fix: Final fixes for demo

feat: Improved UI of streanlit

feat: eval dashboard complete, Plotly charts showing score trends and model comparison

feat: external repo scanning working, 4 targets scanned with real findings documented

feat: MCP server working, Claude Desktop successfully querying VulnGraph attack graph

feat: agentic pipeline complete, 5-step deterministic workflow with RAG-enriched KB queries

feat: eval harness complete, llama3.2:3b selected over phi3/qwen2.5/gemma2 based on eval scores

feat: Pydantic schema validation on all scanner ingestion, 14 Bandit + 3 Trivy findings validated

feat: FastAPI backend complete, all endpoints working including health, stats, findings, graph

feat: complete RAG-augmented LLM pipeline, all findings explained including real CVEs

Added LLM Explanations