terradev-cli 3.7.1__tar.gz → 3.7.2__tar.gz

terradev_cli-3.7.2/CHANGELOG_v3.7.1.md

@@ -0,0 +1,77 @@
+# Changelog v3.7.1
+
+## 🚀 CUDA Graph Optimization with NUMA Awareness
+
+### 🧠 Revolutionary Passive Optimization
+- **Automatic CUDA Graph Detection**: Passively analyzes models for CUDA Graph compatibility without user intervention
+- **NUMA-Aware Scoring**: Rates endpoints by CUDA Graph performance potential (PIX: 1.0, PXB: 0.8, PHB: 0.6, SYS: 0.3)
+- **Model-Specific Intelligence**: Different optimization strategies for transformers, CNNs, and MoE models
+- **Background Analysis**: Runs automatically every 5 minutes to optimize warm pool and endpoint selection
+
+### 🔍 NUMA Topology Intelligence
+- **GPU/NIC Alignment**: Automatically detects and prioritizes endpoints with optimal PCIe topology
+- **Intra-GPU NUMA Support**: AMD MI300X XCD locality awareness for maximum performance
+- **RDMA Optimization**: Boosts scores for endpoints with GPUDirect RDMA capabilities
+- **Performance Estimates**: Provides 1.2-5x speedup estimates based on topology analysis
+
+### 📊 Smart Model Detection
+- **Transformers**: Highest priority (0.9 base score) - benefit most from CUDA Graphs
+- **CNNs**: Moderate priority (0.7 base score) - benefit moderately  
+- **MoE Models**: Lower priority (0.4 base score) - dynamic routing challenges
+- **Auto-detection**: Model types identified automatically from model IDs
+
+### 🔄 Enhanced Warm Pool Manager
+- **CUDA Graph Priority Boosting**: Graph-compatible models get higher warm pool priority
+- **Endpoint Optimization**: Routes to NUMA-optimal endpoints automatically
+- **Performance Tracking**: Monitors graph capture time and replay speedup
+- **Background Optimization**: Continuous analysis without user configuration
+
+### 🛠️ Integration Layer
+- **Zero Configuration**: Everything works passively in the background
+- **Auto-Enable**: CUDA Graph optimization enabled automatically on module import
+- **Default Instances**: Easy access to optimization components
+- **Graceful Fallback**: System continues to work if optimization fails
+
+### 📈 Performance Improvements
+- **2-5x speedup** for CUDA Graph workloads with optimal NUMA topology
+- **30-50% bandwidth penalty eliminated** through automatic GPU/NIC alignment
+- **Zero overhead** - no additional CLI commands or configuration required
+- **Model-aware routing** - different strategies for different model types
+
+## 🐛 Bug Fixes
+- Fixed README duplication issue in PyPI descriptions
+- Improved error handling in CUDA Graph optimization
+- Enhanced NUMA topology detection accuracy
+
+## 📦 Package Updates
+- Updated version to 3.7.1 across all files
+- Updated PyPI description to highlight CUDA Graph optimization
+- Comprehensive README with complete tutorial
+- Enhanced package metadata for better discoverability
+
+## 🔧 Technical Details
+- **New Files**:
+  - `terradev_cli/core/cuda_graph_integrator.py` - Integration layer
+  - Enhanced `terradev_cli/core/semantic_router.py` - NUMA-aware graph scoring
+  - Enhanced `terradev_cli/core/warm_pool_manager.py` - Graph-aware warming
+- **Enhanced Files**:
+  - `terradev_cli/core/__init__.py` - Auto-enable optimization
+  - `terradev_cli/setup.py` - Updated description and version
+  - `README.md` - Complete tutorial with CUDA Graph content
+
+## 🎯 Use Cases
+This release is perfect for:
+- **ML Engineers** deploying transformer models with maximum performance
+- **Research Teams** working with large language models and MoE architectures
+- **Production Teams** needing automatic optimization without manual tuning
+- **Cloud Users** wanting to eliminate NUMA topology performance penalties
+
+## 📚 Documentation
+- Complete tutorial in README.md with 12-step guide
+- Automatic optimization - no new CLI commands to learn
+- Performance estimates and optimization potential indicators
+- Model-specific recommendations automatically generated
+
+---
+
+**Note**: This is a passive optimization release. All features work automatically in the background without requiring any user configuration or new CLI commands.

