Skip to content

Releases: ZedingZhang/ya-agent

Release list

Ya v0.1.6

Ya v0.1.6 Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 27 Jul 06:49
v0.1.6
99be44e

Markdown rendering resilience

  • Terminal rendering now recovers when a model leaves a fenced code block
    unclosed before a subsequent Markdown heading, so later headings, lists, and
    emphasis do not appear as raw Markdown.

Relevant local memory

  • Ya now ranks approved memory for each task instead of always injecting the
    oldest three cards.
  • Ranking is local and dependency-free: exact phrases and meaningful English
    keywords rank before Chinese character n-gram overlap; ties prefer newer
    cards, and low-scoring cards stay out of the model context.
  • ya ask --show-memory displays the cards and scores selected for a task.
    It is opt-in because it can print personal memory text.

Faster, more reliable answers

  • Interactive single-agent answers now stream line-by-line with readable
    Markdown terminal rendering.
  • ya ask --web auto|on|off selects an intelligent default, requires web
    search, or disables web access for a faster direct answer.
  • ya ask --stream auto|off controls interactive streaming. ToA, web research,
    pipes, and Markdown output remain buffered for reliable tool handling.
  • DeepSeek and web-search requests retry transient failures before output is
    shown, without duplicating a partial answer.

Standalone executables

Ya now ships as standalone executables for macOS Apple Silicon, macOS Intel,
Windows x64, and Linux x64. Download the matching asset and run it directly;
Python, pip, and a PATH change are not required.

The macOS and Windows files are currently unsigned. Verify checksums.txt
before overriding a system warning. ya auth deepseek remains macOS-only;
Linux and Windows use DEEPSEEK_API_KEY.

Python packages

The universal wheel and source distribution remain available for Python users
and contributors.

中文说明

独立可执行文件

Ya 现提供 macOS Apple Silicon、macOS Intel、Windows x64 和 Linux x64 的独立
可执行文件。下载对应附件后即可直接运行，无需安装 Python、pip 或修改 PATH。

macOS 和 Windows 文件当前未签名。在绕过系统提示前，请先校验 checksums.txt。
ya auth deepseek 仍仅支持 macOS；Linux 和 Windows 请使用 DEEPSEEK_API_KEY。

更快、更可靠的回答

  • 交互式终端中的单 Agent 回答现在会逐行流式输出，并保持易读的 Markdown 渲染。
  • ya ask --web auto|on|off 分别提供智能默认策略、强制网页检索和关闭网页访问以加快直接回答。
  • ya ask --stream auto|off 控制交互式流式输出。ToA、网页检索、管道和 Markdown 输出会保持缓冲，
    以保证工具处理可靠。
  • DeepSeek 和网页检索在输出前遇到瞬时故障时会重试，且不会重复已经显示的部分回答。

相关本地记忆

  • Ya 现在会为每个任务排序已批准记忆，不再固定注入最早的三张卡片。
  • 排序完全在本地、零依赖完成：精确短语和有意义的英文关键词优先于中文字符 n-gram 重合；
    同分时优先较新的卡片，低分卡片不会进入模型上下文。
  • ya ask --show-memory 可显示本次任务实际选用的卡片和分数。该选项需要显式传入，因为它可能
    打印个人记忆文本。

Markdown 渲染恢复能力

  • 当模型漏掉围栏代码块的结束标记、随后输出 Markdown 标题时，终端渲染现在会自动恢复，后续的
    标题、列表和强调文本不会再原样显示 Markdown 标记。

Python 包

仍保留通用 wheel 与源码包，供 Python 用户和贡献者使用。

Assets 9
Loading

Ya v0.1.5

Ya v0.1.5

Choose a tag to compare

View all tags
@github-actions github-actions released this 27 Jul 06:33
v0.1.5
2a0e050

Relevant local memory

  • Ya now ranks approved memory for each task instead of always injecting the
    oldest three cards.
  • Ranking is local and dependency-free: exact phrases and meaningful English
    keywords rank before Chinese character n-gram overlap; ties prefer newer
    cards, and low-scoring cards stay out of the model context.
  • ya ask --show-memory displays the cards and scores selected for a task.
    It is opt-in because it can print personal memory text.

Faster, more reliable answers

  • Interactive single-agent answers now stream line-by-line with readable
    Markdown terminal rendering.
  • ya ask --web auto|on|off selects an intelligent default, requires web
    search, or disables web access for a faster direct answer.
  • ya ask --stream auto|off controls interactive streaming. ToA, web research,
    pipes, and Markdown output remain buffered for reliable tool handling.
  • DeepSeek and web-search requests retry transient failures before output is
    shown, without duplicating a partial answer.

Standalone executables

Ya now ships as standalone executables for macOS Apple Silicon, macOS Intel,
Windows x64, and Linux x64. Download the matching asset and run it directly;
Python, pip, and a PATH change are not required.

