Zum Inhalt springen

Little-Coder Cheat Sheet

Dieser Inhalt ist noch nicht in deiner Sprache verfügbar.


Key Action
Ctrl+Q Toggle plan mode (research → ask → plan flow; indicator below input)
Ctrl+H Toggle shortcuts help panel (curated key reference)
Shift+Tab Cycle thinking level (pi built-in)
/ Recall prompt history (from empty input; persists across sessions)
/ Open command palette
! Enter bash mode
Key Action
Ctrl+P Cycle to next model
Shift+Ctrl+P Cycle to previous model
Ctrl+L Open model selector
Ctrl+T Toggle thinking blocks
Ctrl+O Toggle tool output
Ctrl+S Save model selection
Key Action
Enter Submit input
Shift+Enter Insert newline
Ctrl+J Insert newline (alternative)
Ctrl+C Clear editor / Copy selection
Ctrl+D Exit (when editor empty)
Ctrl+Z Suspend to background (Linux/macOS)
Ctrl+G Open external editor
Ctrl+U Delete to line start
Ctrl+K Delete to line end
Ctrl+Y Yank (paste from kill ring)
Ctrl+W Delete word backward
Ctrl+A Move to line start
Ctrl+E Move to line end
Ctrl+B Move cursor left
Ctrl+F Move cursor right
Alt+Left Move cursor word left
Alt+Right Move cursor word right
Ctrl+V Paste image from clipboard (Linux)
Key Action
Ctrl+N Toggle named session filter
Ctrl+R Rename session
Ctrl+X Clear all models
Ctrl+Backspace Delete session (when query empty)
Ctrl+Left / Alt+Left Fold tree branch / Move up
Ctrl+Right / Alt+Right Unfold tree branch / Move down
Shift+L Edit tree label
Shift+T Toggle tree label timestamps
Ctrl+U Tree filter: user messages only
Ctrl+A Tree filter: show all entries
Alt+Enter Queue follow-up message
Alt+Up Restore queued messages / Move model up
Alt+Down Move model down
Key Action
Esc Interrupt current turn / Cancel plan / Close panel
Flag Action
little-coder Launch interactive agent
-m <model> Specify model (e.g. llamacpp/qwen3.6-35b-a3b)
-e <ext> Load extra extension (file or directory)
--list-models List all registered models
--no-extensions Run pi with no extensions (base mode)
Provider Format Example
llama.cpp llamacpp/<id> llamacpp/qwen3.6-35b-a3b
Ollama ollama/<id> ollama/qwen3.5
LM Studio lmstudio/<id> lmstudio/local-model
Anthropic anthropic/<id> anthropic/claude-haiku-4-5
OpenAI openai/<id> openai/gpt-4o-mini

Plan mode (Ctrl+Q) replaces the normal coding turn with a research flow:

  1. Decompose — request split into 1–4 exploration tasks
  2. Explore — isolated sub-coders run read-only (repo + web), reports returned
  3. Clarify — 1–3 questions with suggested answers + free-text option
  4. Plan — synthesis written as a plan in the chat (no file edits)
  5. Exit — returns to normal coding mode

Esc cancels a plan mid-run.

Variable Purpose
LLAMACPP_API_KEY Required even for local (use noop)
LLAMACPP_BASE_URL Override llama.cpp endpoint (default: http://127.0.0.1:8888/v1)
OLLAMA_API_KEY Required even for local (use noop)
OLLAMA_BASE_URL Override Ollama endpoint (default: http://127.0.0.1:11434/v1)
LMSTUDIO_API_KEY Required even for local (use noop)
LMSTUDIO_BASE_URL Override LM Studio endpoint (default: http://127.0.0.1:1234/v1)
LITTLE_CODER_SUBCODER_CONCURRENCY Parallel sub-coders (default: 2)
LITTLE_CODER_MAX_TURNS Max turns before auto-halt (0 = unlimited)
LITTLE_CODER_EXTRA_EXTENSIONS Path-delimited extra extensions (: on POSIX, ; on Windows)

For the exhaustive list including every pi built-in: type /hotkeys in the chat. Extensions registered via pi.registerShortcut() also appear in the “Extensions” table there.