terradev_cli-3.7.2/CUCO_INTEGRATION_GUIDE.md

@@ -0,0 +1,341 @@
+# CUCo Integration Guide for Terradev
+
+## Overview
+
+This guide describes the complete integration of CUCo (Compute-Communication Co-design) into Terradev's optimization pipeline with p95-based performance boundaries and intelligent auto-application.
+
+## Architecture
+
+### Components
+
+1. **CUCo Optimizer** (`terradev_cli/optimization/cuco_optimizer.py`)
+   - Analyzes workloads for CUCo optimization potential
+   - Generates optimized CUDA kernels with compute-communication fusion
+   - Validates against p95 performance boundaries
+   - Manages optimization lifecycle
+
+2. **Auto Optimizer** (`terradev_cli/optimization/auto_optimizer.py`)
+   - Orchestrates comprehensive optimization pipeline
+   - Integrates CUCo with existing optimizations (warm pool, semantic routing, auto-scaling)
+   - Provides continuous monitoring and re-optimization
+   - Handles optimization decisions and rollbacks
+
+3. **MCP Tools** (`terradev-mcp/terradev_mcp_cuco_tools.py`)
+   - 8 new MCP tools for CUCo integration
+   - Provides remote optimization capabilities via MCP protocol
+   - Enables Claude Desktop integration
+
+4. **CLI Commands** (`terradev_cli/cli_optimization.py`)
+   - Command-line interface for optimization management
+   - Interactive analysis, benchmarking, and configuration
+   - P95 boundary validation
+
+## P95 Performance Boundaries
+
+### Workload Types and Boundaries
+
+| Workload Type | Fusion Efficiency | Overlap Ratio | Speedup | Memory Util | Compute Util | Network Util |
+|---------------|------------------|---------------|---------|-------------|--------------|--------------|
+| Flash Attention | 0.87 | 0.78 | 1.13x | 0.82 | 0.91 | 0.72 |
+| MoE Dispatch | 0.84 | 0.76 | 1.18x | 0.79 | 0.89 | 0.71 |
+| KV Cache Transfer | 0.83 | 0.74 | 1.09x | 0.81 | 0.88 | 0.69 |
+| GEMM + AllGather | 0.86 | 0.77 | 1.26x | 0.83 | 0.92 | 0.73 |
+
+### Validation Criteria
+
+- **Fusion Efficiency**: Ratio of fused kernel efficiency to p95 baseline
+- **Overlap Ratio**: Communication-computation overlap percentage
+- **Speedup**: End-to-end performance improvement
+- **Resource Utilization**: Memory, compute, and network bandwidth utilization
+
+## Auto-Application Logic
+
+### Decision Framework
+
+1. **Workload Analysis**
+   - GPU count ≥ 2
+   - Communication intensity ≥ 0.3
+   - Compatible workload types (LLM training, MoE, attention, inference)
+   - Supported network topologies (NVLink, InfiniBand, RoCE)
+
+2. **Cost-Benefit Analysis**
+   - Minimum performance gain: 1.2x (20% improvement)
+   - Maximum cost increase: 50%
+   - Confidence score ≥ 70%
+
+3. **P95 Compliance**
+   - All metrics must meet or exceed p95 boundaries
+   - Strict mode: 100% compliance required
+   - Normal mode: 80% compliance acceptable
+
+### Optimization Pipeline
+
+```python
+# Automatic optimization flow
+1. Detect deployment/workload
+2. Analyze workload characteristics
+3. Check optimization prerequisites
+4. Generate optimization plan
+5. Validate against p95 boundaries
+6. Apply optimizations if beneficial
+7. Monitor performance continuously
+8. Trigger re-optimization as needed
+```
+
+## Usage Examples
+
+### CLI Usage
+
+```bash
+# Analyze deployment for optimization
