Merge branch 'main' of github.com:djthorpe/go-whisper

Makefile

@@ -16,11 +16,11 @@ BUILD_DIR := build
 
 # Build flags
 BUILD_MODULE := $(shell cat go.mod | head -1 | cut -d ' ' -f 2)
-BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/whisper/version.GitSource=${BUILD_MODULE}
-BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/whisper/version.GitTag=$(shell git describe --tags --always)
-BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/whisper/version.GitBranch=$(shell git name-rev HEAD --name-only --always)
-BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/whisper/version.GitHash=$(shell git rev-parse HEAD)
-BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/whisper/version.GoBuildTime=$(shell date -u '+%Y-%m-%dT%H:%M:%SZ')
+BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/version.GitSource=${BUILD_MODULE}
+BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/version.GitTag=$(shell git describe --tags --always)
+BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/version.GitBranch=$(shell git name-rev HEAD --name-only --always)
+BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/version.GitHash=$(shell git rev-parse HEAD)
+BUILD_LD_FLAGS += -X $(BUILD_MODULE)/pkg/version.GoBuildTime=$(shell date -u '+%Y-%m-%dT%H:%M:%SZ')
 BUILD_FLAGS = -ldflags "-s -w $(BUILD_LD_FLAGS)" 
 
 # If GGML_CUDA is set, then add a cuda tag for the go ${BUILD FLAGS}
@@ -30,7 +30,7 @@ ifeq ($(GGML_CUDA),1)
 endif
 
 # Targets
-all: whisper
+all: whisper api
 
 # Generate the pkg-config files
 generate: mkdir go-tidy
@@ -42,6 +42,11 @@ whisper: mkdir generate go-tidy libwhisper libggml
 	@echo "Building whisper"
 	@PKG_CONFIG_PATH=${ROOT_PATH}/${BUILD_DIR} ${GO} build ${BUILD_FLAGS} -o ${BUILD_DIR}/whisper ./cmd/whisper
 
+# Make api
+api: mkdir go-tidy
+	@echo "Building api"
+	@${GO} build ${BUILD_FLAGS} -o ${BUILD_DIR}/api ./cmd/api
+
 # Build docker container
 docker: docker-dep submodule
 	@echo build docker image: ${BUILD_TAG} for ${OS}/${ARCH}
@@ -51,14 +56,16 @@ docker: docker-dep submodule
 		--build-arg OS=${OS} \
 		--build-arg SOURCE=${BUILD_MODULE} \
 		--build-arg VERSION=${VERSION} \
-		-f etc/Dockerfile.${OS}-${ARCH} .
+		-f etc/Dockerfile .
 
 # Test whisper bindings
 test: generate libwhisper libggml
 	@echo "Running tests (sys)"
 	@PKG_CONFIG_PATH=${ROOT_PATH}/${BUILD_DIR} ${GO} test -v ./sys/whisper/...
 	@echo "Running tests (pkg)"
-	@PKG_CONFIG_PATH=${ROOT_PATH}/${BUILD_DIR} ${GO} test -v ./pkg/whisper/...
+	@PKG_CONFIG_PATH=${ROOT_PATH}/${BUILD_DIR} ${GO} test -v ./pkg/...
+	@echo "Running tests (whisper)"
+	@PKG_CONFIG_PATH=${ROOT_PATH}/${BUILD_DIR} ${GO} test -v ./
 
 # Build whisper-static-library
 libwhisper: submodule

README.md

@@ -2,7 +2,7 @@
 
 Speech-to-Text in golang. This is an early development version.
 
-* `cmd` contains an OpenAI-API compatible server
+* `cmd` contains an OpenAI-API compatible service
 * `pkg` contains the `whisper` service and client
 * `sys` contains the `whisper` bindings to the `whisper.cpp` library
 * `third_party` is a submodule for the whisper.cpp source
@@ -11,6 +11,7 @@ Speech-to-Text in golang. This is an early development version.
 
 (Note: Docker images are not created yet - this is some forward planning!)
 
+You can either run the whisper service as a CLI command or in a docker container.
 There are docker images for arm64 and amd64 (Intel). The arm64 image is built for
 Jetson GPU support specifically, but it will also run on Raspberry Pi's.
 
@@ -19,14 +20,15 @@ In order to utilize a NVIDIA GPU, you'll need to install the
 
 A docker volume should be created called "whisper" can be used for storing the Whisper language
 models. You can see which models are available to download locally [here](https://huggingface.co/ggerganov/whisper.cpp).
-The following command will run the server on port 8080:
+
+The following command will run the server on port 8080 for an NVIDIA GPU:
 
 ```bash
 docker run \
   --name whisper-server --rm \
   --runtime nvidia --gpus all \ # When using a NVIDIA GPU
   -v whisper:/models -p 8080:8080 -e WHISPER_DATA=/models \
-  ghcr.io/mutablelogic/go-whisper:latest
+  ghcr.io/mutablelogic/go-whisper
 ```
 
 If you include a `-debug` flag at the end, you'll get more verbose output. The API is then
@@ -92,6 +94,10 @@ The following `Makefile` targets can be used:
 
 See all the other targets in the `Makefile` for more information.
 
