
这是一个人工智能代理项目;它用于学习。
概览
OpenNova v0.4.3 是一个用 Python 构建的终端 AI 编码代理,围绕文本 TUI 构建。 OpenNova 结合了聚焦的代理运行时和全屏终端工作台: - OpenAI、Anthropic 和 DeepSeek 提供商 - 流式文本 TUI 聊天及上下文、任务和活动面板 - 持久会话,包含简历选择器和完整文字记录回放 - 计划/行动工作流程、TodoWrite、子代理和工作树 - 40 个内置工具,以及技能、可信项目插件、钩子和 MCP - 上下文压缩和分层项目内存 - 三种权限模式、参数规则、秘密编辑、审计日志和沙盒 - 一个无头 Python SDK 用于脚本和服务 传统交互式命令行界面和独立的 opennova tui 命令不包含在当前产品中。运行 opennova 时无需子命令即可打开文本 TUI。命令选项仍可用于设置、自动化和一次性任务。 版本 0.4。
概述(AI 翻译)
OpenNova v0.4.3 是一个用 Python 构建的终端 AI 编码代理,围绕文本 TUI 构建。
英语 |简体中文
快速入门 |教程****API参考
概述
OpenNova 结合了聚焦的代理运行时和全屏终端工作台:
- OpenAI、Anthropic 和 DeepSeek 服务提供商
- 带聊天和上下文、任务和活动面板的流式文本TUI
- 与简历选择器的持续会话及完整转录回放
- 计划/行动工作流、TodoWrite、子代理和工作树
- 40个内置工具,以及技能、可信项目插件、钩子和MCP
- 上下文压缩和分层项目内存
- 三种权限模式、参数规则、秘密编辑、审计日志和沙盒
- 一个无头 Python SDK 用于脚本和服务
遗留的交互式命令行界面和独立命令opennova tui不包含在当前产品中。运行opennova时无需子命令即可打开文本TUI。命令选项仍可用于设置、自动化和一次性任务。
v0.4.3 新增内容
版本 0.4.3 提升了仓库规模编码工作流程的运行效率和可靠性:
本页由 AI 翻译,技术细节请以原文为准。
README
以下为原文 README 节选,完整内容请查看原文。
OpenNova
OpenNova v0.4.3 is a terminal AI coding agent built in Python around a Textual TUI.
English | 简体中文
Quickstart | Tutorial | API Reference
Overview
OpenNova combines a focused agent runtime with a full-screen terminal workbench:
- OpenAI, Anthropic, and DeepSeek providers
- Streaming Textual TUI with chat and Context, Tasks, and Activity panels
- Persistent sessions with a resume picker and complete transcript replay
- Plan/Act workflows, TodoWrite, sub-agents, and worktrees
- 40 built-in tools plus Skills, trusted project plugins, hooks, and MCP
- Context compression and layered project memory
- Three permission modes, parameter rules, secret redaction, audit logs, and sandboxes
- A headless Python SDK for scripts and services
The legacy interactive command-line interface and standalone opennova tui command are not part of the current product. Run opennova with no subcommand to open the Textual TUI. Command options remain available for setup, automation, and one-shot tasks.
What is new in v0.4.3
Version 0.4.3 improves runtime efficiency and reliability for repository-scale coding workflows:
- executes independent read-only tools concurrently while preserving deterministic result order
- protects edits from stale file snapshots and stores oversized tool output as retrievable artifacts
- adds deferred tool discovery, shared gitignore handling, bootstrap profiles, and diagnostics
- expands MCP lifecycle support, checkpoint safety, session forking, and layered memory controls
- improves model routing with token budgets, retries, fallbacks, and circuit breakers
Installation
Requirements: Python 3.11+ and uv.
git clone https://github.com/Wardell-Stephen-CurryII/OpenNova.git
cd OpenNova
uv sync
uv run opennova init
For a globally available command:
uv tool install .
opennova
Configuration
Configuration is merged in this order: defaults, ~/.opennova/config.yaml, project .opennova/config.yaml, then environment variables.
default_provider: deepseek
default_model: deepseek-v4-pro
providers:
openai:
api_key: ${OPENAI_API_KEY}
default_model: gpt-4o
anthropic:
api_key: ${ANTHROPIC_API_KEY}
default_model: claude-sonnet-4
deepseek:
api_key: ${DEEPSEEK_API_KEY}
base_url: https://api.deepseek.com/v1
default_model: deepseek-v4-pro
agent:
max_iterations: 20
compression:
enabled: true
threshold: 0.55
keep_last_pairs: 6
max_tool_result_tokens: 8000
security:
permission_mode: auto # request | auto | full
sandbox_mode: true
allow_network: true
strict_shell_parsing: false
read_only: false
mcp:
enabled: true
servers: []
skills:
enabled: true
dirs: []
exclude: []
API keys may also be supplied through OPENAI_API_KEY, ANTHROPIC_API_KEY, and DEEPSEEK_API_KEY.
Usage
推荐工具
换一个关键词,或者移除筛选条件。
安装
npx skillfish add wardell-stephen-curryii/opennova