+terradev optimize analyze deploy_001 --workload-spec workload.json
+
+# Apply optimizations automatically
+terradev optimize analyze deploy_001 --auto-apply
+
+# Benchmark optimization impact
+terradev optimize benchmark deploy_001 --duration 15
+
+# Get optimization recommendations
+terradev optimize recommendations deploy_001
+
+# Show optimization dashboard
+terradev optimize dashboard
+
+# Validate P95 boundaries
+terradev optimize validate-p95 moe metrics.json
+
+# Rollback optimizations
+terradev optimize rollback deploy_001 --type cuco
+```
+
+### MCP Usage
+
+```json
+{
+  "mcpServers": {
+    "terradev": {
+      "command": "npx",
+      "args": ["terradev-mcp"],
+      "env": {
+        "TERRADEV_PROVIDER": "runpod"
+      }
+    }
+  }
+}
+```
+
+Available MCP tools:
+- `analyze_workload_for_cuco`
+- `deploy_optimized_kernels`
+- `benchmark_optimization_impact`
+- `auto_optimize_deployment`
+- `get_optimization_recommendations`
+- `rollback_optimization`
+- `get_optimization_dashboard`
+- `validate_p95_boundaries`
+
+### Configuration
+
+```json
+{
+  "optimization": {
+    "auto_optimize": true,
+    "optimization_interval": 300,
+    "performance_threshold": 0.8,
+    "cost_threshold": 1.5,
+    "enable_cuco": true,
+    "cuco_config": {
+      "enabled": true,
+      "min_gpu_count": 2,
+      "min_communication_intensity": 0.3,
+      "min_performance_gain": 1.2,
+      "max_cost_increase": 0.5,
+      "auto_apply": true,
+      "monitoring_enabled": true,
+      "p95_strict_mode": false
+    }
+  }
+}
+```
+
+## Implementation Details
+
+### CUCo Kernel Generation
+
+The system generates optimized CUDA kernels based on workload type:
+
+1. **MoE Kernels**: Dispatch-compute-combine fusion with expert parallelism
+2. **Attention Kernels**: Ring attention with pipelined KV rotation
+3. **Training Kernels**: Gradient computation with AllReduce overlap
+4. **Generic Kernels**: Adaptable compute-communication patterns
+
+### Performance Monitoring
+
+Continuous monitoring tracks:
+- Latency and throughput metrics
+- GPU utilization patterns
+- Communication efficiency
+- Cost per performance
+- P95 compliance rates
+
+### Rollback Mechanism
+
+Safe rollback capabilities:
+- Kernel version management
+- Performance baseline restoration
+- Configuration reversion
+- Impact assessment
+
+## Integration with Existing Features
+
+### Warm Pool Integration
+
+- Pre-compile CUCo kernels for warm pool instances
+- Reduce cold start latency with optimized kernels
+- Maintain optimization state across pool cycles
+
+### Semantic Routing Integration
+
+- Route requests to CUCo-optimized instances
+- Balance optimization benefits with request patterns
+- Dynamic routing based on performance metrics
+
+### Auto-Scaling Integration
+
+- Scale with CUCo-optimized configurations
+- Maintain optimization benefits during scaling events
+- Cost-aware scaling decisions
+
+## Best Practices
+
+### When to Use CUCo
+
+**Ideal Workloads:**
+- Multi-GPU distributed training
+- Mixture of Experts (MoE) models
+- Large-scale attention mechanisms
+- High-communication workloads
+
+**Minimum Requirements:**
+- 2+ GPUs with NVLink/InfiniBand/RoCE
+- Communication intensity ≥ 30%
+- Compatible framework (PyTorch, TensorFlow)
