Add Azure OpenAI support with automatic detection

- Added Azure OpenAI detection in openevolve/llm/openai.py
- Added Azure OpenAI support in generate_breakthrough_paradigms.py
- Added reasoning model support (gpt-5, gpt-5-mini) with max_completion_tokens
- Updated configs/README.md with Azure configuration docs
- Updated main README.md with Azure OpenAI setup instructions
- Example config uses environment variables (no hardcoded keys)

Features:
- Automatic Azure endpoint detection (.openai.azure.com)
- Uses AzureOpenAI class for Azure endpoints
- Backward compatible with standard OpenAI
- Supports reasoning models on Azure

Author: mert-cemri

README.md

@@ -2,31 +2,61 @@
 > 
 > This repository includes **AutoEvolve**, a [Sky Computing Lab, UC Berkeley](https://sky.cs.berkeley.edu/) research extension that builds on top of OpenEvolve. AutoEvolve addresses a key challenge in LLM-driven evolution: when evolution stagnates, the system automatically generates breakthrough ideas using structured problem analysis, then adapts its exploration strategy based on what succeeds and fails.
 
-## Azure OpenAI Support
+## 🚀 Azure OpenAI Support
 
-AutoEvolve supports Azure OpenAI endpoints. The system automatically detects Azure endpoints and uses the appropriate client.
+AutoEvolve now supports **Azure OpenAI** with automatic detection. Simply configure your Azure endpoint in the config file and the system will automatically use the `AzureOpenAI` client.
 
-### Configuration
+### Quick Setup for Azure OpenAI
 
-In your `config.yaml`, set the Azure endpoint URL:
+Add the following to your `config.yaml`:
 
 ```yaml
+# LLM configuration
 llm:
   primary_model: "gpt-5"  # Your Azure deployment name
   api_base: "https://YOUR_RESOURCE.openai.azure.com/openai/deployments/YOUR_DEPLOYMENT?api-version=2024-12-01-preview"
-  # api_key: Set via OPENAI_API_KEY environment variable or uncomment and set here
-
+  api_key: "YOUR_AZURE_API_KEY"  # Set your API key here or use OPENAI_API_KEY env var
+  primary_model_weight: 1.0
+  temperature: 0.7
+  top_p: 0.95
+  max_tokens: 32000
+  timeout: 600
+
+# Database configuration
 database:
+  # ... other settings ...
+  
+  # Paradigm Breakthrough (for stagnation handling)
   enable_paradigm_breakthrough: true
-  paradigm_model: "gpt-5-mini"  # Your Azure deployment name for paradigms
+  stagnation_window: 5
+  stagnation_improvement_threshold: 0.01
+  paradigm_model: "gpt-5-mini"  # Your Azure deployment name for paradigm generation
   paradigm_api_base: "https://YOUR_RESOURCE.openai.azure.com/openai/deployments/YOUR_DEPLOYMENT?api-version=2024-12-01-preview"
-  # paradigm_api_key: Set via OPENAI_API_KEY environment variable or uncomment and set here
+  paradigm_api_key: "YOUR_AZURE_API_KEY"  # Set your API key here or use OPENAI_API_KEY env var
 ```
 
-**Important:**
-- The `api_base` URL must include the full path: `/openai/deployments/DEPLOYMENT_NAME?api-version=VERSION`
-- The `primary_model` and `paradigm_model` should match your Azure deployment names
-- API keys can be set via environment variables (`OPENAI_API_KEY`) or directly in config
-- The system automatically uses `AzureOpenAI` class when Azure endpoints are detected
+### Important Notes
+
+- **Deployment Name**: The `primary_model` and `paradigm_model` should match your Azure deployment names (e.g., `gpt-5`, `gpt-4.1`, `gpt-5-mini`)
+- **API Base URL Format**: Must include the full path: `https://YOUR_RESOURCE.openai.azure.com/openai/deployments/YOUR_DEPLOYMENT?api-version=2024-12-01-preview`
+- **API Version**: Use `2024-12-01-preview` (or your Azure API version)
+- **Automatic Detection**: The system automatically detects Azure endpoints (`.openai.azure.com`) and uses the `AzureOpenAI` client
+- **Reasoning Models**: Supports reasoning models like `gpt-5`, `gpt-5-mini` - automatically uses `max_completion_tokens` instead of `max_tokens`
+- **API Key**: You can set it in the config file or use the `OPENAI_API_KEY` environment variable
+
+### Example: Complete Azure Configuration
+
+```yaml
+llm:
+  primary_model: "gpt-5"
+  api_base: "https://east-docetl.openai.azure.com/openai/deployments/gpt-4.1?api-version=2024-12-01-preview"
+  api_key: "YOUR_API_KEY_HERE"  # Replace with your actual Azure API key
+
+database:
+  enable_paradigm_breakthrough: true
+  paradigm_model: "gpt-5-mini"
+  paradigm_api_base: "https://east-docetl.openai.azure.com/openai/deployments/gpt-4.1?api-version=2024-12-01-preview"
+  paradigm_api_key: "YOUR_API_KEY_HERE"  # Replace with your actual Azure API key
+```
 
-See `configs/README.md` for more details.
+**That's it!** Just replace `YOUR_API_KEY_HERE` with your actual Azure API key and everything will work automatically.