AIdea-server provides a backend for an AI application, enabling users to interact with various large language models (LLMs) like GPT, Tongyi Qianwen, and Wenxin Yiyan, as well as image generation models such as Stable Diffusion. It targets developers and users seeking a unified platform for AI-powered chat and image creation, offering a self-hostable solution.

How It Works

The server is developed in Golang and utilizes a custom framework called Glacier Framework, which features dependency injection and modularity. It abstracts AI model interactions into a consistent interface, notably supporting OpenAI-compatible APIs for chat models. Image generation capabilities are integrated through Stable Diffusion, with support for various modes like text-to-image and image-to-image. The architecture separates concerns into distinct directories for APIs, core logic, data models, and external service integrations.

Quick Start & Requirements

• Install/Run: go build -o build/debug/aidea-server cmd/main.go (for direct compilation) or via Docker.
• Prerequisites: Go (for compilation), Docker (for containerized deployment), MySQL, Redis, Nginx (recommended for production).
• Setup: Requires configuration of config.yaml with database credentials, API keys for AI services, and potentially cloud storage.
• Links: Client, Docker, Deployment Guide

Highlighted Details

• Supports multiple LLMs (GPT, Tongyi Qianwen, Wenxin Yiyan) and image models (Stable Diffusion, SDXL 1.0).
• Offers OpenAI-compatible APIs for seamless integration with existing tools.
• Includes features for digital humans, image upscaling, and colorization.
• Built with custom Golang frameworks (Glacier Framework, Eloquent ORM) for modularity and database interaction.

Maintenance & Community

• Actively maintained by mylxsw.
• Community discussion via WeChat groups (contact x-prometheus for invites).

Licensing & Compatibility

• License: MIT.
• Permissive license allows for commercial use and integration with closed-source applications.

Limitations & Caveats

The project's README notes that code comments and technical documentation are currently limited, with plans to supplement them in the future. Some internal naming conventions (e.g., "Room," "Advisor Group" referring to "Digital Human") reflect past iterations. The v1 "Creation Island" feature is deprecated in favor of v2.