Files
GenericAgent/assets
a60c938e89 refactor(configure): 重构配置向导 — 16 家厂商、启动流程优化、粘贴修复、新增微信平台 (#364)
* feat: add cross-platform GA command entry system

* fix: hide code runner console window

* fix(tgapp): preserve streaming overflow segments (#330)

Keep Telegram streaming overflow messages tied to the primary live message, reuse them across edits, and delete stale overflow replies when streamed content shrinks.

* refactor: rename command/ to ga_cli/ to avoid namespace collision

* feat(tgapp): add /btw command handling (#335)

* fix(launch): skip paste interception when clipboard contains text

PPT/Word copies include both image/png and text/plain items.
Only intercept paste when no text item is present (pure screenshot/image).

* docs: update WeChat group 17 QR code (#338)

* feat(tui): multi-line input with Enter submit, Ctrl+Enter newline, paste-safe

* feat(reflect): add init() for runtime args with fallback to env/config

* feat(tui): add tuiapp_v2.py — refined Textual frontend (#350)

新版 TUI 与 tuiapp.py 并存,共享 GenericAgent 调度/队列/线程协议,
重做视觉与交互:
- 多行 InputArea:≤3 行自适应、>3 行三行滑动窗口、粘贴占位符
- 命令面板(OptionList + 模糊匹配 + Tab 补全)
- /llm、command 选择走 ChoiceList,选完区域收缩
- Markdown HardBreak 渲染 + ANSI 缓存
- 响应式布局(窄屏隐藏侧栏 / 宽屏按比例放大)
- selection-aware 自动滚动(选文本时不抢滚动)
- 新增 /btw、/continue、/export、/restore

不改 cli / hub / README / tuiapp.py 入口。

* docs: add Evaluation — Five Dimensions section (EN & CN) (#339)

* feat: dynamic tool_result maxlen based on parallel tool count

* feat(configure): major rewrite — 16 providers, new startup flow, paste fix

- Expand model providers from 12 to 16 (add Qwen, Stepfun, Tencent,
  Volcengine, Baidu Qianfan, Xiaomi MiMo, Zhipu Coding Plan, GMI)
- Merge duplicate entries (MiniMax/Kimi/Qwen/CRS) with protocol choice
- Add WeChat to messaging platforms (frontends/wechatapp.py existed but
  was not discoverable via configure.py)
- Add coding plan / token plan endpoint selection for applicable providers
- Rework startup flow: no mykey.py → suggest both; has mykey.py → modify/new
- Fix paste race condition in masked_input (batch read + deferred repaint)
- Fix _configure_advanced using provider['type'] instead of cfg['type']
- Fix backup filename containing '/' (invalid on Windows paths)
- Simplify Python version check to recommend 3.12 only
- Add manual model name input option regardless of probe result
- Remove dead code (_read_char, import time)

---------

Co-authored-by: Liang Jiaqing <l.j.q.light@gmail.com>
Co-authored-by: manaskarra <106862611+manaskarra@users.noreply.github.com>
Co-authored-by: YooooEX <me@yooooex.com>
Co-authored-by: AspasZhang <24210980075@m.fudan.edu.cn>
Co-authored-by: nianyucatfish <156216359+nianyucatfish@users.noreply.github.com>
Co-authored-by: Jinyi Han <15224975562@163.com>
2026-05-13 23:10:28 +08:00
..