The macOS and Windows files are currently unsigned. Verify checksums.txt
before overriding a system warning. ya auth deepseek remains macOS-only;
Linux and Windows use DEEPSEEK_API_KEY.

Python packages

The universal wheel and source distribution remain available for Python users
and contributors.

中文说明

独立可执行文件

Ya 现提供 macOS Apple Silicon、macOS Intel、Windows x64 和 Linux x64 的独立
可执行文件。下载对应附件后即可直接运行，无需安装 Python、pip 或修改 PATH。

macOS 和 Windows 文件当前未签名。在绕过系统提示前，请先校验 checksums.txt。
ya auth deepseek 仍仅支持 macOS；Linux 和 Windows 请使用 DEEPSEEK_API_KEY。

更快、更可靠的回答

  • 交互式终端中的单 Agent 回答现在会逐行流式输出，并保持易读的 Markdown 渲染。
  • ya ask --web auto|on|off 分别提供智能默认策略、强制网页检索和关闭网页访问以加快直接回答。
  • ya ask --stream auto|off 控制交互式流式输出。ToA、网页检索、管道和 Markdown 输出会保持缓冲，
    以保证工具处理可靠。
  • DeepSeek 和网页检索在输出前遇到瞬时故障时会重试，且不会重复已经显示的部分回答。

相关本地记忆

  • Ya 现在会为每个任务排序已批准记忆，不再固定注入最早的三张卡片。
  • 排序完全在本地、零依赖完成：精确短语和有意义的英文关键词优先于中文字符 n-gram 重合；
    同分时优先较新的卡片，低分卡片不会进入模型上下文。
  • ya ask --show-memory 可显示本次任务实际选用的卡片和分数。该选项需要显式传入，因为它可能
    打印个人记忆文本。

Python 包

仍保留通用 wheel 与源码包，供 Python 用户和贡献者使用。

Loading

Ya v0.1.4

Ya v0.1.4

Choose a tag to compare

View all tags
@github-actions github-actions released this 27 Jul 06:13
v0.1.4
49ddd53

Faster, more reliable answers

  • Interactive single-agent answers now stream line-by-line with readable
    Markdown terminal rendering.
  • ya ask --web auto|on|off selects an intelligent default, requires web
    search, or disables web access for a faster direct answer.
  • ya ask --stream auto|off controls interactive streaming. ToA, web research,
    pipes, and Markdown output remain buffered for reliable tool handling.
  • DeepSeek and web-search requests retry transient failures before output is
    shown, without duplicating a partial answer.

Standalone executables

Ya now ships as standalone executables for macOS Apple Silicon, macOS Intel,
Windows x64, and Linux x64. Download the matching asset and run it directly;
Python, pip, and a PATH change are not required.

The macOS and Windows files are currently unsigned. Verify checksums.txt
before overriding a system warning. ya auth deepseek remains macOS-only;
Linux and Windows use DEEPSEEK_API_KEY.

Python packages

The universal wheel and source distribution remain available for Python users
and contributors.

中文说明

独立可执行文件

Ya 现提供 macOS Apple Silicon、macOS Intel、Windows x64 和 Linux x64 的独立
可执行文件。下载对应附件后即可直接运行，无需安装 Python、pip 或修改 PATH。

macOS 和 Windows 文件当前未签名。在绕过系统提示前，请先校验 checksums.txt。
ya auth deepseek 仍仅支持 macOS；Linux 和 Windows 请使用 DEEPSEEK_API_KEY。

更快、更可靠的回答

  • 交互式终端中的单 Agent 回答现在会逐行流式输出，并保持易读的 Markdown 渲染。
  • ya ask --web auto|on|off 分别提供智能默认策略、强制网页检索和关闭网页访问以加快直接回答。
  • ya ask --stream auto|off 控制交互式流式输出。ToA、网页检索、管道和 Markdown 输出会保持缓冲，
    以保证工具处理可靠。
  • DeepSeek 和网页检索在输出前遇到瞬时故障时会重试，且不会重复已经显示的部分回答。

Python 包

仍保留通用 wheel 与源码包，供 Python 用户和贡献者使用。

Loading

Ya v0.1.3

Ya v0.1.3

Choose a tag to compare

View all tags
@github-actions github-actions released this 27 Jul 02:49
v0.1.3
5b44a5a

Standalone executables

Ya now ships as standalone executables for macOS Apple Silicon, macOS Intel,
Windows x64, and Linux x64. Download the matching asset and run it directly;
Python, pip, and a PATH change are not required.

The macOS and Windows files are currently unsigned. Verify checksums.txt
before overriding a system warning. ya auth deepseek remains macOS-only;
Linux and Windows use DEEPSEEK_API_KEY.

