Refactor

README.md

@@ -5,31 +5,48 @@
 
 **Status:** 🧪Experimental
 
-## Turn any model into a serverless API
+Kuda's goal is to make it **easy** and **inexpensive** to add cloud GPUs to any webapp.
 
-Easily turn any model into a serverless API that will consume cloud GPUs only
-when it's being called.
+## Serverless GPU inference
 
-Kuda deploys your API as a docker container, so you can use any language, any
-framework, and there is no library to import in your code.
+Kuda builds on [Knative](#) to allocate cloud GPUs only when there is traffic to your app.
 
-All you need is a Dockerfile.
+This is ideal when you want to share your prototypes online without keeping expensive GPUs allocated all the time.
 
-## Easy to use
+It tries to reduce cold starts time (gpu nodes allocation and service instanciation) as much possible and to tries manage cooldown times intelligently.
 
-- `kuda init` Initializes your local & remote configurations
-- `kuda dev` Deploys the API in dev mode (with file sync & live reload)
-- `kuda deploy` Deploys the API in production mode
-- `kuda publish` Publishes the API template to the registry
+## Add GPU models to a webapp easily
 
-## Features
+- Deploy a template from github
 
-- Provision GPUs & scale based on traffic (from zero to N)
-- Interactive development on cloud GPUs from any workstation
-- Protect & control access to your APIs using API Keys
-- HTTPS with TLS termination & automatic certificate management
+```bash
+kuda deploy -f github.com/cyrildiagne/gpt2-http
+```
+
+- Call your deployed API
+
+```bash
+$ curl \
+    -H 'x-api-key: $your_key' \
+    -F 'input=Kuda is' \
+    https://gpt2.<your-namespace>.kuda.cloud/generate
+```
+
+```json
+{
+  "query": "Kuda is",
+  "generated": "a tool that...etc."
+}
+```
 
-## Use the frameworks you know
+Checkout the full list of templates available in [the registry](#).
+
+## Turn any model into a serverless API
+
+Kuda deploys APIs as a docker containers, so you can use any language, any
+framework, and there is no library to import in your code.
+
+All you need is a Dockerfile.
 
 Here's a minimal example that just prints the result of `nvidia-smi` using
 [Flask](http://flask.palletsprojects.com):
@@ -70,10 +87,10 @@ deploy:
 ```
 
 Running `kuda deploy` in this example would build and deploy the API to a url
-such as `https://hello-gpu.default.yourdomain.com` which you can easily call:
+such as `https://hello-gpu.my-namespace.kuda.cloud` which you can easily call:
 
 ```
-$ curl https://hello-gpu.default.yourdomain.com
+$ curl https://hello-gpu.my-namespace.kuda.cloud
 
 Hello GPU!
 
@@ -99,6 +116,13 @@ Hello GPU!
 Checkout the full example with annotations in
 [examples/hello-gpu-flask](examples/hello-gpu-flask).
 
+## Features
+
+- Provision GPUs & scale based on traffic (from zero to N)
+- Interactive development on cloud GPUs from any workstation
+- Protect & control access to your APIs using API Keys
+- HTTPS with TLS termination & automatic certificate management
+
 ## Get Started
 
 - [Install](docs/install_cli.md)

images/deployer/main.go → cmd/api/main.go

@@ -6,7 +6,8 @@ import (
 	"net/http"
 	"os"
 
-	"github.com/cyrildiagne/kuda/pkg/deployer"
+	"github.com/cyrildiagne/kuda/pkg/api"
+	"github.com/cyrildiagne/kuda/pkg/deploy"
 	"github.com/cyrildiagne/kuda/pkg/gcloud"
 
 	"github.com/gorilla/mux"
@@ -36,7 +37,7 @@ func main() {
 		log.Fatalf("error initializing firebase: %v\n", err)
 	}
 
-	env := &deployer.Env{
+	env := &api.Env{
 		GCPProjectID: gcpProjectID,
 		DB:           fs,
 		Auth:         auth,
@@ -53,15 +54,15 @@ func main() {
 	if value, ok := os.LookupEnv("port"); ok {
 		port = value
 	}
-	fmt.Println("Starting deployer on port", port)
+	fmt.Println("Starting api on port", port)
 
 	r := mux.NewRouter()
 	r.HandleFunc("/", handleRoot).Methods("GET")
 
-	deployHandler := deployer.Handler{Env: env, H: deployer.HandleDeploy}
+	deployHandler := api.Handler{Env: env, H: deploy.HandleDeploy}
 	r.Handle("/deploy", deployHandler).Methods("POST")
 
