feat: experimental k9s inspired TUI by n9te9 · Pull Request #1533 · kagent-dev/kagent

n9te9 · 2026-03-23T05:18:57Z

Context

As discussed in #1494, we are revamping the CLI TUI to provide a more intuitive, k9s-inspired experience for agent management.

Since the existing TUI relies heavily on a session-first architecture, modifying it directly carries a high risk of breaking current functionality. To ensure a safe and iterative development process, this PR lays the foundation for the new TUI by introducing a clean-slate charmbracelet/bubbletea architecture behind a hidden experimental-tui command.

This initial PR focuses strictly on establishing the UI framework, state routing, and layout components using mock data, serving as a Proof of Concept (PoC) for the new UX paradigm before we wire it up to the Kubernetes client.

Changes

CLI Entrypoint: Added a hidden experimental-tui command to cmd/kagent/main.go to safely test the new TUI.

New Architecture: Created the internal/experimental_tui package to completely isolate the new implementation from the existing TUI codebase.

State Router: Implemented a stack-based navigation router (screen/root.go) capable of seamlessly pushing and popping different screen models.

Domain Models: Added Agent and ModelProvider structs/interfaces in domain/ to standardize the data structures required for the views.

UI Components: Built reusable lipgloss components in component/, including a responsive data table, status badges, and a context usage gauge.

Agent List Screen: Implemented the initial agent_list.go screen featuring vim-like keybindings (j/k to navigate, q to quit), currently populated with a DemoProvider to demonstrate the layout and interaction mechanics.

…1525) Fixes kagent-dev#1524 --------- Signed-off-by: Jesus Munoz <jesus.munoz@solo.io> Signed-off-by: n9te9 <ai.kanaria.ai@gmail.com>

## Summary - **PyJWT**: `>=2.8.0` → `>=2.12.0` — fixes CVE-2026-32597 (accepts unknown `crit` header extensions) - **pyOpenSSL**: `25.3.0` → `>=26.0.0` — fixes CVE-2026-27459 - **pyasn1**: `0.6.2` → `>=0.6.3` — fixes CVE-2026-30922 - **google.golang.org/grpc**: `v1.79.2` → `v1.79.3` — fixes CVE-2026-33186 (authorization bypass via missing leading slash in :path) - **kagent-tools** helm dep: `0.1.1` → `0.1.2` ## CVE Details | Package | CVE | Severity | Fixed In | |---------|-----|----------|----------| | google.golang.org/grpc | CVE-2026-33186 | CRITICAL | 1.79.3 | | PyJWT | CVE-2026-32597 | HIGH | 2.12.0 | | pyOpenSSL | CVE-2026-27459 | HIGH | 26.0.0 | | pyasn1 | CVE-2026-30922 | HIGH | 0.6.3 | ## Test plan - [ ] `uv sync` in Python workspace resolves without conflicts - [ ] `make -C python test` passes - [ ] `go mod tidy` succeeds with no diff - [ ] Trivy scan passes in CI 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Signed-off-by: n9te9 <ai.kanaria.ai@gmail.com>

Signed-off-by: Dmytro Rashko <dmitriy.rashko@amdocs.com> Signed-off-by: Eitan Yarmush <eitan.yarmush@solo.io> Co-authored-by: Eitan Yarmush <eitan.yarmush@solo.io> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> Signed-off-by: n9te9 <ai.kanaria.ai@gmail.com>

Signed-off-by: n9te9 <ai.kanaria.ai@gmail.com>

Copilot

Pull request overview

This PR introduces an isolated, experimental Bubble Tea–based TUI (k9s-inspired) behind a new CLI entrypoint to prototype a new navigation/layout architecture without impacting the existing session-first TUI.