Python packages

The universal wheel and source distribution remain available for Python users
and contributors.

中文说明

独立可执行文件

Ya 现提供 macOS Apple Silicon、macOS Intel、Windows x64 和 Linux x64 的独立
可执行文件。下载对应附件后即可直接运行，无需安装 Python、pip 或修改 PATH。

macOS 和 Windows 文件当前未签名。在绕过系统提示前，请先校验 checksums.txt。
ya auth deepseek 仍仅支持 macOS；Linux 和 Windows 请使用 DEEPSEEK_API_KEY。

Python 包

仍保留通用 wheel 与源码包，供 Python 用户和贡献者使用。

Loading

Ya v0.1.2

Ya v0.1.2

Choose a tag to compare

View all tags
@github-actions github-actions released this 27 Jul 01:55
v0.1.2
d2ed90c

Standalone executables

Ya now ships as standalone executables for macOS Apple Silicon, macOS Intel,
Windows x64, and Linux x64. Download the matching asset and run it directly;
Python, pip, and a PATH change are not required.

The macOS and Windows files are currently unsigned. Verify checksums.txt
before overriding a system warning. ya auth deepseek remains macOS-only;
Linux and Windows use DEEPSEEK_API_KEY.

Python packages

The universal wheel and source distribution remain available for Python users
and contributors.

中文说明

独立可执行文件

Ya 现提供 macOS Apple Silicon、macOS Intel、Windows x64 和 Linux x64 的独立
可执行文件。下载对应附件后即可直接运行，无需安装 Python、pip 或修改 PATH。

macOS 和 Windows 文件当前未签名。在绕过系统提示前，请先校验 checksums.txt。
ya auth deepseek 仍仅支持 macOS；Linux 和 Windows 请使用 DEEPSEEK_API_KEY。

Python 包

仍保留通用 wheel 与源码包，供 Python 用户和贡献者使用。

Loading

Ya v0.1.1

Ya v0.1.1

Choose a tag to compare

View all tags
@ZedingZhang ZedingZhang released this 27 Jul 00:59
v0.1.1
6cc57f9

Markdown terminal rendering

  • Interactive terminals now render Ya's common Markdown output, including headings, emphasis, lists, block quotes, code blocks, links, and simple tables.
  • --format auto is the default: render in a TTY and preserve raw Markdown when output is redirected or piped.
  • Use --format terminal to force terminal formatting or --format markdown to preserve source Markdown.
  • Set NO_COLOR=1 to disable ANSI styles.

中文说明

Markdown 终端渲染

  • 交互式终端现可渲染 Ya 常见的 Markdown 输出，包括标题、强调、列表、引用、代码块、链接和简单表格。
  • --format auto 为默认行为：TTY 中渲染，重定向或通过管道输出时保留原始 Markdown。
  • 使用 --format terminal 强制终端排版，使用 --format markdown 保留源 Markdown。
  • 设置 NO_COLOR=1 可关闭 ANSI 样式。

Assets

  • ya_agent-0.1.1-py3-none-any.whl: universal Python wheel
  • ya_agent-0.1.1.tar.gz: source distribution
Loading

Ya v0.1.0

Ya v0.1.0

Choose a tag to compare

View all tags
@ZedingZhang ZedingZhang released this 26 Jul 13:48
v0.1.0
4924328

First release

Ya v0.1.0 is a consent-first personal research CLI agent for Python 3.9+.

Platform support

  • macOS: ya auth deepseek stores the API key in the macOS Keychain.
  • Linux: Set DEEPSEEK_API_KEY in the shell before running Ya.
  • Windows (PowerShell): Set $env:DEEPSEEK_API_KEY before running Ya.

The attached py3-none-any wheel works across supported operating systems. Linux and Windows use environment-based credentials; the Keychain authentication command is macOS-only.

Assets

  • ya_agent-0.1.0-py3-none-any.whl: universal Python wheel
  • ya_agent-0.1.0.tar.gz: source distribution

首个版本

Ya v0.1.0 是一个以用户同意为先的个人研究型 CLI Agent，需要 Python 3.9 或更高版本。

平台支持

  • macOS：ya auth deepseek 会将 API Key 保存到 macOS 钥匙串。
  • **Linux：**运行 Ya 前，请在 Shell 中设置 DEEPSEEK_API_KEY。
  • **Windows（PowerShell）：**运行 Ya 前，请设置 $env:DEEPSEEK_API_KEY。

随附的 py3-none-any wheel 可在支持的操作系统上使用。Linux 和 Windows 使用环境变量认证；钥匙串认证命令仅适用于 macOS。

附件

  • ya_agent-0.1.0-py3-none-any.whl：跨平台通用 Python wheel
  • ya_agent-0.1.0.tar.gz：源码发行包
Loading