Repository navigation

[Feat] Add visual Sokoban mode - #62

Draft

MrDongsls wants to merge 10 commits into

ThinkFlowLab:mainfrom

MrDongsls:feat/sokoban-visual

Draft

[Feat] Add visual Sokoban mode#62
MrDongsls wants to merge 10 commits into
ThinkFlowLab:mainfrom
MrDongsls:feat/sokoban-visual

Commits on Oct 17, 2026
- [Feat] Add text-board Sokoban agent with Valen levels
  
  hsliuustc0106
  authored and
  Mr.Dong
  committed
  c1def35
- An env may swap the rules a model reads, per episode
  Mr.Dong
  committed
  e09ee1c
- Port Valen's Sokoban renderer behind the visual extra
  Mr.Dong
  committed
  f03a788
- Add sokoban --visual: Valen's rendering in Valen's words
  Mr.Dong
  committed
  cb30050
- [Docs] sokoban --visual: what runs today, and what jev cannot
  Mr.Dong
  committed
  b8558e4
- Refuse a text-only backend where the image is the observation
  Mr.Dong
  committed
  961c278
- fix minor docs problem
  Mr.Dong
  committed
  6682ed0
- Pin Cua 4B's supports_images to its modality
  Mr.Dong
  committed
  9047a5b
- Name the image-capable backends; admission is capability, not a result
  Mr.Dong
  committed
  2b19790
- Omnijev is browser-front only in the visual Sokoban docs; s1a run cannot select it
  Mr.Dong
  committed
  d0fb0c9