+## Developing
+
+TODO
+
 ## Status
 
 Still in development. See this [issue](https://github.com/mutablelogic/go-whisper/issues/1) for

cmd/api/delete.go

@@ -0,0 +1,12 @@
+package main
+
+type DeleteCmd struct {
+	Model string `arg:"" name:"model" help:"Model to delete"`
+}
+
+func (cmd *DeleteCmd) Run(ctx *Globals) error {
+	if err := ctx.client.DeleteModel(ctx.ctx, cmd.Model); err != nil {
+		return err
+	}
+	return nil
+}

cmd/api/download.go

@@ -0,0 +1,22 @@
+package main
+
+import (
+	"fmt"
+
+	"github.com/djthorpe/go-tablewriter"
+)
+
+type DownloadCmd struct {
+	Model string `arg:"" name:"model" help:"Model to download (must end in .bin)"`
+}
+
+func (cmd *DownloadCmd) Run(ctx *Globals) error {
+	model, err := ctx.client.DownloadModel(ctx.ctx, cmd.Model, func(status string, cur, total int64) {
+		pct := fmt.Sprintf("%02d%%", int(100*float64(cur)/float64(total)))
+		ctx.writer.Writeln(pct, status)
+	})
+	if err != nil {
+		return err
+	}
+	return ctx.writer.Write(model, tablewriter.OptHeader())
+}

cmd/api/main.go

@@ -0,0 +1,104 @@
+package main
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"syscall"
+
+	// Packages
+	kong "github.com/alecthomas/kong"
+	tablewriter "github.com/djthorpe/go-tablewriter"
+	opt "github.com/mutablelogic/go-client"
+	ctx "github.com/mutablelogic/go-server/pkg/context"
+	client "github.com/mutablelogic/go-whisper/pkg/client"
+)
+
+////////////////////////////////////////////////////////////////////////////////
+// TYPES
+
+type Globals struct {
+	Url   string `name:"url" help:"URL of whisper service (can be set from WHISPER_URL env)" default:"${WHISPER_URL}"`
+	Debug bool   `name:"debug" help:"Enable debug output"`
+
+	// Writer, service and context
+	writer *tablewriter.Writer
+	client *client.Client
+	ctx    context.Context
+}
+
+type CLI struct {
+	Globals
+
+	Ping     PingCmd     `cmd:"ping" help:"Ping the whisper service"`
+	Models   ModelsCmd   `cmd:"models" help:"List models"`
+	Download DownloadCmd `cmd:"download" help:"Download a model"`
+	Delete   DeleteCmd   `cmd:"delete" help:"Delete a model"`
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// GLOBALS
+
+const (
+	defaultEndpoint = "http://localhost:8080/api/v1"
+)
+
+////////////////////////////////////////////////////////////////////////////////
+// MAIN
+
+func main() {
+	// The name of the executable
+	name, err := os.Executable()
+	if err != nil {
+		panic(err)
+	} else {
+		name = filepath.Base(name)
+	}
+
+	// Create a cli parser
+	cli := CLI{}
+	cmd := kong.Parse(&cli,
+		kong.Name(name),
+		kong.Description("speech transcription and translation service client"),
+		kong.UsageOnError(),
+		kong.ConfigureHelp(kong.HelpOptions{Compact: true}),
+		kong.Vars{
+			"WHISPER_URL": envOrDefault("WHISPER_URL", defaultEndpoint),
+		},
+	)
+
+	// Set whisper client options
+	opts := []opt.ClientOpt{}
+	if cli.Globals.Debug {
+		opts = append(opts, opt.OptTrace(os.Stderr, true))
+	}
+
+	// Create a whisper client
+	client, err := client.New(cli.Globals.Url, opts...)
+	if err != nil {
+		cmd.FatalIfErrorf(err)
+		return
+	} else {
+		cli.Globals.client = client
+	}
+
+	// Create a tablewriter object with text output
+	writer := tablewriter.New(os.Stdout, tablewriter.OptOutputText())
+	cli.Globals.writer = writer
+
+	// Create a context
+	cli.Globals.ctx = ctx.ContextForSignal(os.Interrupt, syscall.SIGQUIT)
+
+	// Run the command
+	if err := cmd.Run(&cli.Globals); err != nil {
+		cmd.FatalIfErrorf(err)
+	}
+}
+
+func envOrDefault(name, def string) string {
+	if value := os.Getenv(name); value != "" {
+		return value
+	} else {
+		return def
+	}
+}

cmd/api/models.go

@@ -0,0 +1,13 @@
+package main
+
+import "github.com/djthorpe/go-tablewriter"
+
+type ModelsCmd struct{}
+
+func (cmd *ModelsCmd) Run(ctx *Globals) error {
+	if models, err := ctx.client.ListModels(ctx.ctx); err != nil {
+		return err
+	} else {
+		return ctx.writer.Write(models, tablewriter.OptHeader())
+	}
+}

cmd/api/ping.go

@@ -0,0 +1,10 @@
+package main
+
+type PingCmd struct{}
+
+func (cmd *PingCmd) Run(ctx *Globals) error {
+	if err := ctx.client.Ping(ctx.ctx); err != nil {
+		return err
+	}
+	return ctx.writer.Writeln("OK")
+}