Add Hugging Face Spaces deployment configuration

Added files for deploying AI Personas to Hugging Face Spaces with GPU:

1. requirements_hf.txt - Dependencies for HF Spaces deployment
- Includes PyTorch, transformers, PEFT for Be.FM model
- Optimized for GPU inference on T4/A10G hardware

2. README_HF.md - Space homepage with metadata
- Hugging Face Space configuration in frontmatter
- User-facing documentation
- Model selection guide

3. .streamlit/config.toml - Streamlit configuration
- Custom theme colors
- Server settings for HF deployment
- Browser optimizations

4. DEPLOY_TO_HF.md - Complete deployment guide
- Step-by-step instructions for HF Spaces setup
- Cost estimates for GPU tiers
- Troubleshooting tips
- Secret management for API keys

This enables colleagues to access the app via browser with GPU-accelerated
Be.FM model at ~$0.60/hour on T4 small GPU.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

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

.streamlit/config.toml

@@ -0,0 +1,14 @@
+[theme]
+primaryColor = "#667eea"
+backgroundColor = "#FFFFFF"
+secondaryBackgroundColor = "#f0f2f6"
+textColor = "#262730"
+font = "sans serif"
+
+[server]
+headless = true
+enableCORS = false
+enableXsrfProtection = true
+
+[browser]
+gatherUsageStats = false

DEPLOY_TO_HF.md

@@ -0,0 +1,142 @@
+# Deploying AI Personas to Hugging Face Spaces
+
+## Prerequisites
+
+1. **Hugging Face Account**: Sign up at https://huggingface.co/join
+2. **Git installed** on your Mac (you already have this)
+
+## Step-by-Step Deployment
+
+### 1. Create a New Space
+
+1. Go to https://huggingface.co/spaces
+2. Click "Create new Space"
+3. Fill in:
+   - **Space name**: `ai-personas-urban-planning` (or your choice)
+   - **License**: MIT
+   - **SDK**: Streamlit
+   - **Space hardware**: **T4 small** (GPU, ~$0.60/hour) or **A10G small** (~$3/hour, faster)
+   - **Visibility**: Public or Private (your choice)
+4. Click "Create Space"
+
+### 2. Clone Your New Space
+
+```bash
+# In terminal on your Mac
+cd ~/Desktop  # or wherever you want
+
+# Clone the empty space (replace USERNAME and SPACENAME)
+git clone https://huggingface.co/spaces/USERNAME/SPACENAME
+cd SPACENAME
+```
+
+### 3. Copy Your AI Personas Files
+
+```bash
+# From your AI_Personas directory, copy everything except git history
+cd ~/path/to/AI_Personas
+
+# Copy all necessary files
+cp -r data/ pages/ src/ ../SPACENAME/
+cp web_app.py ../SPACENAME/
+cp requirements_hf.txt ../SPACENAME/requirements.txt
+cp README_HF.md ../SPACENAME/README.md
+cp -r .streamlit/ ../SPACENAME/
+
+# Copy these if they exist
+cp *.py ../SPACENAME/ 2>/dev/null || true
+```
+
+### 4. Commit and Push to Hugging Face
+
+```bash
+cd ../SPACENAME
+
+# Add all files
+git add .
+
+# Commit
+git commit -m "Initial deployment of AI Personas app"
+
+# Push to Hugging Face (will trigger deployment)
+git push
+```
+
+### 5. Wait for Build
+
+1. Go to your Space page: `https://huggingface.co/spaces/USERNAME/SPACENAME`
+2. Watch the "Building" status (takes 5-10 minutes first time)
+3. Once it says "Running", click the app!
+
+### 6. Configure Secrets (for Anthropic Claude)
+
+If you want the Anthropic Claude option to work:
+
+1. Go to your Space → Settings → Repository secrets
+2. Add secret:
+   - **Name**: `ANTHROPIC_API_KEY`
+   - **Value**: Your Claude API key
+3. Save
+
+Then update `src/llm/anthropic_client.py` to use it:
+
+```python
+# In __init__ method, change:
+self.api_key = api_key or os.environ.get("ANTHROPIC_API_KEY")
+```
+
+## Cost Estimates
+
+- **T4 small GPU**: ~$0.60/hour = ~$14/day if running 24/7
+- **Free tier**: CPU only (Be.FM won't work well)
+- **Pause when not in use**: Turn off in Space settings to save costs
+
+## Sharing with Colleagues
+
+Once deployed, share the URL:
+```
+https://huggingface.co/spaces/USERNAME/SPACENAME
+```
+
+Anyone can use it! No installation needed.
+
+## Troubleshooting
+
+**Space keeps crashing?**
+- Check the "Logs" tab in your Space
+- You might need more RAM → upgrade to A10G
+
+**Be.FM loading slowly?**
+- First load takes 2-3 minutes to download model
+- After that, it's cached and loads in ~30 seconds
+
+**Want to update the app?**
+```bash
+# Make changes locally
+git add .
+git commit -m "Update app"
+git push
+
+# Space will auto-rebuild in 2-5 minutes
+```
+
+## Alternative: Free CPU-Only Deployment
+
+If you don't want to pay for GPU:
+
+1. Change hardware to "CPU basic" (free)
+2. Modify `src/llm/local_model_client.py`:
+   ```python
+   # Force CPU mode
+   self.device = "cpu"
+   ```
+3. Set expectation with colleagues: responses take 2-3 minutes each
+
+Free but slow!
+
+---
+
+## Need Help?
+
+Check the Hugging Face Spaces documentation:
+https://huggingface.co/docs/hub/spaces-overview

