vibectl 0.8.6

A vibes-based alternative to kubectl

vibectl

A vibes‑first alternative to kubectl — because clusters deserve good vibes too.

---

✨ Why vibectl?

Managing Kubernetes shouldn't feel like editing an INI file through a periscope. vibectl wraps plain‑English intent, a dash of emoji, and an LLM planner around ordinary kubectl, giving you memory‑aware, conversational control of any cluster. Keep using every manifest, context, and kube‑config you already have – just add vibes.

---

🚀 Feature Highlights

Category	What it does
Conversational Commands	vibectl get vibe pods → natural‑language queries for any verb.
LLM Planner	Structured JSON contracts ensure predictable plans that you confirm.
Memory System	Context persists between invocations (vibectl memory ...).
Semi‑Autonomous Loops	vibectl semiauto iteratively proposes & executes safe changes.
Full Autonomy	vibectl vibe plans, confirms, executes, summarises, updates memory.
Rich TUI	Live watch, logs ‑f, port‑forward with pause/filter/save key‑bindings.
Traffic Proxy	Optional middle‑proxy shows per‑session throughput & errors.
Chaos / Demo Tooling	Drop‑in sandbox demos for CTFs, Kafka tuning, Chaos‑Monkey battles.
Intelligent Apply	vibectl apply vibe autocorrects or generates manifests before applying.

New in 0.8.x vibectl patch for intelligent resource patching, vibectl edit for smart editing workflows, vibectl diff for live comparisons, and the new intelligent apply workflow.

---

🛠 Installation

Via pip (any OS)

pip install vibectl            # core CLI
pip install llm-anthropic      # Claude (default)
# or: pip install llm-openai   # OpenAI models
# or: pip install llm-ollama   # local Ollama models

Via Nix flakes (NixOS & friends)

git clone https://github.com/othercriteria/vibectl.git
cd vibectl
flake develop        # drops you into a fully wired shell

---

🔑 Configure an LLM key

export ANTHROPIC_API_KEY=sk-ant-...        # quickest
vibectl config set model claude-3.7-sonnet

More options? See docs/MODEL_KEYS.md.

---

⏱ 60‑Second Tour

1 – Ask for a vibe check

❯ vibectl get vibe pods
🔄 Consulting claude-3.7-sonnet for a plan…
🔄 Running: kubectl get pods -n sandbox
✨ Vibe check:
🚀 3 nginx‑demo pods in deployment 🌟

2 – Persist a fact

vibectl memory set "We're working in 'sandbox' namespace."

3 – Iterate with semiauto

vibectl semiauto "set up a demo redis"
# …you confirm each command (`y / n / a / b / m / e`)

4 – One‑liner natural language

vibectl scale vibe "nginx‑demo down to 1 replica"

5 – Smart patching with natural language

vibectl patch vibe "set nginx‑demo image to nginx:1.21"

6 – Rich port‑forward session

vibectl port-forward vibe "nginx-demo-service to 8090"
# Keybindings:  P‑pause  W‑wrap  F‑filter  S‑save  E‑exit

7 – Live event watch

vibectl events -n sandbox --watch

---

📚 Command Cheatsheet

Mode	When to use	Example
Just pass‑through	You want raw kubectl	vibectl just get pods -A
get / describe / scale / delete / logs / events / patch / edit	Familiar verbs, plus AI summary	vibectl get pods
vibe	Full autonomous planner	vibectl vibe "deploy redis with persistence"
semiauto	Step‑wise interactive planner	vibectl semiauto
auto	Non‑interactive loops (used by agents)	vibectl auto "keep latency <50 ms"

---

🧠 Memory Commands

vibectl memory show          # view
vibectl memory set "..."        # replace
vibectl memory set --edit    # $EDITOR
vibectl memory disable|enable
vibectl memory clear

The planner sees memory every turn, so write facts, goals, and preferences there.

---

🎮 Interactive UI Details

Watch / Logs / Events

• Live table updates with elapsed time & line count
• P pause display W wrap F regex filter S save buffer E exit
• Upon exit vibectl prints a metrics table plus an LLM‑generated summary.

Port‑Forward Enhancements

Configure a proxy port range once:

vibectl config set intermediate_port_range 10000-11000

Every port-forward thereafter shows bytes ⇅, connection duration, errors, and a colourful recap.

---

⚙️ Key Configuration Knobs (vibectl config)

Key	Default	Why you'd change it
model	claude-3.7-sonnet	Switch to gpt-4o, ollama:llama3:latest, etc.
intelligent_edit	true	Disable to use standard kubectl edit behavior.
show_metrics	false	View LLM tokens & latency.
show_raw_output	false	Always print raw kubectl output.
theme	dark	light / system / custom.
live_display_max_lines	20	Default visible buffer for watch/logs.

Full schema in docs/CONFIG.md.

---

📦 Demo Environments

Demo	What it shows	Path
Bootstrap	k3d + Ollama self‑contained playground	examples/k8s-sandbox/bootstrap/
CTF Sandbox	vibectl autonomously solves staged flags	examples/k8s-sandbox/ctf/
Chaos Monkey	Red vs Blue vibectl agents battle	examples/k8s-sandbox/chaos-monkey/
Kafka Throughput	vibectl tunes Kafka via agent loop	examples/k8s-sandbox/kafka-throughput/

Each demo has its own README.md with step‑by‑step instructions.

---

🧪 Development & Testing

flake develop     # or: make install-dev
make check        # ruff + mypy + pytest
make test-fast    # quick subset
make bump-patch   # bump version with changelog guard

Pre‑commit hooks enforce Ruff lint/format; CI targets 100 % coverage (see TESTING.md).

---

🤝 Contributing

Pull requests welcome! Start with an issue or draft PR so we can vibe‑check the idea. All contributors agree to the MIT License.

---

© 2025 Daniel Klein & the vibectl community. Spread good vibes — even to your clusters.