Convert from Lua filter to TypeScript engine extension

- Rewrite marimo execution as a Quarto engine extension (TypeScript)
- Support three cell syntaxes: {python .marimo}, {python.marimo}, python {.marimo}
- Implement claimsLanguage for automatic engine detection (priority 2 for class syntax)
- Add cell-execution-regex.ts with comprehensive test suite
- Fix PDF output handling (mimebundle, raw HTML blocks)
- Update all examples to use preferred {python .marimo} syntax
- Requires quarto >= 1.9.19

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: gordonwoodhull

.gitignore

@@ -23,3 +23,5 @@ result
 # pixi environments
 .pixi
 *.egg-info
+
+**/*.quarto_ipynb

README.md

@@ -103,4 +103,16 @@ Javascript.
 
 ---
 
+## Development
+
+To build the TypeScript engine extension:
+
+```bash
+quarto call build-ts-extension
+```
+
+This bundles `src/marimo-engine.ts` into `_extensions/marimo/marimo-engine.js`.
+
+---
+
 Credits: [holoviz-quarto](https://github.com/awesome-panel/holoviz-quarto) for ideas on layout

_extensions/marimo/_extension.yml

@@ -1,6 +1,6 @@
 title: Quarto Marimo Extension
-version: 0.4.3
-quarto-required: ">=1.3.0"
+version: 0.5.0
+quarto-required: ">=1.9.19"
 contributes:
-  filters:
-    - marimo-execute.lua
+  engines:
+    - path: marimo-engine.js

_extensions/marimo/command.py

@@ -31,7 +31,7 @@ def extract_command(header: str) -> list[str]:
         mode="w", delete=False, suffix=".txt"
     ) as temp_file:
         flags = construct_uv_flags(pyproject, temp_file, [], [])
-
+        temp_file.flush()
     return ["run"] + flags
 
 

_extensions/marimo/extract.py

@@ -91,6 +91,19 @@ def get_mime_render(
         if config["output"] and mime_sensitive:
             if mimetype.startswith("image"):
                 return {"type": "figure", "value": f"{output.data}", **render_options}
+            # Handle mimebundle - extract image data if present
+            if mimetype == "application/vnd.marimo+mimebundle":
+                try:
+                    bundle = json.loads(output.data) if isinstance(output.data, str) else output.data
+                    # Look for image data in the bundle
+                    for key in ["image/png", "image/jpeg", "image/svg+xml"]:
+                        if key in bundle:
+                            return {"type": "figure", "value": bundle[key], **render_options}
+                    # Fall back to text if no image
+                    if "text/plain" in bundle:
+                        return {"type": "para", "value": bundle["text/plain"], **render_options}
+                except (json.JSONDecodeError, TypeError):
+                    pass  # Fall through to default handling
             if mimetype.startswith("text/plain") or mimetype.startswith(
                 "text/markdown"
             ):