AIR CLI Integration: air run end to end command

acceptance/experimental/air/run/invalid.yaml

@@ -0,0 +1,5 @@
+experiment_name: bad.name
+command: x
+compute:
+  accelerator_type: GPU_8xH100
+  num_accelerators: 3

acceptance/experimental/air/run/output.txt

@@ -0,0 +1,39 @@
+
+=== dry-run (text)
+>>> [CLI] experimental air run -f valid.yaml --dry-run
+Dry run: configuration for "smoke-test" is valid; not submitting.
+
+=== dry-run (json)
+>>> [CLI] experimental air run -f valid.yaml --dry-run -o json
+{
+  "v": 1,
+  "ts": "[TIMESTAMP]",
+  "data": {
+    "status": "DRY_RUN_OK",
+    "dry_run": true
+  }
+}
+
+=== override not yet supported
+>>> [CLI] experimental air run -f valid.yaml --dry-run --override a=b
+Error: --override is not yet supported
+
+Exit code: 1
+
+=== watch not yet supported
+>>> [CLI] experimental air run -f valid.yaml --dry-run --watch
+Error: --watch is not yet supported
+
+Exit code: 1
+
+=== invalid config is rejected
+>>> [CLI] experimental air run -f invalid.yaml --dry-run
+Error: invalid experiment_name "bad.name": only alphanumeric characters, hyphens (-), and underscores (_) are allowed
+
+Exit code: 1
+
+=== missing --file
+>>> [CLI] experimental air run --dry-run
+Error: required flag(s) "file" not set
+
+Exit code: 1

acceptance/experimental/air/run/script

@@ -0,0 +1,17 @@
+title "dry-run (text)"
+trace $CLI experimental air run -f valid.yaml --dry-run
+
+title "dry-run (json)"
+trace $CLI experimental air run -f valid.yaml --dry-run -o json
+
+title "override not yet supported"
+errcode trace $CLI experimental air run -f valid.yaml --dry-run --override a=b
+
+title "watch not yet supported"
+errcode trace $CLI experimental air run -f valid.yaml --dry-run --watch
+
+title "invalid config is rejected"
+errcode trace $CLI experimental air run -f invalid.yaml --dry-run
+
+title "missing --file"
+errcode trace $CLI experimental air run --dry-run

acceptance/experimental/air/run/test.toml

@@ -0,0 +1,4 @@
+# `air run --dry-run` validates the config locally and makes no workspace calls,
+# so no engine matrix or server stubs are needed.
+[EnvMatrix]
+DATABRICKS_BUNDLE_ENGINE = []

acceptance/experimental/air/run/valid.yaml

@@ -0,0 +1,5 @@
+experiment_name: smoke-test
+command: python train.py
+compute:
+  accelerator_type: GPU_1xH100
+  num_accelerators: 1

acceptance/experimental/air/unimplemented/output.txt

@@ -1,10 +1,4 @@
 
-=== run
->>> [CLI] experimental air run
-Error: `air run` is not implemented yet
-
-Exit code: 1
-
 === logs
 >>> [CLI] experimental air logs 123
 Error: `air logs` is not implemented yet

acceptance/experimental/air/unimplemented/script

@@ -1,8 +1,5 @@
 # Each stub must fail with "not implemented"; errcode records the exit code.
 
-title "run"
-errcode trace $CLI experimental air run
-
 title "logs"
 errcode trace $CLI experimental air logs 123
 

experimental/air/cmd/run.go

@@ -1,10 +1,25 @@
 package aircmd
 
 import (
+	"errors"
+	"fmt"
+	"strconv"
+
 	"github.com/databricks/cli/cmd/root"
+	"github.com/databricks/cli/libs/cmdctx"
+	"github.com/databricks/cli/libs/cmdio"
+	"github.com/databricks/cli/libs/flags"
 	"github.com/spf13/cobra"
 )
 
+// runResult is the JSON payload for `air run`.
+type runResult struct {
+	Status       string `json:"status"`
+	DryRun       bool   `json:"dry_run,omitempty"`
+	RunID        string `json:"run_id,omitempty"`
+	DashboardURL string `json:"dashboard_url,omitempty"`
+}
+
 func newRunCommand() *cobra.Command {
 	var (
 		file           string
@@ -21,16 +36,63 @@ func newRunCommand() *cobra.Command {
 		Long: `Submit a training workload to Databricks serverless GPU compute.
 