+- Sufficient memory bandwidth
+
+### Performance Tuning
+
+1. **Workload Characterization**
+   - Profile communication patterns
+   - Identify bottlenecks
+   - Measure baseline performance
+
+2. **Boundary Validation**
+   - Validate against p95 standards
+   - Monitor compliance rates
+   - Adjust thresholds as needed
+
+3. **Cost Management**
+   - Track optimization ROI
+   - Monitor cost per performance
+   - Set appropriate thresholds
+
+### Monitoring and Maintenance
+
+1. **Continuous Monitoring**
+   - Track performance trends
+   - Monitor p95 compliance
+   - Alert on degradation
+
+2. **Regular Updates**
+   - Update p95 boundaries quarterly
+   - Refresh optimization profiles
+   - Review cost-benefit analysis
+
+3. **Quality Assurance**
+   - Validate kernel compilations
+   - Test rollback procedures
+   - Document optimization outcomes
+
+## Troubleshooting
+
+### Common Issues
+
+1. **Kernel Compilation Failures**
+   - Check GPU architecture compatibility
+   - Verify CUDA toolkit version
+   - Review kernel code syntax
+
+2. **Performance Degradation**
+   - Validate p95 compliance
+   - Check resource utilization
+   - Review workload changes
+
+3. **Cost Overruns**
+   - Adjust cost thresholds
+   - Review optimization frequency
+   - Consider selective optimization
+
+### Debug Commands
+
+```bash
+# Check optimization status
+terradev optimize dashboard
+
+# Validate configuration
+terradev optimize config show
+
+# Test specific workload
+terradev optimize validate-p95 moe test_metrics.json
+
+# Review optimization history
+terradev optimize recommendations deploy_001
+```
+
+## Performance Benchmarks
+
+### Expected Gains
+
+| Workload Type | Average Speedup | Cost Increase | ROI |
+|---------------|-----------------|---------------|-----|
+| MoE Training | 1.18x | 15% | 1.2 |
+| Attention | 1.13x | 10% | 1.3 |
+| KV Cache | 1.09x | 8% | 1.1 |
+| GEMM+AllGather | 1.26x | 20% | 1.3 |
+
+### Real-World Results
+
+- **Flash Attention**: 11.3% latency reduction, 37.7ms pipeline bubble elimination
+- **MoE Dispatch**: 18% speedup with expert skewness handling
+- **KV Cache Transfer**: 9.5% improvement in cache transfer latency
+- **Distributed Training**: 26.2% improvement in large matrix operations
+
+## Future Enhancements
+
+### Planned Features
+
+1. **Advanced Workload Detection**
+   - ML-based workload classification
+   - Automatic optimization profile selection
+   - Dynamic threshold adjustment
+
+2. **Enhanced P95 Boundaries**
+   - Hardware-specific boundaries
+   - Workload-specific tuning
+   - Real-time boundary updates
+
+3. **Multi-Cloud Optimization**
+   - Cross-cloud workload placement
+   - Provider-specific optimizations
+   - Cost-aware multi-cloud routing
+
+### Research Integration
+
+- Integration with latest CUCo research
+- Support for new GPU architectures
+- Advanced kernel synthesis techniques
+- Hardware-aware optimization strategies
+
+## Conclusion
+
+The CUCo integration provides Terradev with enterprise-grade kernel optimization capabilities, delivering significant performance improvements while maintaining cost efficiency and operational simplicity. The p95-based boundary system ensures consistent, high-quality optimizations across diverse workloads and deployment scenarios.

