Add academic models and focus

- Add anthropic/claude-3-haiku and nvidia/nemotron-4-340b-instruct
- Update tagline and system prompt for academic research
- Include recent .gitignore and requirement updates

.gitignore

@@ -0,0 +1,26 @@
+# Environment variables
+.env
+.env.local
+
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+env/
+venv/
+ENV/
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+
+# OS
+.DS_Store
+Thumbs.db
+
+# Logs
+*.log

CLAUDE.md

@@ -17,33 +17,60 @@ pip install -r requirements.txt
 
 ## Code Architecture
 
-This is a single-file Gradio application (`app.py`) that helps users create and deploy chat interfaces on HuggingFace Spaces. The application has two main features:
+This is a single-file Gradio 4.32.0 application (`app.py`) that helps users create and deploy chat interfaces on HuggingFace Spaces. The application has two main features:
 
 1. **Spaces Configuration Tool**: Generates complete deployment packages for chat interfaces
-   - Uses the `SPACE_TEMPLATE` string (lines 9-81) to generate Python code
-   - Creates zip files containing app.py, requirements.txt, and README.md
+   - Uses the `SPACE_TEMPLATE` string (lines 15-137) to generate Python code
+   - Creates zip files containing app.py, requirements.txt, README.md, and config.json
    - Supports multiple LLM models via OpenRouter API
+   - **URL Grounding**: Supports up to 4 discrete URLs for context grounding
+     - Fetches web content from provided URLs using BeautifulSoup
+     - Injects content as context into the system prompt
+     - Automatically truncates content to ~4000 characters per URL
 
 2. **Chat Support Assistant**: Provides configuration help using Gemma-2-27B
    - Located in the second tab of the interface
    - Uses OpenRouter API with environment variable `OPENROUTER_API_KEY`
+   - Includes URL grounding functionality in collapsible accordion
 
 ### Key Functions
-- `create_readme()`: Generates deployment instructions (lines 99-117)
-- `create_requirements()`: Generates requirements.txt content (lines 119-127)
-- `generate_space_config()`: Main function that creates the deployment package (lines 129-180)
-- `create_support_assistant()`: Creates the chat support interface (lines 396-449)
+- `create_readme()`: Generates deployment instructions with HuggingFace YAML frontmatter
+- `create_requirements()`: Generates requirements.txt content including beautifulsoup4
+- `generate_zip()`: Main function that creates the deployment package with URL grounding support
+- `on_generate()`: Callback function for the generate button handling all input parameters
+- `fetch_url_content()`: Fetches and extracts text from URLs (both in template and main app)
+- `get_grounding_context()`: Combines URL content into context (both in template and main app)
+- `respond()`: Chat support function with URL grounding parameters
 
 ### Template System
 The app uses string templating to generate Python code. The main template (`SPACE_TEMPLATE`) contains placeholders that are filled with user configuration:
 - `{name}`, `{description}`, `{system_prompt}`, `{model}`
 - `{api_key_var}`, `{temperature}`, `{max_tokens}`, `{examples}`
+- `{grounding_urls}`: JSON array of URLs for context grounding
+
+### Model Configuration
+Available models are limited to three open-source options:
+- `google/gemma-2-27b-it` (default)
+- `mistralai/mixtral-8x7b-instruct`
+- `meta-llama/llama-3.1-70b-instruct`
+
+### URL Grounding Architecture
+URL grounding is implemented in both the main app and generated templates:
+- Web scraping uses BeautifulSoup4 with 10-second timeout
+- Content is cleaned (scripts/styles removed) and truncated to 4000 characters
+- Error handling for failed requests returns descriptive error messages
+- Context is injected into system prompts for enhanced responses
 
 ## Environment Variables
 - `OPENROUTER_API_KEY`: Required for the Chat Support feature to work
+- Uses python-dotenv to load from `.env` file automatically
 
