AutoGPT

Context #

  • AutoGPT, developed by a gaming company Significant Gravitas, is a platform built by to create, deploy, and manage continuous AI agents that automate complex workflows.
  • Initially supported only OpenAI’s GPT 3 & 4 models, now it supports multiple LLM Providers: Anthropic, Groq, Llama
  • It has a hybrid licensing model - mixed of MIT and Polyform Shield License - so not truly an open source.
  • GitHub Repo

Platform Architecture #

  • Primarily, have AutoGPT Server and AutoGPT Frontend
  • Platform Components:
    • Agents and Workflows: customized workflows to build agents for various tasks such as data processing, task management, integration, etc.
    • Blocks as Integrations: connecting with external services, data processing, AI model invocations, conditional logic and custom scripts, and more.
  • AutoGPT employs the agent protocol standard by the AI Engineer Foundation. This standardizes the communication pathways from your agent to the frontend and benchmark.
  • Built Benchmark for measuring agent’s performnance.
  • Technologies: Supabase/PostgreSQL as Vector database, Kong as gateway, Redis for Caching, Next.js for Frontend, Docker for Containers

Local Setup #

Running backend (read instructions here):

git clone https://github.com/Significant-Gravitas/AutoGPT.git
cd AutoGPT
git submodule update --init --recursive
cd autogpt_platform
cp supabase/docker/.env.example .env
docker compose up -d --build

Running frontend:

cd sandbox/genai/AutoGPT/autogpt_platform/frontend
cp .env.example .env
nvm use 18.17.0
npm install
npm run dev