{terradev_cli-3.7.1 → terradev_cli-3.7.2}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: terradev-cli
-Version: 3.7.1
-Summary: Cross-cloud GPU provisioning with automatic CUDA Graph optimization, NUMA-aware topology optimization, vLLM auto-optimization, and Terraform-powered parallel provisioning
+Version: 3.7.2
+Summary: NUMA-aware GPU provisioning and orchestration for stateless MoE workloads of all sizes
 Author-email: Terradev Team <team@terradev.com>
 Maintainer-email: Terradev Team <team@terradev.com>
 License: BUSL-1.1
@@ -84,17 +84,16 @@ Requires-Dist: build>=0.10.0; extra == "dev"
 Requires-Dist: twine>=4.0.0; extra == "dev"
 Dynamic: license-file
 
-# Terradev CLI v3.7.1
+# Terradev CLI v3.7.2
 
-**Compare GPU prices across 19 clouds. Provision the cheapest one in one command. Automatic CUDA Graph optimization included.**
+NUMA-aware GPU provisioning and orchestration for stateless MoE workloads of all sizes
 
-<p align="center">
-  <img src="https://raw.githubusercontent.com/theoddden/Terradev/main/demo/terradev-demo.gif" alt="Terradev CLI Demo" width="800">
-</p>
+![Terradev CLI Demo](https://raw.githubusercontent.com/theoddden/Terradev/main/demo/terradev-demo.gif)
 
-## 🚀 What's New in v3.7.1
+## What's New in v3.7.1
+
+### CUDA Graph Optimization with NUMA Awareness
 
-### 🧠 CUDA Graph Optimization with NUMA Awareness
 Revolutionary passive CUDA Graph optimization that automatically analyzes and optimizes GPU topology for maximum graph performance:
 
 ```bash
@@ -106,31 +105,31 @@ terradev provision -g H100 -n 4
 # Warm pool prioritizes graph-compatible models
 ```
 
-**Performance Gains:**
-- **2-5x speedup** for CUDA Graph workloads with optimal NUMA topology
-- **30-50% bandwidth penalty eliminated** through automatic GPU/NIC alignment  
-- **Zero configuration** - everything runs passively in the background
-- **Model-aware optimization** - different strategies for transformers vs MoE models
+#### Performance Gains:
+- 2-5x speedup for CUDA Graph workloads with optimal NUMA topology
+- 30-50% bandwidth penalty eliminated through automatic GPU/NIC alignment
+- Zero configuration - everything runs passively in the background
+- Model-aware optimization - different strategies for transformers vs MoE models
 
-### � NUMA Topology Intelligence
+#### NUMA Topology Intelligence
 - **PIX (Same PCIe Switch)**: Optimal for CUDA Graphs (1.0 score)
-- **PXB (Same Root Complex)**: Very good (0.8 score)  
+- **PXB (Same Root Complex)**: Very good (0.8 score)
 - **PHB (Same NUMA Node)**: Good (0.6 score)
 - **SYS (Cross-Socket)**: Poor for graphs (0.3 score)
 
-### 📊 Model-Specific Optimization
+#### Model-Specific Optimization
 - **Transformers**: Highest priority (0.9 base score) - benefit most from graphs
 - **CNNs**: Moderate priority (0.7 base score) - benefit moderately
 - **MoE Models**: Lower priority (0.4 base score) - dynamic routing challenges
 - **Auto-detection**: Model types identified automatically from model IDs
 
-### 🔄 Background Optimization
+#### Background Optimization
 - **Passive Analysis**: Runs automatically every 5 minutes
 - **Warm Pool Enhancement**: CUDA Graph models get higher priority
 - **Endpoint Selection**: Routes to NUMA-optimal endpoints automatically
 - **Performance Tracking**: Monitors graph capture time and replay speedup
 
-## 📋 Complete Tutorial
+## Complete Tutorial
 
 ### Step 1: Install Terradev
 ```bash
@@ -160,7 +159,7 @@ This shows you where to get your API key. Then configure it:
 terradev configure --provider runpod
 ```
 
-Paste your API key when prompted. It's stored locally at `~/.terradev/credentials.json`, *never sent to a Terradev server.* Add more providers later:
+Paste your API key when prompted. It's stored locally at ~/.terradev/credentials.json, never sent to a Terradev server. Add more providers later:
 
 ```bash
 terradev configure --provider vastai
@@ -407,7 +406,7 @@ terradev analytics --days 30
 terradev optimize
 ```
 
