Roshan10000/MFDE

Fix: Added missing main() function and entry point to app.py and server/app.py for multi-mode deployment compliance

Final synchronization: server/app.py matched to root app.py

Final cleanup: Updated openenv.yaml and synchronized app.py for HF submission

Integrated HF_TOKEN across all components and switched default AI backend to Hugging Face Inference API

Revert: Remove 'safe' priority tier to comply with hackathon spec [low, medium, high]

Upgrade to 4-tier risk classification system (High, Medium, Low, Safe) for OpenEnv compliance

feat: implement advanced threat detection scanner with 4 risk tiers

feat: implement Semantic Noise Engine and Threat Matrix Analytics dashboard

style: revert decimal precision to 2 places for cleaner UI

feat: calibrate scoring compliance and enhance dashboard precision

fix: implement strict (0.0, 1.0) grading interval with 0.05-0.95 rewards and 0.02-0.98 overall score clamping

fix: restore standard 0.0-1.0 scoring range and revert task graders to deterministic in openenv.yaml while retaining explicit grader logic in grader.py

fix: finalize scoring recalibration to 0.00-0.99 range and implement per-task explicit graders

feat: implement per-task grader mapping and heuristic grading for Gmail in openenv.yaml

fix: explicitly define deterministic graders for all simulation tasks to satisfy Phase 2 validation

fix: complete inference.py rewrite following mandatory OpenEnv stdout format and proxy guidelines exactly

fix: critical - remove Dockerfile API_BASE_URL override that silently blocked OpenEnv proxy injection; split MFDE_SERVER_URL from LLM proxy API_BASE_URL

fix: fully decouple OpenEnv injected API_BASE_URL from internal server ENV_BASE_URL to prevent silent API bypass, and explicitly enforce openai client proxy

fix: strictly implement OpenEnv validator proxy guidelines without fallbacks

chore: revert inference.py to use original Anthropic models and API base as requested by user

fix: ensure Anthropic fallback accurately inherits the original MODEL_NAME environment variable

feat: implement dual-system AI triage to safely fallback to Anthropic if OpenEnv OpenAI proxy is not locally injected

fix: completely remove heuristic fallback and enforce LLM API usage via openai client in ai_triage

fix: rewire inference.py to route through OpenEnv liteLLM proxy using API_BASE_URL and OpenAI client

fix: add missing main() block to server/app.py for OpenEnv multi-mode deployment validation

fix: change Yield Pos UI to display 2 decimal places to capture exact 0.99/0.01 bounds

fix: add missing HF Spaces YAML configuration to README

feat: complete TriageCenter Neural Defense Grid UI rewrite ensuring OpenEnv compliance

Fix: step auto-reset, error handling, env edge cases

Fix: port 7860, requirements, env.py edge cases

Upgrade to v2.0: Gmail integration and AI triage

Fix total combined XP exact precision

Fix UI yield rounding bug

Supercharge HTML Dashboard UI & Analytics Leaderboard

Fix environment score bounds and indexing bug