update

backend_deploy.py

@@ -288,29 +288,46 @@ def prettify_comfyui_json_for_html(json_content: str) -> str:
         return json_content
 
 
-def parse_transformers_js_output(code: str) -> Dict[str, str]:
-    """Parse transformers.js output into separate files (index.html, index.js, style.css)
-    
-    Uses comprehensive parsing patterns to handle various LLM output formats.
+# Note: parse_transformers_js_output, parse_python_requirements, strip_tool_call_markers,
+# remove_code_block, extract_import_statements, generate_requirements_txt_with_llm,
+# and parse_multi_file_python_output are now imported from backend_parsers.py
+
+
+def is_streamlit_code(code: str) -> bool:
+    """Check if code is Streamlit"""
+    return 'import streamlit' in code or 'streamlit.run' in code
+
+
+def is_gradio_code(code: str) -> bool:
+    """Check if code is Gradio"""
+    return 'import gradio' in code or 'gr.' in code
+
+
+def detect_sdk_from_code(code: str, language: str) -> str:
+    """Detect the appropriate SDK from code and language"""
+    if language == "html":
+        return "static"
+    elif language == "transformers.js":
+        return "static"
+    elif language == "comfyui":
+        return "static"
+    elif language == "react":
+        return "docker"
+    elif language == "streamlit" or is_streamlit_code(code):
+        return "docker"
+    elif language == "gradio" or is_gradio_code(code):
+        return "gradio"
+    else:
+        return "gradio"  # Default
+
+
+def add_anycoder_tag_to_readme(api, repo_id: str, app_port: Optional[int] = None) -> None:
     """
-    files = {
-        'index.html': '',
-        'index.js': '',
-        'style.css': ''
-    }
-    
-    # Multiple patterns to match the three code blocks with different variations
-    html_patterns = [
-        r'```html\s*\n([\s\S]*?)(?:```|\Z)',
-        r'```htm\s*\n([\s\S]*?)(?:```|\Z)',
-        r'```\s*(?:index\.html|html)\s*\n([\s\S]*?)(?:```|\Z)'
-    ]
+    Download existing README, add anycoder tag and app_port if needed, and upload back.
+    Preserves all existing README content and frontmatter.
     
-    js_patterns = [
-        r'```javascript\s*\n([\s\S]*?)(?:```|\Z)',
-        r'```js\s*\n([\s\S]*?)(?:```|\Z)',
-        r'```\s*(?:index\.js|javascript|js)\s*\n([\s\S]*?)(?:```|\Z)'
-    ]
+    Args:
+        api: HuggingFace API client
     
     css_patterns = [
         r'```css\s*\n([\s\S]*?)(?:```|\Z)',

backend_parsers.py

@@ -15,12 +15,17 @@ def parse_transformers_js_output(code: str) -> Dict[str, str]:
     Uses comprehensive parsing patterns to handle various LLM output formats.
     Updated to use transformers.js v3.8.0 CDN.
     """
+    print(f"[Parser] Received code length: {len(code)} characters")
+    print(f"[Parser] First 200 chars: {code[:200]}")
+    
     # Auto-fix: If code doesn't start with === index.html ===, add it
     code_stripped = code.strip()
     if not code_stripped.startswith('==='):
         print("[Parser] Auto-fixing: Adding missing === index.html === marker")
         code = '=== index.html ===\n' + code
         code_stripped = code.strip()
+    else:
+        print("[Parser] Code starts with === marker, proceeding normally")
     
     # Check if code starts with HTML instead of markers (common LLM mistake)
     if code_stripped.startswith('<!DOCTYPE') or code_stripped.startswith('<html'):
@@ -125,9 +130,12 @@ def parse_transformers_js_output(code: str) -> Dict[str, str]:
     # Fallback: support === index.html === format if any file is missing
     if not (files['index.html'] and files['index.js'] and files['style.css']):
         # Use regex to extract sections - match === markers with optional whitespace and newlines
-        html_fallback = re.search(r'===\s*index\.html\s*===\s*[\r\n]+([\s\S]+?)(?=\n===|$)', code, re.IGNORECASE)
-        js_fallback = re.search(r'===\s*index\.js\s*===\s*[\r\n]+([\s\S]+?)(?=\n===|$)', code, re.IGNORECASE)
-        css_fallback = re.search(r'===\s*style\.css\s*===\s*[\r\n]+([\s\S]+?)(?=\n===|$)', code, re.IGNORECASE)
+        # Made [\r\n]+ optional with * instead of + to handle cases where content follows immediately
+        html_fallback = re.search(r'===\s*index\.html\s*===\s*[\r\n]*([\s\S]+?)(?=\n===|$)', code, re.IGNORECASE)
+        js_fallback = re.search(r'===\s*index\.js\s*===\s*[\r\n]*([\s\S]+?)(?=\n===|$)', code, re.IGNORECASE)
+        css_fallback = re.search(r'===\s*style\.css\s*===\s*[\r\n]*([\s\S]+?)(?=\n===|$)', code, re.IGNORECASE)
+        
+        print(f"[Parser] Fallback extraction - HTML found: {bool(html_fallback)}, JS found: {bool(js_fallback)}, CSS found: {bool(css_fallback)}")
         
         if html_fallback:
             content = html_fallback.group(1).strip()