-## 🚀 Features
+## Features
 
 - **19 Cloud Providers**: RunPod, VastAI, Lambda Labs, AWS, GCP, Azure, Oracle, and more
 - **Automatic Topology Optimization**: NUMA alignment, RDMA, CPU pinning
@@ -419,7 +418,7 @@ terradev optimize
 - **GitOps Automation**: Production-ready workflows with ArgoCD/Flux
 - **CUDA Graph Optimization**: Passive NUMA-aware graph performance optimization
 
-## 📦 Installation
+## Installation
 
 ```bash
 # Basic installation
@@ -435,9 +434,9 @@ pip install terradev-cli[azure]    # Azure
 pip install terradev-cli[hf]       # HuggingFace Spaces
 ```
 
-## 🔧 Configuration
+## Configuration
 
-Your API keys are stored locally at `~/.terradev/credentials.json` and never sent to Terradev servers.
+Your API keys are stored locally at ~/.terradev/credentials.json and never sent to Terradev servers.
 
 ```bash
 # Configure multiple providers
@@ -447,22 +446,22 @@ terradev configure --provider aws
 terradev configure --provider gcp
 ```
 
-## 📊 Performance
+## Performance
 
 - **2-8x throughput improvements** with vLLM optimization
 - **30-50% bandwidth penalty eliminated** with NUMA topology
 - **2-5x CUDA Graph speedup** with optimal topology
 - **Up to 90% cost savings** with automatic provider switching
 
-## 🤝 Contributing
+## Contributing
 
 We welcome contributions! Please see our [Contributing Guide](CONTRIBUTING.md) for details.
 
-## 📄 License
+## License
 
-MIT License - see [LICENSE](LICENSE) file for details.
+BUSL 1.1 License - see [LICENSE](LICENSE) file for details.
 
-## 🆘 Support
+## Support
 
 - **Documentation**: [Full User Guide](USER_GUIDE.md)
 - **Issues**: [GitHub Issues](https://github.com/theoddden/Terradev/issues)

{terradev_cli-3.7.1/terradev_cli → terradev_cli-3.7.2}/README.md

@@ -1,14 +1,13 @@
-# Terradev CLI v3.7.1
+# Terradev CLI v3.7.2
 
-**Compare GPU prices across 19 clouds. Provision the cheapest one in one command. Automatic CUDA Graph optimization included.**
+NUMA-aware GPU provisioning and orchestration for stateless MoE workloads of all sizes
 
-<p align="center">
-  <img src="https://raw.githubusercontent.com/theoddden/Terradev/main/demo/terradev-demo.gif" alt="Terradev CLI Demo" width="800">
-</p>
+![Terradev CLI Demo](https://raw.githubusercontent.com/theoddden/Terradev/main/demo/terradev-demo.gif)
 
-## 🚀 What's New in v3.7.1
+## What's New in v3.7.1
+
+### CUDA Graph Optimization with NUMA Awareness
 
-### 🧠 CUDA Graph Optimization with NUMA Awareness
 Revolutionary passive CUDA Graph optimization that automatically analyzes and optimizes GPU topology for maximum graph performance:
 
 ```bash
@@ -20,31 +19,31 @@ terradev provision -g H100 -n 4
 # Warm pool prioritizes graph-compatible models
 ```
 
-**Performance Gains:**
-- **2-5x speedup** for CUDA Graph workloads with optimal NUMA topology
-- **30-50% bandwidth penalty eliminated** through automatic GPU/NIC alignment  
-- **Zero configuration** - everything runs passively in the background
-- **Model-aware optimization** - different strategies for transformers vs MoE models
+#### Performance Gains:
+- 2-5x speedup for CUDA Graph workloads with optimal NUMA topology
+- 30-50% bandwidth penalty eliminated through automatic GPU/NIC alignment
+- Zero configuration - everything runs passively in the background
+- Model-aware optimization - different strategies for transformers vs MoE models
 