README_HF.md

@@ -0,0 +1,42 @@
+---
+title: AI Personas for Urban Planning
+emoji: 🏙️
+colorFrom: blue
+colorTo: purple
+sdk: streamlit
+sdk_version: "1.28.0"
+app_file: web_app.py
+pinned: false
+license: mit
+---
+
+# 🏙️ AI Personas for Urban Planning
+
+An AI-powered simulation system that models diverse stakeholder perspectives on urban planning issues using LLM-based personas.
+
+## Features
+
+- **Interactive Chat**: Query individual personas to understand their unique perspectives
+- **Population Analysis**: Analyze responses across population distributions
+- **Opinion Equilibria**: Model how opinions evolve through social influence dynamics
+- **Edit Personas**: Customize stakeholder profiles and urban contexts
+
+## Models
+
+- **Anthropic Claude**: High-quality API-based responses (requires API key)
+- **Local Be.FM**: Stanford's Be.FM model running on GPU (free, no API key needed)
+
+## Usage
+
+1. Select a model from the sidebar (Be.FM recommended for this Space)
+2. Choose a persona or use Population Analysis
+3. Ask questions about urban planning scenarios
+4. Explore how different stakeholders respond
+
+## About Be.FM
+
+Be.FM is a fine-tuned Llama-3.1-8B model trained on behavioral economics data, designed to simulate realistic human decision-making in urban planning contexts.
+
+## Credits
+
+Built with Streamlit, Anthropic Claude, and Stanford's Be.FM model.

requirements_hf.txt

@@ -0,0 +1,21 @@
+# Hugging Face Spaces Requirements
+# For GPU-accelerated deployment with Be.FM model
+
+streamlit>=1.28.0
+anthropic>=0.7.0
+pydantic>=2.0.0
+plotly>=5.17.0
+networkx>=3.1
+numpy>=1.24.0
+pandas>=2.0.0
+
+# PyTorch and ML dependencies
+torch>=2.1.0
+transformers>=4.35.0
+accelerate>=0.24.0
+peft>=0.6.0
+sentencepiece>=0.1.99
+protobuf>=3.20.0
+
+# For better performance
+bitsandbytes>=0.41.0