Restructure architecture from 10 to 8 components

Merge Simulation & Exploration Layer into Conversational Interface Layer
since specification editing and what-if analysis are UI-driven features.
Reposition vLLM Simulator as a development tool rather than a core
architecture component.

Changes:
- Update component count from 10 to 8 across all docs
- Remove component numbering; reference by name for flexibility
- Add Phase 1 note to DeploymentIntent schema clarification
- Enhance Future Enhancements with comprehensive simulation features
- Update architecture diagrams to reflect UI-driven exploration

Files updated:
- docs/ARCHITECTURE.md
- docs/architecture-diagram.md
- CLAUDE.md
- README.md

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Andre Fredette <afredette@redhat.com>

Author: anfredette

CLAUDE.md

@@ -65,23 +65,26 @@ The system translates high-level user intent into technical specifications:
   - GPU capacity plan (e.g., "2x NVIDIA L4 GPUs, independent replicas")
   - Cost estimate ($800/month)
 
-### The 10 Core Components
+### The 8 Core Components
 
-1. **Conversational Interface Layer** - Streamlit UI
+1. **Conversational Interface Layer** - Streamlit UI with interactive exploration features
+   - Specification review and editing
+   - What-if analysis (Phase 2: advanced simulation)
 2. **Context & Intent Engine** - Extract structured specs from conversation
    - Use case → SLO template mapping
    - Auto-generate traffic profiles
 3. **Recommendation Engine** (3 sub-components):
    - Traffic Profile Generator
    - Model Recommendation Engine
    - Capacity Planning Engine
-4. **Simulation & Exploration Layer** - What-if analysis, spec editing
-5. **Deployment Automation Engine** - Generate YAML, deploy to K8s
-6. **Knowledge Base** - Benchmarks, SLO templates, model catalog, outcomes
-7. **LLM Backend** - Powers conversational AI (Ollama with llama3.1:8b)
-8. **Orchestration & Workflow Engine** - Coordinate multi-step flows
-9. **Inference Observability** - Monitor deployed models (TTFT, TPOT, GPU utilization)
-10. **vLLM Simulator** - GPU-free development and testing
+4. **Deployment Automation Engine** - Generate YAML, deploy to K8s
+5. **Knowledge Base** - Benchmarks, SLO templates, model catalog, outcomes
+6. **LLM Backend** - Powers conversational AI (Ollama with llama3.1:8b)
+7. **Orchestration & Workflow Engine** - Coordinate multi-step flows
+8. **Inference Observability** - Monitor deployed models (TTFT, TPOT, GPU utilization)
+
+**Development Tools:**
+- **vLLM Simulator** - GPU-free development and testing (not part of core architecture)
 
 ### Critical Data Collections (Knowledge Base)
 - **Model Benchmarks**: TTFT/TPOT/throughput for (model, GPU, tensor_parallel) tuples
@@ -96,7 +99,7 @@ The system translates high-level user intent into technical specifications:
 **docs/ARCHITECTURE.md and docs/architecture-diagram.md must stay synchronized**:
 - If you change component descriptions in ARCHITECTURE.md, update architecture-diagram.md diagrams
 - If you add/remove components, update both files
-- Component numbering must match (e.g., "Component 3" in both docs)
+- Components are referenced by name (not numbered) for clarity and flexibility
 
 ### Key Architectural Decisions to Preserve
 

README.md

@@ -88,18 +88,19 @@ The POC includes 3 pre-configured scenarios (see [data/demo_scenarios.json](data
 
 ## Architecture Highlights
 
-Compass implements a **10-component architecture** with:
+Compass implements an **8-component architecture** with:
 
-- **Conversational Interface** (Streamlit) - Chat-based requirement gathering
+- **Conversational Interface** (Streamlit) - Chat-based requirement gathering with interactive exploration
 - **Context & Intent Engine** - LLM-powered extraction of deployment specs
 - **Recommendation Engine** - Traffic profiling, model scoring, capacity planning
-- **Simulation & Exploration** - What-if analysis and spec editing
 - **Deployment Automation** - YAML generation and Kubernetes deployment
 - **Knowledge Base** - Benchmarks, SLO templates, model catalog
 - **LLM Backend** - Ollama (llama3.1:8b) for conversational AI
 - **Orchestration** - Multi-step workflow coordination
 - **Inference Observability** - Real-time deployment monitoring
-- **vLLM Simulator** - GPU-free local development
+
+**Development Tools:**
+- **vLLM Simulator** - GPU-free local development and testing
 
 See [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md) for detailed system design.