πŸš€ GhostHarness Live Web GUI Demo β€” Running in client simulation mode for GitHub Pages
SOVEREIGN GhostHarness ● llama-server :8080 (Qwen 2.5 Coder 7B)
VRAM
4.3 / 6.0 GB
RAM
12.4 / 32.0 GB
OPTIMAL: ZERO-SPILL
Try interactive demo prompts:
⚑ GhostHarness Initialized (Local Sovereign Lab)

Connected to abliterated Qwen 2.5 Coder 7B Instruct (Q4_K_M) running 100% in 6GB VRAM.

πŸ›‘οΈ Zero Admin Privileges: Runs completely in user-space on campus PCs.
⚑ Zero-Spill Guard: 4.3 GB weights + 850 MB KV cache = 0% spill into host RAM.
πŸ—ΊοΈ AST Semantic RepoMap: Symbol hierarchy compressed to token budget.
β†Ί Instant Rollback: Atomic snapshots recorded before every disk write.

Click any prompt chip above or type a command below to preview autonomous agentic tool execution.