 ## Important Notes
 - This is a HuggingFace Spaces application (see YAML frontmatter in README.md)
 - No test suite or linting configuration exists
 - The generated packages are self-contained chat applications ready for deployment
-- All generated code uses OpenRouter API for LLM access
+- All generated code uses OpenRouter API for LLM access
+- Generated spaces include BeautifulSoup4 for URL content extraction
+- URL grounding fetches content at runtime, adding context to each conversation
+- Generated README.md files include proper HuggingFace Spaces YAML frontmatter
+- Template uses double braces `{{}}` for literal braces in generated code, single braces `{}` for template variables

app.py

@@ -4,30 +4,86 @@ import zipfile
 import io
 import os
 from datetime import datetime
+from dotenv import load_dotenv
+import requests
+from bs4 import BeautifulSoup
+
+# Load environment variables from .env file
+load_dotenv()
 
 # Template for generated space app (based on mvp_simple.py)
 SPACE_TEMPLATE = '''import gradio as gr
 import os
 import requests
 import json
+from bs4 import BeautifulSoup
 
 # Configuration
 SPACE_NAME = "{name}"
 SPACE_DESCRIPTION = "{description}"
 SYSTEM_PROMPT = """{system_prompt}"""
 MODEL = "{model}"
+GROUNDING_URLS = {grounding_urls}
 
 # Get API key from environment - customizable variable name
 API_KEY = os.environ.get("{api_key_var}")
 
+def fetch_url_content(url):
+    """Fetch and extract text content from a URL"""
+    try:
+        response = requests.get(url, timeout=10)
+        response.raise_for_status()
+        soup = BeautifulSoup(response.content, 'html.parser')
+        
+        # Remove script and style elements
+        for script in soup(["script", "style"]):
+            script.decompose()
+        
+        # Get text content
+        text = soup.get_text()
+        
+        # Clean up whitespace
+        lines = (line.strip() for line in text.splitlines())
+        chunks = (phrase.strip() for line in lines for phrase in line.split("  "))
+        text = ' '.join(chunk for chunk in chunks if chunk)
+        
+        # Truncate to ~4000 characters
+        if len(text) > 4000:
+            text = text[:4000] + "..."
+        
+        return text
+    except Exception as e:
+        return f"Error fetching {{url}}: {{str(e)}}"
+
+def get_grounding_context():
+    """Fetch context from grounding URLs"""
+    if not GROUNDING_URLS:
+        return ""
+    
+    context_parts = []
+    for i, url in enumerate(GROUNDING_URLS, 1):
+        if url.strip():
+            content = fetch_url_content(url.strip())
+            context_parts.append(f"Context from URL {{i}} ({{url}}):\\n{{content}}")
+    
+    if context_parts:
+        return "\\n\\n" + "\\n\\n".join(context_parts) + "\\n\\n"
+    return ""
+
 def generate_response(message, history):
     """Generate response using OpenRouter API"""
     
     if not API_KEY:
         return "Please set your {api_key_var} in the Space settings."
     
+    # Get grounding context
+    grounding_context = get_grounding_context()
+    
+    # Build enhanced system prompt with grounding context
+    enhanced_system_prompt = SYSTEM_PROMPT + grounding_context
+    
     # Build messages array for the API
-    messages = [{{"role": "system", "content": SYSTEM_PROMPT}}]
+    messages = [{{"role": "system", "content": enhanced_system_prompt}}]
     
     # Add conversation history - compatible with Gradio 5.x format
     for chat in history:
@@ -84,12 +140,67 @@ if __name__ == "__main__":
 MODELS = [
     "google/gemma-2-27b-it",
     "mistralai/mixtral-8x7b-instruct",
-    "meta-llama/llama-3.1-70b-instruct"
+    "meta-llama/llama-3.1-70b-instruct",
+    "anthropic/claude-3-haiku",
+    "nvidia/nemotron-4-340b-instruct"
 ]
 