 The workload is described by a YAML config file (see --file).`,
-		RunE: func(cmd *cobra.Command, args []string) error {
-			return notImplemented("run")
-		},
 	}
 
 	cmd.Flags().StringVarP(&file, "file", "f", "", "Path to the workload YAML config")
 	cmd.Flags().BoolVar(&watch, "watch", false, "Stream logs until the run completes")
 	cmd.Flags().StringArrayVar(&overrides, "override", nil, "Override a YAML field, e.g. compute.num_accelerators=8 (repeatable)")
 	cmd.Flags().BoolVar(&dryRun, "dry-run", false, "Validate the config without submitting")
 	cmd.Flags().StringVar(&idempotencyKey, "idempotency-key", "", "Return the existing run if this key was already used")
+	_ = cmd.MarkFlagRequired("file")
+
+	// --dry-run only validates the config locally, so it needs no workspace.
+	// Submission requires an authenticated client.
+	cmd.PreRunE = func(cmd *cobra.Command, args []string) error {
+		if dryRun {
+			return nil
+		}
+		return root.MustWorkspaceClient(cmd, args)
+	}
+
+	cmd.RunE = func(cmd *cobra.Command, args []string) error {
+		ctx := cmd.Context()
+
+		// These flags' pipelines are not ported yet; reject rather than silently
+		// ignore them.
+		if len(overrides) > 0 {
+			return errors.New("--override is not yet supported")
+		}
+		if watch {
+			return errors.New("--watch is not yet supported")
+		}
+
+		cfg, err := loadRunConfig(file)
+		if err != nil {
+			return err
+		}
+
+		if dryRun {
+			if root.OutputType(cmd) == flags.OutputText {
+				cmdio.LogString(ctx, fmt.Sprintf("Dry run: configuration for %q is valid; not submitting.", cfg.ExperimentName))
+				return nil
+			}
+			return renderEnvelope(ctx, runResult{Status: "DRY_RUN_OK", DryRun: true})
+		}
+
+		w := cmdctx.WorkspaceClient(ctx)
+		runID, dashboardURL, err := submitWorkload(ctx, w, cfg, file, idempotencyKey)
+		if err != nil {
+			return err
+		}
+
+		runIDStr := strconv.FormatInt(runID, 10)
+		if root.OutputType(cmd) == flags.OutputText {
+			cmdio.LogString(ctx, "Submitted run "+runIDStr)
+			cmdio.LogString(ctx, "View at: "+dashboardURL)
+			return nil
+		}
+		return renderEnvelope(ctx, runResult{Status: "SUBMITTED", RunID: runIDStr, DashboardURL: dashboardURL})
+	}
 
