tahp0604/multi-agent-debate-defense

edited model

edited model

made optimizations

made optimizations

edited model

added model

Fix Hugging Face Space metadata colorTo value

Update RAG pipeline, config, and requirements; remove legacy IPI detector models

merged user edits with LLM warmup logic

fixed prompt formatting and agent refusals

added llama-3.2-1b and ram optimizations

added model warmup to prevent timeouts

added knowledge base document list to UI

fixed phi-3 bug and attention implementation

enabled data and model cache persistence in HF Bucket

switched prosecutor to qwen 2.5 3b

fixed import error for local_model_id

decoupled prosecutor and defender models

diversified agents with llama 3.2 3b

added safety disclaimers to prevent agent refusals

added local inference

added local inference

Increase gunicorn timeout for long debate rounds

Fix clear_index crash on empty collection

Add missing flask dependency

Initial commit

initial commit