+def fetch_url_content(url):
+    """Fetch and extract text content from a URL"""
+    try:
+        response = requests.get(url, timeout=10)
+        response.raise_for_status()
+        soup = BeautifulSoup(response.content, 'html.parser')
+        
+        # Remove script and style elements
+        for script in soup(["script", "style"]):
+            script.decompose()
+        
+        # Get text content
+        text = soup.get_text()
+        
+        # Clean up whitespace
+        lines = (line.strip() for line in text.splitlines())
+        chunks = (phrase.strip() for line in lines for phrase in line.split("  "))
+        text = ' '.join(chunk for chunk in chunks if chunk)
+        
+        # Truncate to ~4000 characters
+        if len(text) > 4000:
+            text = text[:4000] + "..."
+        
+        return text
+    except Exception as e:
+        return f"Error fetching {url}: {str(e)}"
+
+def get_grounding_context(urls):
+    """Fetch context from grounding URLs"""
+    if not urls:
+        return ""
+    
+    context_parts = []
+    for i, url in enumerate(urls, 1):
+        if url and url.strip():
+            content = fetch_url_content(url.strip())
+            context_parts.append(f"Context from URL {i} ({url}):\n{content}")
+    
+    if context_parts:
+        return "\n\n" + "\n\n".join(context_parts) + "\n\n"
+    return ""
+
 def create_readme(config):
     """Generate README with deployment instructions"""
-    return f"""# {config['name']}
+    return f"""---
+title: {config['name']}
+emoji: 🤖
+colorFrom: blue
+colorTo: red
+sdk: gradio
+sdk_version: 4.32.0
+app_file: app.py
+pinned: false
+---
+
+# {config['name']}
 
 {config['description']}
 
@@ -174,9 +285,9 @@ Generated on {datetime.now().strftime('%Y-%m-%d %H:%M:%S')} with Chat U/I Helper
 
 def create_requirements():
     """Generate requirements.txt"""
-    return "gradio==5.34.0\nrequests==2.31.0"
+    return "gradio==4.44.1\nrequests==2.32.3\nbeautifulsoup4==4.12.3"
 
-def generate_zip(name, description, system_prompt, model, api_key_var, temperature, max_tokens, examples_text):
+def generate_zip(name, description, system_prompt, model, api_key_var, temperature, max_tokens, examples_text, url1="", url2="", url3="", url4=""):
     """Generate deployable zip file"""
     
     # Process examples
@@ -190,6 +301,12 @@ def generate_zip(name, description, system_prompt, model, api_key_var, temperatu
             "What can you do?"
         ])
     
+    # Process grounding URLs
+    grounding_urls = []
+    for url in [url1, url2, url3, url4]:
+        if url and url.strip():
+            grounding_urls.append(url.strip())
+    
     # Create config
     config = {
         'name': name,
@@ -199,7 +316,8 @@ def generate_zip(name, description, system_prompt, model, api_key_var, temperatu
         'api_key_var': api_key_var,
         'temperature': temperature,
         'max_tokens': int(max_tokens),
-        'examples': examples_json
+        'examples': examples_json,
+        'grounding_urls': json.dumps(grounding_urls)
     }
     
     # Generate files
@@ -226,7 +344,7 @@ def generate_zip(name, description, system_prompt, model, api_key_var, temperatu
     return filename
 
 # Define callback functions outside the interface
-def on_generate(name, description, system_prompt, model, api_key_var, temperature, max_tokens, examples_text):
+def on_generate(name, description, system_prompt, model, api_key_var, temperature, max_tokens, examples_text, url1, url2, url3, url4):
     if not name or not name.strip():
         return gr.update(value="Error: Please provide a Space Title", visible=True), gr.update(visible=False)
     
@@ -234,7 +352,7 @@ def on_generate(name, description, system_prompt, model, api_key_var, temperatur
         return gr.update(value="Error: Please provide a System Prompt", visible=True), gr.update(visible=False)
     
     try:
-        filename = generate_zip(name, description, system_prompt, model, api_key_var, temperature, max_tokens, examples_text)
+        filename = generate_zip(name, description, system_prompt, model, api_key_var, temperature, max_tokens, examples_text, url1, url2, url3, url4)
         
         success_msg = f"""**Deployment package ready!**
 
@@ -258,7 +376,7 @@ def on_generate(name, description, system_prompt, model, api_key_var, temperatur
     except Exception as e:
         return gr.update(value=f"Error: {str(e)}", visible=True), gr.update(visible=False)
 