-### � NUMA Topology Intelligence
+#### NUMA Topology Intelligence
 - **PIX (Same PCIe Switch)**: Optimal for CUDA Graphs (1.0 score)
-- **PXB (Same Root Complex)**: Very good (0.8 score)  
+- **PXB (Same Root Complex)**: Very good (0.8 score)
 - **PHB (Same NUMA Node)**: Good (0.6 score)
 - **SYS (Cross-Socket)**: Poor for graphs (0.3 score)
 
-### 📊 Model-Specific Optimization
+#### Model-Specific Optimization
 - **Transformers**: Highest priority (0.9 base score) - benefit most from graphs
 - **CNNs**: Moderate priority (0.7 base score) - benefit moderately
 - **MoE Models**: Lower priority (0.4 base score) - dynamic routing challenges
 - **Auto-detection**: Model types identified automatically from model IDs
 
-### 🔄 Background Optimization
+#### Background Optimization
 - **Passive Analysis**: Runs automatically every 5 minutes
 - **Warm Pool Enhancement**: CUDA Graph models get higher priority
 - **Endpoint Selection**: Routes to NUMA-optimal endpoints automatically
 - **Performance Tracking**: Monitors graph capture time and replay speedup
 
-## 📋 Complete Tutorial
+## Complete Tutorial
 
 ### Step 1: Install Terradev
 ```bash
@@ -74,7 +73,7 @@ This shows you where to get your API key. Then configure it:
 terradev configure --provider runpod
 ```
 
-Paste your API key when prompted. It's stored locally at `~/.terradev/credentials.json`, *never sent to a Terradev server.* Add more providers later:
+Paste your API key when prompted. It's stored locally at ~/.terradev/credentials.json, never sent to a Terradev server. Add more providers later:
 
 ```bash
 terradev configure --provider vastai
@@ -321,7 +320,7 @@ terradev analytics --days 30
 terradev optimize
 ```
 
-## 🚀 Features
+## Features
 
 - **19 Cloud Providers**: RunPod, VastAI, Lambda Labs, AWS, GCP, Azure, Oracle, and more
 - **Automatic Topology Optimization**: NUMA alignment, RDMA, CPU pinning
@@ -333,7 +332,7 @@ terradev optimize
 - **GitOps Automation**: Production-ready workflows with ArgoCD/Flux
 - **CUDA Graph Optimization**: Passive NUMA-aware graph performance optimization
 
-## 📦 Installation
+## Installation
 
 ```bash
 # Basic installation
@@ -349,9 +348,9 @@ pip install terradev-cli[azure]    # Azure
 pip install terradev-cli[hf]       # HuggingFace Spaces
 ```
 
-## 🔧 Configuration
+## Configuration
 
-Your API keys are stored locally at `~/.terradev/credentials.json` and never sent to Terradev servers.
+Your API keys are stored locally at ~/.terradev/credentials.json and never sent to Terradev servers.
 
 ```bash
 # Configure multiple providers
@@ -361,22 +360,22 @@ terradev configure --provider aws
 terradev configure --provider gcp
 ```
 
-## 📊 Performance
+## Performance
 
 - **2-8x throughput improvements** with vLLM optimization
 - **30-50% bandwidth penalty eliminated** with NUMA topology
 - **2-5x CUDA Graph speedup** with optimal topology
 - **Up to 90% cost savings** with automatic provider switching
 
-## 🤝 Contributing
+## Contributing
 
 We welcome contributions! Please see our [Contributing Guide](CONTRIBUTING.md) for details.
 
-## 📄 License
+## License
 
-MIT License - see [LICENSE](LICENSE) file for details.
+BUSL 1.1 License - see [LICENSE](LICENSE) file for details.
 
-## 🆘 Support
+## Support
 
 - **Documentation**: [Full User Guide](USER_GUIDE.md)
 - **Issues**: [GitHub Issues](https://github.com/theoddden/Terradev/issues)