ai-shell

https://gb-docs.snaily.top/
Gemini-Balance
Docker compose

services:
  gemini-balance:
    image: ghcr.io/snailyp/gemini-balance:latest
    container_name: gemini-balance
    restart: unless-stopped
    ports:
      - "9001:8000"
    volumes:
      - ~/docker/gemini-balance/data:/app/data
    env_file:
      - .env
    healthcheck:
      test: ["CMD-SHELL", "python -c \"import requests; exit(0) if requests.get('http://localhost:8000/health').status_code == 200 else exit(1)\""]
      interval: 1800s
      timeout: 5s
      retries: 1
      start_period: 20s

Content of .env file

DATABASE_TYPE=sqlite
SQLITE_DATABASE=default_db
API_KEYS=[""]
ALLOWED_TOKENS=["sk-123456"]

• this is the MVP config
  OpenAI

http://10.10.120.12:9001/v1/chat/completions

Python
https://github.com/mceck/shy-sh

python3 -m venv venv
. venv/bin/activate
pip install shy-sh

Configuration in ~/.config/shy/config.yaml, similar all chat history are stored here

language: ''
llm:
  agent_pattern: flow
  api_key: sk-123456
  base_url: http://10.10.120.12:9001/v1
  name: gemini-2.5-flash-lite
  provider: openai
  temperature: 0.0
sandbox_mode: false

• the v1 is needed to add the the path to be compatible with gemini-balance
  To execute

~/venv/bin/shy 

or add to .bash_aliases

alias shy="~/venv/bin/shy"

Pure Bash
https://github.com/Hezkore/bash-ai

curl -sS https://raw.githubusercontent.com/hezkore/bash-ai/main/install.sh | bash

The config is located at ~/.config/bai.cfg

key=sk-
api=http://10.10.120.12:9001/v1/chat/completions
model=gemini-2.5-flash-lite