-def respond(message, chat_history):
+def respond(message, chat_history, url1="", url2="", url3="", url4=""):
     # Make actual API request to OpenRouter
     import os
     import requests
@@ -268,25 +386,38 @@ def respond(message, chat_history):
     
     if not api_key:
         response = "Please set your OPENROUTER_API_KEY in the Space settings to use the chat support."
-        chat_history.append({"role": "user", "content": message})
-        chat_history.append({"role": "assistant", "content": response})
+        chat_history.append([message, response])
         return "", chat_history
     
-    # Build conversation history for API
-    messages = [{
-        "role": "system", 
-        "content": """You are a helpful assistant specializing in creating chat UIs for HuggingFace Spaces. You help users configure assistants for education and research. Provide concise, practical advice about:
+    # Get grounding context from URLs
+    grounding_urls = [url1, url2, url3, url4]
+    grounding_context = get_grounding_context(grounding_urls)
+    
+    # Build enhanced system prompt with grounding context
+    base_system_prompt = """You are a helpful assistant specializing in creating chat UIs for HuggingFace Spaces. You help users configure assistants for education and research. Provide concise, practical advice about:
 - System prompts for different use cases (courses, research, tutoring)
 - Model selection (recommending google/gemma-2-27b-it as a great balance)
 - HuggingFace Space deployment tips
 - Customization options
 
 Keep responses brief and actionable. Focus on what the user is specifically asking about."""
+    
+    enhanced_system_prompt = base_system_prompt + grounding_context
+    
+    # Build conversation history for API
+    messages = [{
+        "role": "system", 
+        "content": enhanced_system_prompt
     }]
     
-    # Add conversation history
+    # Add conversation history - Gradio 4.x uses list/tuple format
     for chat in chat_history:
-        messages.append(chat)
+        if isinstance(chat, (list, tuple)) and len(chat) >= 2:
+            user_msg, assistant_msg = chat[0], chat[1]
+            if user_msg:
+                messages.append({"role": "user", "content": user_msg})
+            if assistant_msg:
+                messages.append({"role": "assistant", "content": assistant_msg})
     
     # Add current message
     messages.append({"role": "user", "content": message})
@@ -315,8 +446,7 @@ Keep responses brief and actionable. Focus on what the user is specifically aski
     except Exception as e:
         assistant_response = f"Error: {str(e)}"
     
-    chat_history.append({"role": "user", "content": message})
-    chat_history.append({"role": "assistant", "content": assistant_response})
+    chat_history.append([message, assistant_response])
     return "", chat_history
 
 def clear_chat():
@@ -340,7 +470,7 @@ with gr.Blocks(title="Chat U/I Helper") as demo:
                     label="Space Description", 
                     placeholder="A customizable AI chat interface for...",
                     lines=2,
-                    value="A customizable AI chat interface for your specific needs"
+                    value="An AI research assistant tailored for academic inquiry and scholarly dialogue"
                 )
                 
                 model = gr.Dropdown(
@@ -360,7 +490,7 @@ with gr.Blocks(title="Chat U/I Helper") as demo:
                     label="System Prompt",
                     placeholder="You are a research assistant...",
                     lines=4,
-                    value="You are a clear and concise research assistant. Provide  accurate, succinct, and responsive support."
+                    value="You are a knowledgeable academic research assistant. Provide thoughtful, evidence-based guidance for scholarly work, literature reviews, and academic writing. Support students and researchers with clear explanations and critical thinking."
                 )
                 
                 examples_text = gr.Textbox(
@@ -370,6 +500,33 @@ with gr.Blocks(title="Chat U/I Helper") as demo:
                     info="These will appear as clickable examples in the chat interface"
                 )
                 
