fix(paper-review): enable vision mode by default and correct DashScope model name (#135)

- Set `vision=True` as default in __main__.py so PDF review uses vision mode without needing --vision flag
- Update SKILL.md to reflect vision mode is on by default
- Fix DashScope model name from non-existent qwen3.5-plus back to dashscope/qwen3.5-plus (correct litellm prefix format)

Made-with: Cursor

Author: XiaoBoAI

cookbooks/paper_review/__main__.py

@@ -270,7 +270,7 @@ def main(
     no_correctness: bool = False,
     no_criticality: bool = False,
     no_bib: bool = False,
-    vision: bool = False,
+    vision: bool = True,
     vision_max_pages: int = 30,
     format_vision_max_pages: int = 10,
     timeout: int = 7500,
@@ -299,7 +299,7 @@ def main(
         no_correctness: Skip correctness check.
         no_criticality: Skip criticality verification.
         no_bib: Skip BibTeX verification.
-        vision: Use vision mode for PDF (requires pypdfium2).
+        vision: Use vision mode for PDF (default: True, requires pypdfium2).
         vision_max_pages: Max pages in vision mode (default: 30, 0 = all).
         format_vision_max_pages: Max pages for format check in vision mode (default: 10).
         timeout: API timeout in seconds (default: 7500).

skills/paper-review/SKILL.md

@@ -104,7 +104,7 @@ python -m cookbooks.paper_review --bib_only references.bib --email your@email.co
 | `--no_correctness` | off | Skip correctness check |
 | `--no_criticality` | off | Skip criticality verification |
 | `--no_bib` | off | Skip BibTeX verification |
-| `--vision` | off | Use vision mode (requires pypdfium2) |
+| `--vision` | **on** | Use vision mode (requires pypdfium2); enabled by default |
 | `--vision_max_pages` | `30` | Max pages in vision mode (0 = all) |
 | `--format_vision_max_pages` | `10` | Max pages for format check (0 = use `--vision_max_pages`) |
 | `--timeout` | `7500` | API timeout in seconds |
@@ -147,8 +147,9 @@ will fail or produce empty reviews.
 2. `OPENAI_API_KEY` set → use the latest vision-capable OpenAI model (search web if unsure)
 3. `ANTHROPIC_API_KEY` set → use the latest vision-capable Anthropic model (search web if unsure)
 
-**Always add `--vision` flag for PDF review** to enable vision mode. Without it, the pipeline
-uses text extraction which may lose formatting, figures, and tables.
+**Vision mode is enabled by default for PDF review.** Pages are rendered as images, which
+preserves formatting, figures, and tables. To disable, pass `--no_vision` (not recommended).
+The model **must** support multimodal (vision) input.
 
 ## Additional resources