AI Local

AI Local contains two Python clients for experimenting with Gemini desktop chat and OpenAI-compatible local inference endpoints.

Features

• Runs a Tkinter Gemini chat interface with local JSON conversation history.
• Creates, switches, and deletes local chat history files from the desktop UI.
• Streams responses from an OpenAI-compatible local server through a small CLI client.
• Supports configurable local server URL, API key placeholder, model name, and sample prompt through environment variables.

Requirements

• Python 3.10 or newer.
• A Google API key for the Gemini desktop chat client.
• An OpenAI-compatible local inference server for localAI.py.
• Python packages listed in requirements.txt: google-generativeai, openai, python-dotenv, and Pillow.
• On Linux, a Tkinter-capable Python installation is required for the desktop GUI.

Setup

1. Create and activate a virtual environment.
2. Install dependencies:

pip install -r requirements.txt

3. Copy .env.example to a local .env file.
4. Configure the values needed for the client you want to run:

OPENAI_BASE_URL=http://127.0.0.1:8000/v1
OPENAI_API_KEY=EMPTY
OPENAI_MODEL=your-local-model-name
GOOGLE_API_KEY=your-google-api-key

localAI.py also supports an optional OPENAI_SAMPLE_PROMPT value. If it is not set, the script uses its built-in sample prompt.

Before running ai_main.py, create a history folder in this repository if it does not already exist. The UI expects at least one JSON file in that folder when it starts.

Usage

Run the Gemini desktop chat application:

python ai_main.py

Run the OpenAI-compatible local server client:

python localAI.py

Project Structure

• ai_main.py: Tkinter Gemini chat client with local history file management.
• localAI.py: streaming CLI request against an OpenAI-compatible chat completions endpoint.
• requirements.txt: Python runtime dependencies.
• .env.example: sample configuration for Gemini and OpenAI-compatible local inference settings.
• .gitignore: excludes local environment files, virtual environments, cache files, and generated history outputs.

摘要

AI Local 收錄兩個本機 AI 實驗用 Python 程式。 ai_main.py 是使用 Gemini 的 Tkinter 桌面聊天介面，會把對話紀錄存成 JSON 檔。 localAI.py 是連接 OpenAI 相容本機推論服務的命令列範例，並以串流方式輸出回覆。 使用前需安裝 requirements.txt，並依 .env.example 建立本機環境檔。 若要啟動桌面聊天介面，執行 python ai_main.py；若要測試本機推論服務，執行 python localAI.py。 桌面聊天介面需要 GOOGLE_API_KEY，本機推論範例需要 OPENAI_BASE_URL、OPENAI_API_KEY 與 OPENAI_MODEL。