Changes:

Added internal/experimental_tui with initial router (RootModel), an agent list screen, demo data/provider, and basic UI components (table, status, gauge).
Wired a new experimental-tui Cobra command into the CLI (currently gated to ENVIRONMENT=local).
Introduced initial domain models (Agent, ModelProvider) to standardize data needed by views.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
go/core/cli/internal/experimental_tui/screen/root.go	Adds stack-based screen router for push/pop navigation.
go/core/cli/internal/experimental_tui/screen/agent_list.go	Adds initial agent list screen with keybindings and table rendering.
go/core/cli/internal/experimental_tui/domain/provider.go	Defines `ModelProvider` interface for model metadata/cost/context info.
go/core/cli/internal/experimental_tui/domain/agent.go	Adds `Agent` domain model and helper methods for metrics/labels.
go/core/cli/internal/experimental_tui/demo.go	Provides a demo `ModelProvider` implementation for mock rendering.
go/core/cli/internal/experimental_tui/component/table.go	Implements table rendering for agents (ID/name/status/context gauge).
go/core/cli/internal/experimental_tui/component/status.go	Adds (currently minimal) status badge rendering helper.
go/core/cli/internal/experimental_tui/component/guage.go	Adds context usage gauge renderer (note: filename typo).
go/core/cli/internal/experimental_tui/app.go	Adds `Run()` entrypoint to launch the experimental program with demo data.
go/core/cli/internal/cli/agent/run.go	Adds `RunExperimentalTUI` wrapper called from Cobra.
go/core/cli/cmd/kagent/main.go	Registers the new `experimental-tui` command (currently env-gated).

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-23T05:22:19Z

go/core/cli/internal/experimental_tui/app.go

+func Run() error {
+	m := screen.NewRootModel(screen.NewAgentListModel(demoAgents))
+
+	p := tea.NewProgram(m)
+	if _, err := p.Run(); err != nil {
+		fmt.Printf("Error running program: %v", err)
+		os.Exit(1)
+	}


Run() returns an error but currently prints the error and calls os.Exit(1), which makes it hard for callers (and tests) to handle failures and also makes the return value misleading. Consider returning the p.Run() error instead of exiting, and let the Cobra command decide whether to os.Exit. Also consider using tea.WithAltScreen() (as the existing TUI does) to avoid leaving the UI rendered in the user’s terminal buffer.

Copilot · 2026-03-23T05:22:19Z

go/core/cli/internal/cli/agent/run.go

+	experimentaltui.Run()
+	return nil
+}


RunExperimentalTUI ignores the error returned by experimentaltui.Run() and always returns nil. This will hide failures from the CLI command and makes the error handling in main.go ineffective; return the error from experimentaltui.Run() instead.

Suggested change

experimentaltui.Run()

return nil

}

return experimentaltui.Run()

}

Copilot · 2026-03-23T05:22:20Z

go/core/cli/cmd/kagent/main.go

+	experimentalTuiCmd := &cobra.Command{
+		Use:   "experimental-tui",
+		Short: "Run the experimental TUI (for development/testing)",
+		Long:  `Run the experimental TUI (for development/testing)`,
+		Run: func(cmd *cobra.Command, args []string) {
+			if err := cli.RunExperimentalTUI(cmd.Context(), runCfg); err != nil {
+				fmt.Fprintf(os.Stderr, "Error running experimental TUI: %v\n", err)
+				os.Exit(1)
+			}
+		},
+	}
+
+	commands := []*cobra.Command{installCmd, uninstallCmd, invokeCmd, bugReportCmd, versionCmd, dashboardCmd, getCmd, initCmd, buildCmd, deployCmd, addMcpCmd, runCmd, mcp.NewMCPCmd(), envdoc.NewEnvCmd()}
+	if os.Getenv("ENVIRONMENT") == "local" {
+		commands = append(commands, experimentalTuiCmd)
+	}


The PR description says this command is a "hidden experimental-tui command", but the implementation conditionally registers it only when ENVIRONMENT == local. If the intent is "hidden but available", prefer always registering the command and marking it hidden via Cobra (e.g., Hidden: true) rather than gating it behind an environment variable.

