vybe 0.8.3

Vibe coding terminal capture toolkit

Vybe

  ____   ____       ___
  \   \ /   /___ __ | |__   ____
   \   Y   /(   |  || __ \ / __ \
    \     /  \___  || \_\ \  ___/
     \___/   (_____||_____/\____)

Vybe is a vibe coding terminal toolkit: run a command, capture its output, and instantly reuse it — copy to clipboard, wrap in Markdown, search errors, jump to the last failure, or grab tmux scrollback.

Highlights

• vybe run ... streams output live and saves it.
• vybe retry reruns your last vybe run command.
• vybe snipclip copies output only (perfect for issues/ChatGPT).
• vybe snipclip --redact copies output with common secrets masked.
• vybe errors extracts likely error blocks from latest capture.
• vybe export --last --json emits machine-readable context for agents.
• vybe run --tag <name> groups captures by task/session.
• vybe diff shows what changed between your latest two captures.
• vybe share builds a Markdown-ready report for issues/LLM chats.
• vybe doctor prints a fast environment snapshot for debugging setup issues.
• vybe fail jumps back to the most recent failing run.
• Works great on Kali (zsh) and supports tmux scrollback capture.

Install (dev / from source)

python -m venv .venv
source .venv/bin/activate
pip install -e .
vybe --help
python -m vybe --help

Install (recommended on Kali/Ubuntu)

Use pipx to avoid PEP 668 "externally-managed-environment" issues:

sudo apt install pipx
pipx ensurepath
pipx install vybe

From a local checkout:

cd ~/dev/Vybe
pipx install . --force

Check install/update guidance:

vybe self-check
vybe self-check --json

Publishing (maintainers)

Vybe publishes to PyPI from GitHub tags via Trusted Publishing.

One-time setup in PyPI project settings:

• Add a Trusted Publisher for repo homer1013/Vybe
• Workflow file: .github/workflows/publish.yml
• Environment: pypi

Release flow:

git tag vX.Y.Z
git push origin vX.Y.Z

GitHub Actions will build and publish automatically.

Usage

vybe run pytest -q
vybe retry
vybe r pytest -q
vybe run --tag auth pytest -q
vybe rr --cwd
vybe fail
vybe s
vybe sc
vybe sc --redact
vybe ll 5
vybe ls --tag auth
vybe snipclip
vybe errors
vybe export --last --json --snip
vybe diff
vybe share --redact --errors
vybe share --clip
vybe share --json
vybe share --json --errors --redact
vybe doctor
vybe self-check
vybe cfg
vybe init
vybe completion install zsh
vybe md bash
vybe grep "Traceback|ERROR" --i

Speed aliases

• vybe r ... is the same as vybe run ...
• vybe rr is the same as vybe retry
• vybe rr --cwd retries in the original working directory
• vybe rr --tag <name> retries and assigns a tag
• vybe l is the same as vybe last
• vybe s is the same as vybe snip
• vybe sc is the same as vybe snipclip
• vybe o is the same as vybe open
• vybe ll [N] is the same as vybe ls [N]
• Full commands remain the canonical docs and are recommended in scripts/automation

Quick recipes

Fast debug loop:

vybe r pytest -q
vybe errors
vybe share --redact --errors --clip
vybe rr

Tagged task loop:

vybe run --tag auth pytest -q
vybe rr --tag auth
vybe ls --tag auth
vybe diff --tag auth

Agent handoff loop:

vybe export --last --json --snip --redact
vybe share --json --errors --redact
vybe doctor --json

LLM-friendly JSON export

Use this to hand structured context to coding agents.

vybe export --last --json
vybe export --last --json --snip
vybe export --last --json --snip --redact

Tagging and diffs

Use tags to keep one debugging thread grouped:

vybe run --tag auth pytest -q
vybe rr --tag auth
vybe ls --tag auth
vybe tags

See exactly what changed between your latest two captures:

vybe diff
vybe diff --tag auth
vybe diff --full

Share bundles and doctor

Generate a ready-to-paste Markdown bundle:

vybe share
vybe share --redact --errors
vybe share --clip
vybe share --json
vybe share --json --errors --redact

Get quick environment diagnostics:

vybe doctor
vybe doctor --json
vybe self-check
vybe self-check --json
vybe cfg --json
vybe init

CLI stability

Vybe keeps a stable v1 CLI contract for humans, scripts, and agents:

• See docs/CLI_CONTRACT.md
• Machine-readable JSON outputs are additive: existing keys remain, new keys may be added

Agent quickstart (human + LLM loop)

Use this when pairing with ChatGPT/Codex/Claude during debugging.

1. Run and capture

vybe r pytest -q

2. Copy output-only to clipboard for your LLM

vybe sc

3. Apply changes, then retry quickly

vybe rr

4. If you moved directories, retry in the original working dir

vybe rr --cwd

5. Check recent attempts fast

vybe ll 8

Failure-first loop:

vybe fail
vybe s
vybe sc

Tip for agents and scripts:

• Prefer full commands in automation (vybe run, vybe retry) for clarity.
• Use aliases interactively for speed.

Command reference

Run:

vybe --help

Clipboard support

Vybe auto-detects clipboard tools:

• X11: xclip or xsel
• Wayland: wl-copy

Shell completion install

Install directly from the CLI:

vybe completion install zsh
vybe completion install bash
vybe completion install fish

tmux scrollback capture

vybe pane 4000
vybe open

Environment variables

• VYBE_DIR log dir (default ~/.cache/vybe)
• VYBE_STATE state file (default ~/.config/vybe/state.json)
• VYBE_INDEX index file (default ~/.cache/vybe/index.jsonl)
• VYBE_CONFIG config file (default ~/.config/vybe/config.json)
• VYBE_MAX_INDEX max index entries (default 2000)

Shell completions

See completions/:

• bash: completions/vybe.bash
• zsh: completions/_vybe
• fish: completions/vybe.fish

Use vybe completion install <shell> instead of copying files manually.

License

MIT