+                gr.Markdown("### URL Grounding (Optional)")
+                gr.Markdown("Add up to 4 URLs to provide context. Content will be fetched and added to the system prompt.")
+                
+                url1 = gr.Textbox(
+                    label="URL 1",
+                    placeholder="https://example.com/page1",
+                    info="First URL for context grounding"
+                )
+                
+                url2 = gr.Textbox(
+                    label="URL 2", 
+                    placeholder="https://example.com/page2",
+                    info="Second URL for context grounding"
+                )
+                
+                url3 = gr.Textbox(
+                    label="URL 3",
+                    placeholder="https://example.com/page3", 
+                    info="Third URL for context grounding"
+                )
+                
+                url4 = gr.Textbox(
+                    label="URL 4",
+                    placeholder="https://example.com/page4",
+                    info="Fourth URL for context grounding"
+                )
+                
                 with gr.Row():
                     temperature = gr.Slider(
                         label="Temperature",
@@ -396,7 +553,7 @@ with gr.Blocks(title="Chat U/I Helper") as demo:
             # Connect the generate button
             generate_btn.click(
                 on_generate,
-                inputs=[name, description, system_prompt, model, api_key_var, temperature, max_tokens, examples_text],
+                inputs=[name, description, system_prompt, model, api_key_var, temperature, max_tokens, examples_text, url1, url2, url3, url4],
                 outputs=[status, download_file]
             )
         
@@ -409,14 +566,36 @@ with gr.Blocks(title="Chat U/I Helper") as demo:
                 chatbot = gr.Chatbot(
                     value=[],
                     label="Chat Support Assistant",
-                    height=400,
-                    type="messages"
+                    height=400
                 )
                 msg = gr.Textbox(
                     label="Ask about configuring chat UIs for courses, research, or custom HuggingFace Spaces",
                     placeholder="How can I configure a chat UI for my senior seminar?",
                     lines=2
                 )
+                
+                with gr.Accordion("URL Grounding (Optional)", open=False):
+                    gr.Markdown("Add URLs to provide additional context for more informed responses")
+                    chat_url1 = gr.Textbox(
+                        label="URL 1",
+                        placeholder="https://example.com/context1",
+                        info="First URL for context"
+                    )
+                    chat_url2 = gr.Textbox(
+                        label="URL 2", 
+                        placeholder="https://example.com/context2",
+                        info="Second URL for context"
+                    )
+                    chat_url3 = gr.Textbox(
+                        label="URL 3",
+                        placeholder="https://example.com/context3",
+                        info="Third URL for context"
+                    )
+                    chat_url4 = gr.Textbox(
+                        label="URL 4",
+                        placeholder="https://example.com/context4",
+                        info="Fourth URL for context"
+                    )
                 with gr.Row():
                     submit = gr.Button("Send", variant="primary")
                     clear = gr.Button("Clear")
@@ -434,9 +613,9 @@ with gr.Blocks(title="Chat U/I Helper") as demo:
                 )
             
             # Connect the chat functionality
-            submit.click(respond, [msg, chatbot], [msg, chatbot])
-            msg.submit(respond, [msg, chatbot], [msg, chatbot])
+            submit.click(respond, [msg, chatbot, chat_url1, chat_url2, chat_url3, chat_url4], [msg, chatbot])
+            msg.submit(respond, [msg, chatbot, chat_url1, chat_url2, chat_url3, chat_url4], [msg, chatbot])
             clear.click(clear_chat, outputs=[msg, chatbot])
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(share=True)

