-
Notifications
You must be signed in to change notification settings - Fork 1
commands
Quick reference. For the full guide — every built-in with its arguments, the web/desktop subset, aliases, and worked examples — see Slash commands.
Type / at the start of a message to run a command (in the REPL or the web
chat). Built-ins cover the session controls — /help, /model, /effort,
/fast, /plan, /auto, /goal, /compact, /reset, /tools, /json,
/notools, /forcetool, /image, /audio, /reload. Run /help for the
full list.
Drop a markdown file at ~/.evi/commands/<name>.md and it becomes /<name>.
The file's contents are sent as your next message, after substitution. Modelled
on Claude Code's custom commands.
---
description: Draft a conventional-commit message
argument-hint: [scope]
model: qwen2.5-coder:14b-instruct-q4_K_M
---
Look at the staged diff and write a commit message for $ARGUMENTS.-
description— shown in/help. -
argument-hint— documents the expected arguments. -
model— surfaced to the UI as the command's preferred model.
-
$ARGUMENTS— everything typed after the command name. -
$1,$2, …$9— positional args (shell-split; quote to group:/review "two words" tail→$1=two words). -
{args}— legacy alias for$ARGUMENTS(still works).
/commit auth with the file above sends:
"Look at the staged diff and write a commit message for auth."
@path/to/file inlines that file's contents (fenced) if it exists — handy for
templates that should always include a checklist or style guide. Tokens that
aren't readable files (emails, @handles) are left untouched.
Subdirectories become :-separated names:
~/.evi/commands/git/commit.md → /git:commit.
!bash execution blocks are intentionally not run — auto-executing shell on
command expansion is too sharp an edge for eVi's permission model. Use a tool
(with its permission prompt) or a Skill for anything that needs to
run code.
Generated from docs/commands.md — edit there, not here.
Start here
Guides
- Architecture
- [[Agent SDK (
evi.sdk)|sdk]] - SDK coverage + borrowable features
- Multi-machine setup
- Self-update design (Phase 29 proposal)
- [[Self-build — developing and building eVi with eVi|self-build]]
- Development notes
- Releasing
- Desktop bundling
- Code signing policy
- Surface parity — CLI ↔ Web ↔ Desktop
- eVi vs Claude Code — feature comparison
- Future integrations — backlog
- Roadmap
Feature deep-dives
- eVi feature guides
- Agents & Orchestration
- Recipes, Routines, Scheduled tasks, Channels
- Evals & LLM-as-judge
- Content Guardrails
- Hooks (tool + lifecycle, command/url)
- MCP (client + serve)
- Memory & Context management
- Observability (OpenTelemetry, stats, crash reports)
- Permissions & Sandbox
- Plugins & Marketplace
- Sessions, Resume, Handoff, Checkpoints
- Skills
- Slash commands
- Structured Outputs & Batch
- Ultracode
- Voice (TTS engines, STT, AutoSpeaker)
- Web & Desktop (settings, multi-user, deep links, updater)