yume

# yume

<div align="center">
  <img src="assets/yume.png" alt="yume logo" width="120">
  <h3>claude code, unchained</h3>
  <p>orchestration flow. memory system. background agents.<br>4 built-in agents. plugins & skills.</p>
  <p>
    <a href="https://aofp.github.io/yume">Download</a>
  </p>
</div>

---

## why yume?

Claude Code is transformative for development. Terminal interfaces hold it back.

### terminal limitations

- **flickering** - 700+ upvoted github issue, fundamentally unfixed
- **performance degradation** - multi-second delays in extended sessions
- **terminal crashes** - IDE instability after 10-20 minutes of heavy use
- **hidden rate limits** - requires manual `/usage` checks
- **fragile sessions** - complete loss on crash, no recovery mechanism

### yume's approach

- ✅ **native rendering** - hardware-accelerated UI, zero terminal artifacts
- ✅ **consistent <50ms latency** - responsive input across unlimited session length
- ✅ **standalone architecture** - immune to IDE/terminal crashes
- ✅ **persistent rate limit display** - 5-hour and 7-day metrics always visible
- ✅ **automatic session recovery** - 5-minute checkpoint intervals with 24-hour retention

---

## features

### orchestration flow (unique to yume)
- **automatic task decomposition** - understand → decompose → act → verify workflow
- **complex task handling** - 3+ step tasks get architect planning, implementer execution, guardian verification
- **trivial task passthrough** - simple tasks proceed directly without overhead
- **no commands needed** - baked into every session via --append-system-prompt

### agent system
- **4 built-in agents** - synced to ~/.claude/agents/yume-*.md
  - **architect** - plans, decomposes tasks, identifies risks, todowrite tracking
  - **explorer** - read-only codebase search, glob/grep/read, never modifies
  - **implementer** - focused code changes, edit/write, minimal diffs
  - **guardian** - reviews for bugs, security, performance after changes
- **4 concurrent background agents** - async execution with git branch isolation (yume-async-{type}-{id})
- **no timeout** - agents run until completion, output to ~/.yume/agent-output/
- **streaming isolation** - background agents don't interfere with main session
- **merge conflict detection** - pre-merge checks before integrating agent work

### memory v2 system
- **per-project markdown** - organized files in ~/.yume/memory/ (global + project-specific)
- **5 importance levels** - ephemeral (1d), low (7d), normal (30d), high (90d), permanent
- **auto-learning** - extracts patterns from error/fix conversations, architecture decisions
- **ttl-based expiration** - automatic pruning of outdated entries
- **mcp integration** - custom yume-mcp-memory.cjs server for claude access

### multi-provider support
- **6 models, 3 providers** - via yume-cli shim
  - claude sonnet 4.5, opus 4.5
  - gemini 2.5 pro, flash
  - gpt-5.2 codex, codex mini
- **protocol normalization** - stream-json output for all providers
- **same interface** - switch providers without changing workflow

### core features
- **persistent rate limit visibility** - 5h and 7d limits always visible, no /usage needed
- **zero visual artifacts** - native rendering eliminates terminal flickering
- **real-time thinking display** - all thinking steps stream live as they happen
- **keyboard-first** - 32+ shortcuts, full mouse-free navigation
- **<50ms response time** - instant UI regardless of session length
- **standalone reliability** - independent process, immune to IDE crashes
- **auto-compaction** - variable threshold (default 75%), user configurable or disable

### plugin & skills system
- **complete extensibility** - commands, agents, hooks, skills, MCP servers
- **skills auto-injection** - context based on file extensions, keywords, regex
- **plugin directory** - install/enable/disable per component
- **bundled yume plugin** - default commands and agents

### additional features
- **crash recovery** - 30s auto-save, 5-min snapshots, 24hr retention, complete session restore
- **history rollback** - restore points per message, conflict detection, undo edits
- **mid-stream context** - real-time token updates during streaming
- **analytics dashboard** - usage by project/model/date, cost tracking, export
- **voice dictation** - F5 to toggle, native speech-to-text
- **@ mention system** - @r recent, @m modified, folder navigation

### customization
- **12 themes, infinite combinations** - accents, backgrounds, foregrounds fully customizable
- **OLED optimization** - pure black (#000000) default
- **100% CLI compatibility** - subagents, MCP, hooks, skills, claude.md, /commands

---

## tech stack

- **tauri 2.9** - rust backend, native speed
- **react 19** - modern ui with virtualized rendering
- **sqlite + wal** - sessions, messages, analytics persistence
- **socket.io** - real-time streaming with mid-stream context updates
- **compiled binaries** - no node.js dependency for end users (@yao-pkg/pkg)
- **

... (truncated)