Building a Multi-Agent Debate Platform with LangGraph

Why LangGraph?

Among the many AI frameworks, what attracted me most about LangGraph is its StateGraph mechanism. Compared to simple Chains, StateGraph provides more flexible control flow, making it ideal for scenarios requiring multiple agents to interact.

Architecture Design

Role Design

class DebateRole(Enum):
    MODERATOR = "moderator"
    PRO = "pro"
    CON = "con"

State Design

class DebateState(TypedDict):
    topic: str
    messages: list[Message]
    current_speaker: DebateRole
    round: int

Lessons Learned

1. Message Ordering Issues

Initially, I didn’t handle message ordering correctly, leading to incoherent agent responses.

2. Token Limits

As debates progressed, historical messages became too long and exceeded token limits. The solution was implementing a summarization mechanism.

Results

DebateAI can now conduct complete debate processes, including opening, interactive debate, and conclusion phases.

Next Steps

1. Add more role types
2. Support multilingual debates
3. Add scoring mechanism