Copilot · 2026-03-23T05:22:20Z

go/core/cli/internal/experimental_tui/component/guage.go

+package component
+
+import (


File name appears to be misspelled (guage vs gauge). Renaming it to gauge.go will improve discoverability/searchability and avoid propagating the typo in future references.

Copilot · 2026-03-23T05:22:20Z

go/core/cli/internal/experimental_tui/screen/root.go

+func (m RootModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
+	switch msg := msg.(type) {
+	case PushScreenMessage:
+		m.stack = append(m.stack, msg.Screen)
+		return m, msg.Screen.Init()
+	case PopScreenMessage:
+		if len(m.stack) > 1 {
+			m.stack = m.stack[:len(m.stack)-1]
+		}
+		return m, nil
+	}


The new stack-based router logic in RootModel.Update (push/pop behavior) is core navigation infrastructure but currently has no unit tests. Since this logic is easy to regress when new screens are added, add focused tests that assert push/pop ordering and that messages are forwarded to the top screen.

Copilot · 2026-03-23T05:22:20Z

go/core/cli/internal/experimental_tui/domain/agent.go

+func (a Agent) ContextUsageRate() float64 {
+	max := a.Provider.MaxContextTokens()
+	if max == 0 {
+		return 0
+	}
+
+	current := float64(a.Provider.InputTokens() + a.Provider.OutputTokens())
+	return current / float64(max)
+}


Several methods (e.g., ContextUsageRate, TokenRatio, Model) dereference a.Provider without a nil check, so the zero-value Agent{} (or any agent built without setting Provider) will panic at runtime. Consider enforcing non-nil provider by construction (e.g., unexport Provider and require NewAgent), or make these methods nil-safe (return 0/"" when Provider is nil).

Copilot · 2026-03-23T05:22:20Z

go/core/cli/internal/experimental_tui/component/table.go

+		statusBadge := agent.Status
+		if w := lipgloss.Width(string(statusBadge)); w > columnWidth.Status {
+			columnWidth.Status = w
+		}
+		gauge := RenderUsageGauge(agent.ContextUsageRate(), colGaugeWidth)
+		if w := lipgloss.Width(gauge); w > columnWidth.Gauge {
+			columnWidth.Gauge = w
+		}
+	}


columnWidth.Gauge is being updated using the rendered gauge string width, but later it is passed back into RenderUsageGauge as the bar width. This causes the gauge to grow (since the rendered string includes brackets/percent), breaking column alignment. Keep a separate fixed bar width (e.g., colGaugeWidth) and compute the gauge column width deterministically, or store bar width in columnWidth.Gauge and never overwrite it with lipgloss.Width(gauge).

feat: experimental agent tui

536ac09

n9te9 requested review from EItanya, ilackarms and yuval-k as code owners March 23, 2026 05:18

Copilot AI review requested due to automatic review settings March 23, 2026 05:18

Copilot started reviewing on behalf of n9te9 March 23, 2026 05:19 View session

jmunozro and others added 4 commits March 23, 2026 14:22

feat: adding configurable ssh-hosts for git-based skills (kagent-dev#…

0a4e80f

…1525) Fixes kagent-dev#1524 --------- Signed-off-by: Jesus Munoz <jesus.munoz@solo.io> Signed-off-by: n9te9 <ai.kanaria.ai@gmail.com>

fix: experimentalTuiCmd is only applied in local environment

17e69f4

Signed-off-by: n9te9 <ai.kanaria.ai@gmail.com>

Copilot AI reviewed Mar 23, 2026

View reviewed changes

n9te9 force-pushed the main branch from a1ebc64 to 17e69f4 Compare March 23, 2026 05:23

n9te9 requested a review from peterj as a code owner March 23, 2026 05:23

n9te9 marked this pull request as draft March 23, 2026 09:13

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: experimental k9s inspired TUI#1533

feat: experimental k9s inspired TUI#1533
n9te9 wants to merge 5 commits intokagent-dev:mainfrom
n9te9:main

n9te9 commented Mar 23, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

n9te9 commented Mar 23, 2026

Context

Changes

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants