Building an AI Devil's Advocate: A Multi-Agent Debate App with LangGraph & Streamlit

Description and use

The AI Devil's Advocate app showcases a resilient multi‑agent debate system that automatically switches LLM providers when rate limits are hit, maintaining continuous operation and offering transcript export.

Problem

Managing API rate limits while running a multi-agent debate application.

Outcome

The app demonstrates a resilient multi‑agent debate system that automatically switches LLM providers on rate‑limit errors, maintaining continuous operation and providing a polished UI with transcript export and usage monitoring.

Workflow

• Configure multiple LLM providers in llm_factory.py
• Wrap providers with FallbackLLMWrapper
• Run debate agents via LangGraph state machine
• If a 429 error occurs, wrapper routes prompt to next provider
• Continue debate until verdict is generated
• Export transcript as PDF or Markdown
• Display real‑time API call counter in sidebar

Evidence

If the primary LLM throws a 429 Too Many Requests error, the wrapper catches it and immediately routes the prompt to the next available provider in the list.

Source

Multi-Provider Architecture: Switch between Groq, Gemini, and OpenRouter on the fly.

Source