-	publishHandler := deployer.Handler{Env: env, H: deployer.HandlePublish}
+	publishHandler := api.Handler{Env: env, H: deploy.HandlePublish}
 	r.Handle("/publish", publishHandler).Methods("POST")
 
 	http.ListenAndServe(":"+port, r)

cli/cmd/auth.go → cmd/cli/auth.go

@@ -1,4 +1,4 @@
-package cmd
+package main
 
 import (
 	"context"

cli/cmd/deploy.go → cmd/cli/deploy.go

@@ -1,4 +1,4 @@
-package cmd
+package main
 
 import (
 	"bufio"
@@ -43,7 +43,7 @@ func deployFromPublished(published string) error {
 	params.Set("from_published", published)
 	body := strings.NewReader(params.Encode())
 
-	url := cfg.Provider.DeployerURL + "/deploy"
+	url := cfg.Provider.ApiURL + "/deploy"
 	req, err := http.NewRequest("POST", url, body)
 	if err != nil {
 		return err
@@ -118,7 +118,7 @@ func deploy(manifest *latest.Manifest) error {
 	writer.Close()
 
 	// Create request.
-	url := cfg.Provider.DeployerURL + "/deploy"
+	url := cfg.Provider.ApiURL + "/deploy"
 	req, err := http.NewRequest("POST", url, body)
 	if err != nil {
 		return err

cli/cmd/dev.go → cmd/cli/dev.go

@@ -1,4 +1,4 @@
-package cmd
+package main
 
 import (
 	"fmt"

cli/cmd/init.go → cmd/cli/init.go

@@ -1,4 +1,4 @@
-package cmd
+package main
 
 import (
 	"fmt"
@@ -30,14 +30,14 @@ var initCmd = &cobra.Command{
 		if authURLFlag != "" {
 			authURL = authURLFlag
 		}
-		deployerURL := "https://deployer." + provider
-		deployerURLFlag, _ := cmd.Flags().GetString("deployer_url")
-		if deployerURLFlag != "" {
-			deployerURL = deployerURLFlag
+		apiURL := "https://api." + provider
+		apiURLFlag, _ := cmd.Flags().GetString("api_url")
+		if apiURLFlag != "" {
+			apiURL = apiURLFlag
 		}
 		newCfg.Provider = config.ProviderConfig{
-			AuthURL:     authURL,
-			DeployerURL: deployerURL,
+			AuthURL: authURL,
+			ApiURL:  apiURL,
 		}
 
 		// Start login flow.
@@ -60,7 +60,7 @@ func init() {
 
 	initCmd.Flags().StringP("provider", "p", "kuda.cloud", "Knative namespace.")
 	initCmd.Flags().String("auth_url", "", "Authentication URL.")
-	initCmd.Flags().String("deployer_url", "", "Deployer URL.")
+	initCmd.Flags().String("api_url", "", "Deployer URL.")
 }
 
 func writeConfig(cfg config.UserConfig) error {

cli/cmd/root.go → cmd/cli/main.go

@@ -1,4 +1,4 @@
-package cmd
+package main
 
 import (
 	"fmt"
@@ -26,9 +26,7 @@ var RootCmd = &cobra.Command{
 	Version: version,
 }
 
-// Execute adds all child commands to the root command and sets flags appropriately.
-// This is called by main.main(). It only needs to happen once to the RootCmd.
-func Execute() {
+func main() {
 	RootCmd.Version = version
 	if err := RootCmd.Execute(); err != nil {
 		fmt.Println(err)

cli/cmd/publish.go → cmd/cli/publish.go

@@ -1,4 +1,4 @@
-package cmd
+package main
 
 import (
 	"bytes"
@@ -51,7 +51,7 @@ func publish(manifest *latest.Manifest) error {
 	writer.Close()
 
 	// Create request.
-	url := cfg.Provider.DeployerURL + "/publish"
+	url := cfg.Provider.ApiURL + "/publish"
 	req, err := http.NewRequest("POST", url, body)
 	if err != nil {
 		return err

docs/cli.md

@@ -15,8 +15,8 @@ Initializes the local configuration.
 **Flags**
 
 - `[-p, --provider]` The provider root URL (default: `kuda.cloud`).
-- `[--auth-url]` Specify which url to use for authentication when using a remote deployer. (default: `auth.<provider>`)
-- `[--deployer-url]` Specify which url to use for deployment when using a remote deployer. (default: `deployer.<provider>`)
+- `[--auth-url]` Specify which url to use for authentication. (default: `auth.<provider>`)
+- `[--api-url]` Specify which url to use as remote api. (default: `api.<provider>`)
 
 **Examples**
 

examples/hello-gpu-flask/kuda.yaml

@@ -6,12 +6,12 @@ name: hello-gpu
 # (Optional) Version of the API.
 version: 0.1.0
 