local-files/claude.md

@@ -1,108 +0,0 @@
-# CLAUDE.md
-
-This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
-
-## Architecture Overview
-
-This is a **Chat UI Helper** that generates ready-to-deploy HuggingFace Spaces for educational and research chat interfaces. The project combines a Gradio-based package generator with an AI-powered configuration assistant.
-
-### Core Components
-
-1. **Main Application (`app.py`)**: Two-tab Gradio interface:
-   - **Space Generator Tab**: Creates deployable HuggingFace Space packages
-   - **Chat UI Helper Tab**: AI assistant for configuration guidance (imported from `chat_helper.py`)
-
-2. **Template System**: 
-   - `SPACE_TEMPLATE` string contains complete deployable chat interface code
-   - Uses Python `.format()` to inject user configurations into template
-   - Generates zip packages with `app.py`, `requirements.txt`, `README.md`, and `config.json`
-
-3. **Chat Helper (`chat_helper.py`)**: 
-   - Modular AI assistant component for configuration guidance
-   - Rule-based responses for common educational/research chat UI questions
-   - Emphasizes Mixtral-8x7B as cost-effective recommended model
-
-### Key Architecture Patterns
-
-- **Template Injection**: Uses `{{}}` for literal braces in generated code, `{}` for template variables
-- **Modular Design**: Chat helper is separate module imported by main app
-- **Configuration-Driven**: All generated spaces use environment-based API key configuration
-- **Educational Focus**: Specialized for academic and research use cases
-
-## Development Commands
-
-### Local Development
-```bash
-# Install dependencies
-pip install -r requirements.txt
-
-# Run the main application
-python app.py
-
-# Test a basic assistant template
-python default.py
-```
-
-### Testing Generated Packages
-```bash
-# Extract and test a generated zip package
-unzip my_custom_space.zip
-cd extracted_package
-export OPENROUTER_API_KEY="your-key-here"
-python app.py
-```
-
-### Requirements
-- All generated chat interfaces require OpenRouter API keys
-- Current Gradio version: 5.34.0 (updated from legacy 4.16.0)
-- Template generates Gradio ChatInterface applications
-
-## Template System Details
-
-### SPACE_TEMPLATE Structure
-The template generates complete HuggingFace Spaces with:
-- OpenRouter API integration using configurable environment variables
-- Gradio 5.x compatible ChatInterface implementation  
-- Error handling for missing API keys
-- Configurable model, temperature, and token limits
-- Support for custom example prompts
-
-### Configuration Variables
-Standard variables injected into templates:
-- `{name}`: Space title/name
-- `{description}`: Space description
-- `{system_prompt}`: AI behavior instructions
-- `{model}`: OpenRouter model identifier
-- `{api_key_var}`: Environment variable name for API key
-- `{temperature}`: Model temperature setting
-- `{max_tokens}`: Maximum response tokens
-- `{examples}`: JSON array of example prompts
-
-### Template Compatibility
-- Template uses double braces `{{API_KEY}}` for variables that should appear literally in generated code
-- Single braces `{model}` are replaced during template formatting
-- Generated code is compatible with HuggingFace Spaces deployment
-
-## Model Configuration
-
-### Default Model List
-Available models prioritized by use case:
-- **Budget**: `google/gemma-2-9b-it`, `mistralai/mistral-7b-instruct`
-- **Recommended**: `mistralai/mixtral-8x7b-instruct` (balanced cost/performance)
-- **Premium**: `anthropic/claude-3.5-sonnet`, `openai/gpt-4o`
-
-### Chat Helper Integration
-The `chat_helper.py` module provides AI-powered guidance specifically recommending Mixtral-8x7B for most educational and research use cases, with configuration examples and cost considerations.
-
-## File Structure Patterns
-
-- `app.py`: Main two-tab application
-- `chat_helper.py`: Modular AI assistant component  
-- `default.py`: Basic working assistant template for testing
-- `config.json`: Example configuration file
-- `exp/`: Legacy experimental versions and documentation
-- Generated zip files follow pattern: `{name}_{timestamp}.zip` → `{name}.zip`
-
-## Current State
-
-The project has evolved from experimental multi-SDK generator (`exp/` folder) to focused educational/research chat UI helper. Current production stack generates HuggingFace Spaces optimized for academic use cases with built-in configuration guidance.

my_custom_space.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c11e277aba67311812d79895394435b16238996ac1e6cad634a0dd5c1c09297
+size 3683

requirements.txt

@@ -1,2 +1,4 @@
-gradio==5.34.0
-requests==2.31.0
+gradio==4.32.0
+requests==2.32.3
+beautifulsoup4==4.12.3
+python-dotenv==1.0.0

test_assistant_with_grounding.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66147023e9f806ed65916773337a4f7a6a874f080300c097824b95de322aea85
+size 3489

test_bot.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ba01cfc6f9dd205102f0f065fb9184397126aaba5b3225ee98b8cb369433eee
+size 3353