# niuma_code

> A companion tool for Claude Code — subtract context, multiply attention. Every LLM token lands exactly where it matters.

niuma_code is a Python-based terminal tool that enhances LLM coding tools like Claude Code through context management and attention focusing. Not a replacement — a sharper complement.

## Core Features

- **harness autonomous chat** — Default mode. LLM explores autonomously in a tool loop, thinking and doing simultaneously, finishing tasks in one go
- **loop engineered orchestration** — `/loop <goal>` enters goal-oriented self-loop: build a checklist with verification commands, execute sequentially, verify each step, self-correct on failure (3-strike)
- **TUI full-screen mode** — prompt_toolkit-based full-screen interface, input box fixed at bottom, append new messages anytime during streaming output
- **IDE orchestration mode** — Full-screen code editor with Python-native syntax for orchestration scripts, F5 preview / F6 unattended execution
- **multi-Provider routing** — Configure multiple API endpoints in settings.json, auto-route by model, switch freely within a session
- **multi-context parallelism** — N parallel contexts each maintaining independent history, token counts, and compression state, with LRU eviction
- **code knowledge graph** — Parse code structure via tree-sitter, build symbol definitions, calls, and dependency graphs
- **perception-driven memory** — 10 perception events transcribed into memory in real-time, 4-layer retrieval + Bayesian decay, experience accumulates across sessions
- **outcome reward tracking** — OutcomeTracker computes 0.0~1.0 reward scores per task lifecycle
- **sub-Agent parallel research** — Read-only sub-agents execute tools in parallel within isolated contexts, return summaries

## Quick Start

1. Download [niuma.exe](https://niumacode.cn/download/niuma.exe) (no installation needed, double-click to run)
2. Configure API key: create `~/.niuma/settings.json`
3. Double-click to run `niuma.exe`

## Links

- [Official Site](https://niumacode.cn/)
- [GitHub](https://github.com/zhiguoliu/niuma-code)
- [GitHub Issues](https://github.com/zhiguoliu/niuma-code/issues)
- [Full Feature Docs](/llms-full.txt)