc StoresGet (StoresGetOptions, StoresGetResult)

  backend.proto:27

  message StoresGetOptions

  backend.proto:87

  • repeated StoresKey Keys = 1

  message StoresGetResult

  backend.proto:91

  • repeated StoresKey Keys = 1

  • repeated StoresValue Values = 2

• rpc StoresSet (StoresSetOptions, Result)

  backend.proto:25

  message StoresSetOptions

  backend.proto:78

  • repeated StoresKey Keys = 1

  • repeated StoresValue Values = 2

• rpc TTS (TTSRequest, Result)

  backend.proto:19

  message TTSRequest

  backend.proto:329

  • string text = 1

  • string model = 2

  • string dst = 3

  • string voice = 4

  • optional string language = 5

• rpc TokenizeString (PredictOptions, TokenizationResponse)

  backend.proto:21

  message TokenizationResponse

  backend.proto:361

  • int32 length = 1

  • repeated int32 tokens = 2

• rpc VAD (VADRequest, VADResponse)

  backend.proto:34

  message VADRequest

  backend.proto:337

  • repeated float audio = 1

  message VADResponse

  backend.proto:346

  • repeated VADSegment segments = 1

Used in: DetectResponse

• float x = 1

• float y = 2

• float width = 3

• float height = 4

• float confidence = 5

• string class_name = 6

Used in: RerankResult

• int32 index = 1

• string text = 2

• float relevance_score = 3

Used in: ModelOptions

• string word = 1

Used as request type in: Backend.Health, Backend.Status

(message has no fields)

Used in: StatusResponse

• uint64 total = 1

• map<string, uint64> breakdown = 2

Used in: PredictOptions

• string role = 1

• string content = 2

The request message containing the user's name.

Used as request type in: Backend.Embedding, Backend.Predict, Backend.PredictStream, Backend.TokenizeString

• string Prompt = 1

• int32 Seed = 2

• int32 Threads = 3

• int32 Tokens = 4

• int32 TopK = 5

• int32 Repeat = 6

• int32 Batch = 7

• int32 NKeep = 8

• float Temperature = 9

• float Penalty = 10

• bool F16KV = 11

• bool DebugMode = 12

• repeated string StopPrompts = 13

• bool IgnoreEOS = 14

• float TailFreeSamplingZ = 15

• float TypicalP = 16

• float FrequencyPenalty = 17

• float PresencePenalty = 18

• int32 Mirostat = 19

• float MirostatETA = 20

• float MirostatTAU = 21

• bool PenalizeNL = 22

• string LogitBias = 23

• bool MLock = 25

• bool MMap = 26

• bool PromptCacheAll = 27

• bool PromptCacheRO = 28

• string Grammar = 29

• string MainGPU = 30

• string TensorSplit = 31

• float TopP = 32

• string PromptCachePath = 33

• bool Debug = 34

• repeated int32 EmbeddingTokens = 35

• string Embeddings = 36

• float RopeFreqBase = 37

• float RopeFreqScale = 38

• float NegativePromptScale = 39

• string NegativePrompt = 40

• int32 NDraft = 41

• repeated string Images = 42

• bool UseTokenizerTemplate = 43

• repeated Message Messages = 44

• repeated string Videos = 45

• repeated string Audios = 46

• string CorrelationId = 47

The response message containing the result

Used as response type in: Backend.Health, Backend.Predict, Backend.PredictStream

• bytes message = 1

• int32 tokens = 2

• int32 prompt_tokens = 3

• double timing_prompt_processing = 4

• double timing_token_generation = 5

• bytes audio = 6

Used as response type in: Backend.GenerateImage, Backend.GenerateVideo, Backend.LoadModel, Backend.SoundGeneration, Backend.StoresDelete, Backend.StoresSet, Backend.TTS

• string message = 1

• bool success = 2

Used in: StatusResponse

• UNINITIALIZED = 0

• BUSY = 1

• READY = 2

• ERROR = -1

Used in: StoresDeleteOptions, StoresFindOptions, StoresFindResult, StoresGetOptions, StoresGetResult, StoresSetOptions

• repeated float Floats = 1

Used in: StoresFindResult, StoresGetResult, StoresSetOptions

• bytes Bytes = 1

Used in: TranscriptResult

• int32 id = 1

• int64 start = 2

• int64 end = 3

• string text = 4

• repeated int32 tokens = 5

Used in: RerankResult

• int32 total_tokens = 1

• int32 prompt_tokens = 2

Used in: VADResponse

• float start = 1

• float end = 2