HN Showcase
Back to browse
GitHub Repository

Run a council of local LLMs that debate, critique, and synthesize — no API keys needed.

74 starsPython

Run an Agent Council of LLMs that debate and synthesize answers

by JitseLambrichts·Mar 12, 2026·8 points·2 comments
Visit ProjectView on HN

AI Analysis

●●SolidBig BrainNiche Gem

Agent council debate architecture with GSM8K benchmarks showing accuracy gains.

Strengths
  • •Dual architectures: sequential Planning/Execute/Critique and parallel expert councils.
  • •Zero-config auto-discovery of Ollama and LM Studio endpoints, no .env setup.
  • •Assigns distinct models to different reasoning stages for optimized performance.
Weaknesses
  • •Multi-agent local LLM orchestration is becoming a crowded category.
  • •Accuracy gains on GSM8K may not translate to real-world reasoning tasks.
Category
AI/ML
Target Audience

Developers running local LLMs who want multi-agent reasoning

Similar To

LangGraph · AutoGen · CrewAI

Post Description

I built a local-first UI that adds two reasoning architectures on top of small models like Qwen, Llama and Mistral: a sequential Thinking Pipeline (Plan → Execute → Critique) and a parallel Agent Council where multiple expert models debate in parallel and a Judge synthesizes the best answer. No API keys, zero .env setup — just pip install multimind. Benchmark on GSM8K shows measurable accuracy gains vs. single-model inference.

Similar Projects

AI/ML●Mid

AI Council Debate

The five-role council (Analyst, Muse, Logician, Ethicist, Pragmatist) is a neat way to force diversity of perspective and makes for entertaining, shareable threads; live chat, voting and a verdict mechanic add community glue. It feels like a well-polished demo rather than a research advance — interesting and fun, but derivative of existing multi-agent/LLM playgrounds and likely limited by shallow or repetitive model outputs unless they invest in moderation, grounding, or stronger agent orchestration.

Crowd PleaserShip It
smakosh
314mo ago