XinHai

XinHai: Sea of Minds Framework for Multimodal Multi-agent Simulation and Evolution

Quick Start

Docker Compose

Build local docker images in docker directory of both backend and frontend.

docker build --add-host raw.githubusercontent.com:185.199.108.133 --build-context root=../.. -t xinhai_backend .

docker build --add-host raw.githubusercontent.com:185.199.108.133 -t xinhai_frontend .

Modify env file

cp .env.example .env

Change docker-compose.yml accordingly.

Start docker compose in project root directory.

docker-compose up

Check status of docker images:

(base) vimos@vimos-Z270MX-Gaming5 XinHaiLLM  % docker-compose images
      Container          Repository    Tag       Image Id       Size
----------------------------------------------------------------------
xinhaillm_bridge_1       xinhai       latest   ad1a7389388c   17.56 GB
xinhaillm_controller_1   xinhai       latest   ad1a7389388c   17.56 GB
xinhaillm_frontend_1     xinhai       latest   ad1a7389388c   17.56 GB
xinhaillm_llm_1          xinhai       latest   ad1a7389388c   17.56 GB
xinhaillm_storage_1      xinhai       latest   ad1a7389388c   17.56 GB

If you need to stop the services, use docker-compose down.

Acknowledgment

XinHai learns a lot from the following repos:

• AgentVerse
• LLaMA-Factory
• MGM
• FlashRAG