-# 'deploy' is the config used when running `kuda deploy`.
+# `deploy` is the config used when running `kuda deploy`.
 # It has sensible defaults but you can override all the properpties.
 deploy:
   dockerfile: ./Dockerfile
 
-# 'dev' is the config used when running `kuda dev`.
+# (Optional) `dev` is used when running `kuda dev`.
 # It inherits all properties from 'deploy' which you can override individually.
 dev:
   # Use python3 to start the Flask dev server rather than gunicorn.
@@ -26,7 +26,7 @@ dev:
     - name: FLASK_ENV
       value: development
 
-# 'paths' is optional. It lets you specify how to interact with the API.
+# (Optional) `paths` lets you specify how clients interact with the API.
 # It follows the OpenAPI 3.0 specification.
 # It is only required when publishing to a registry.
 paths:

images/adapter/README.md → install/adapter/README.md

@@ -143,7 +143,7 @@ docker run --rm \
     -p 44225:44225 \
     -e FIRESTORE_CREDENTIALS="/path/YOUR_CREDENTIALS.json" \
     -v /path/to/local/credentials/folder:/secret \
-    gcr.io/kuda-project/kuda-mixer-adapter
+    gcr.io/kuda-cloud/kuda-mixer-adapter
 ```
 
 Send a mock Authorization request:

images/adapter/install/adapter.yaml → install/adapter/install/adapter.yaml

@@ -34,7 +34,7 @@ spec:
     spec:
       containers:
         - name: kuda-mixer-adapter
-          image: gcr.io/kuda-project/kuda-mixer-adapter
+          image: gcr.io/kuda-cloud/kuda-mixer-adapter
           imagePullPolicy: Always
           ports:
             - containerPort: 44225

images/adapter/install/kuda_cfg.yaml → install/adapter/install/kuda_cfg.yaml

@@ -59,7 +59,8 @@ metadata:
 spec:
   # match: match(destination.service.host, "*.svc.cluster.local") && !match(destination.service.host, "*.kuda.svc.cluster.local") && match(context.reporter.uid, "kubernetes://istio-ingressgateway-*")
   # match: destination.service.namespace != "kuda" && destination.service.namespace != "istio-system" && destination.service.namespace != "knative-serving" && destination.service.namespace != "kube-system" && destination.service.namespace != "cert-manager"
-  match: source.labels["istio"] == "ingressgateway" && destination.service.namespace != "kuda"
+  # match: source.labels["istio"] == "ingressgateway" && destination.service.namespace != "kuda"
+  match: match(source.labels["istio"], "nothing")
   actions:
     - handler: kuda-handler.istio-system
       instances:

images/adapter/install/skaffold.yaml → install/adapter/install/skaffold.yaml

@@ -2,7 +2,7 @@ apiVersion: skaffold/v1
 kind: Config
 build:
   artifacts:
-  - image: gcr.io/kuda-project/kuda-mixer-adapter
+  - image: gcr.io/kuda-cloud/kuda-mixer-adapter
     docker:
       dockerfile: ./Dockerfile
 deploy:

images/deployer/Dockerfile → install/api/Dockerfile

@@ -13,7 +13,7 @@ WORKDIR /go/src/github.com/cyrildiagne/kuda
 RUN go mod download
 
 COPY pkg ./pkg
-WORKDIR /go/src/github.com/cyrildiagne/kuda/pkg/deployer
+WORKDIR /go/src/github.com/cyrildiagne/kuda/pkg/deploy
 RUN GO111MODULE=on CGO_ENABLED=0 GOOS=linux go build -installsuffix cgo -o deployer .
 
 #
@@ -50,7 +50,7 @@ RUN chmod +x /usr/local/bin/kubectl
 COPY --from=builder /tmp/skaffold /usr/local/bin/skaffold
 RUN chmod +x /usr/local/bin/skaffold
 
-COPY --from=builder /go/src/github.com/cyrildiagne/kuda/pkg/deployer/deployer /root/
+COPY --from=builder /go/src/github.com/cyrildiagne/kuda/pkg/deploy/deployer /root/
 RUN chmod +x /root/deployer
 
 # Launch the app on port 80.

images/deployer/README.md → install/api/README.md

@@ -2,7 +2,7 @@
 
 ```bash
 docker build \
-  -t gcr.io/kuda-project/deployer \
+  -t gcr.io/kuda-cloud/deployer \
   -f images/deployer/Dockerfile \
   .
 ```
@@ -16,7 +16,7 @@ docker run --rm \
   -v `dirname $GOOGLE_APPLICATION_CREDENTIALS`:/credentials \
   -e PORT=80 \
   -p 8080:80 \
-  gcr.io/kuda-project/deployer
+  gcr.io/kuda-cloud/deployer
 ```
 
 ## Deploy