 	return cmd
 }

experimental/air/cmd/runconfig_launch.go

@@ -0,0 +1,62 @@
+package aircmd
+
+// This file flattens the validated runConfig schema into the derived values the
+// launch path consumes, replacing the Python CLI's _convert_to_run_config step.
+// There is no separate internal config type: handle_run reads runConfig directly,
+// using these accessors for the values that need computing rather than a plain
+// field read.
+
+const defaultMaxRetries = 3
+
+// timeoutSeconds converts timeout_minutes to seconds. Zero means the user set no
+// timeout and the backend default applies.
+func (c *runConfig) timeoutSeconds() int {
+	if c.TimeoutMinutes == nil {
+		return 0
+	}
+	return *c.TimeoutMinutes * 60
+}
+
+// maxRetries returns the retry count, applying the schema default when unset.
+func (c *runConfig) maxRetries() int {
+	if c.MaxRetries == nil {
+		return defaultMaxRetries
+	}
+	return *c.MaxRetries
+}
+
+// dockerImageURL returns the custom docker image URL, or "" when none is set.
+func (c *runConfig) dockerImageURL() string {
+	if c.Environment != nil && c.Environment.DockerImage != nil {
+		return c.Environment.DockerImage.URL
+	}
+	return ""
+}
+
+// requirementsFile returns the path to a requirements file when
+// environment.dependencies is a string, and whether it was set.
+func (c *runConfig) requirementsFile() (string, bool) {
+	if c.Environment == nil || !c.Environment.Dependencies.set || c.Environment.Dependencies.isList {
+		return "", false
+	}
+	return c.Environment.Dependencies.path, true
+}
+
+// inlineDependencies returns the inline package list when
+// environment.dependencies is a list, and whether it was set.
+func (c *runConfig) inlineDependencies() ([]string, bool) {
+	if c.Environment == nil || !c.Environment.Dependencies.set || !c.Environment.Dependencies.isList {
+		return nil, false
+	}
+	return c.Environment.Dependencies.list, true
+}
+
+// runtimeVersion returns the client image version from environment.version when
+// set. For a requirements-file dependency set, the version lives in that file and
+// is resolved at launch, not here.
+func (c *runConfig) runtimeVersion() (string, bool) {
+	if c.Environment == nil || !c.Environment.Version.set {
+		return "", false
+	}
+	return c.Environment.Version.raw, true
+}

experimental/air/cmd/runconfig_launch_test.go

@@ -0,0 +1,80 @@
+package aircmd
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestRunConfigTimeoutSeconds(t *testing.T) {
+	c := &runConfig{}
+	assert.Equal(t, 0, c.timeoutSeconds())
+
+	c.TimeoutMinutes = new(2)
+	assert.Equal(t, 120, c.timeoutSeconds())
+}
+
+func TestRunConfigMaxRetries(t *testing.T) {
+	c := &runConfig{}
+	assert.Equal(t, defaultMaxRetries, c.maxRetries())
+
+	c.MaxRetries = new(0)
+	assert.Equal(t, 0, c.maxRetries())
+
+	c.MaxRetries = new(7)
+	assert.Equal(t, 7, c.maxRetries())
+}
+
+func TestRunConfigDockerImageURL(t *testing.T) {
+	c := &runConfig{}
+	assert.Empty(t, c.dockerImageURL())
+
+	c.Environment = &environmentConfig{}
+	assert.Empty(t, c.dockerImageURL())
+
+	c.Environment.DockerImage = &dockerImageConfig{URL: "org/repo:tag"}
+	assert.Equal(t, "org/repo:tag", c.dockerImageURL())
+}
+
+func TestRunConfigDependencies(t *testing.T) {
+	t.Run("unset", func(t *testing.T) {
+		c := &runConfig{}
+		_, ok := c.requirementsFile()
+		assert.False(t, ok)
+		_, ok = c.inlineDependencies()
+		assert.False(t, ok)
+	})
+
+	t.Run("file path", func(t *testing.T) {
+		c := &runConfig{Environment: &environmentConfig{
+			Dependencies: dependencies{set: true, isList: false, path: "req.yaml"},
+		}}
+		path, ok := c.requirementsFile()
+		assert.True(t, ok)
+		assert.Equal(t, "req.yaml", path)
+		_, ok = c.inlineDependencies()
+		assert.False(t, ok)
+	})
+
+	t.Run("inline list", func(t *testing.T) {
+		c := &runConfig{Environment: &environmentConfig{
+			Dependencies: dependencies{set: true, isList: true, list: []string{"torch", "numpy"}},
+		}}
+		list, ok := c.inlineDependencies()
+		assert.True(t, ok)
+		assert.Equal(t, []string{"torch", "numpy"}, list)
+		_, ok = c.requirementsFile()
+		assert.False(t, ok)
+	})
+}
+
+func TestRunConfigRuntimeVersion(t *testing.T) {
+	c := &runConfig{}
+	_, ok := c.runtimeVersion()
+	assert.False(t, ok)
+
+	c.Environment = &environmentConfig{Version: stringOrInt{set: true, raw: "5"}}
+	v, ok := c.runtimeVersion()
+	assert.True(t, ok)
+	assert.Equal(t, "5", v)
+}