Intelligent Stock Analysis System Based on LLM.This is an LLM-driven intelligent stock analysis tool that supports daily automatic analysis and push for A-shares, Hong Kong stocks, and US stocks. It obtains real-time market data from data sources like AkShare, Tushare, and YFinance, and uses large model API services such as DeepSeek to conduct multi-dimensional analysis (technical aspects, position distribution, public sentiment) on selected stocks, generating decision-making dashboards. It supports scheduled execution via GitHub Actions (no server required) or one-click deployment via Docker.

Building an AI Agent from Scratch.This project demonstrates how to construct an AI Agent tool similar to Claude Code from the ground up, consisting of 12 lessons. Each lesson comes with a runnable Python file. The content progresses from the most fundamental Agent loop, incrementally incorporating functions like tool invocation, task planning, sub-agents, context compression, multi-agent collaboration, and autonomous execution, ultimately building a comprehensive AI Agent system.

def agent_loop(messages):
    while True:
        response = client.messages.create(
            model=MODEL, system=SYSTEM,
            messages=messages, tools=TOOLS,
        )
        messages.append({"role": "assistant",
                         "content": response.content})

        if response.stop_reason != "tool_use":
            return

        results = []
        for block in response.content:
            if block.type == "tool_use":
                output = TOOL_HANDLERS[block.name](**block.input)
                results.append({
                    "type": "tool_result",
                    "tool_use_id": block.id,
                    "content": output,
                })
        messages.append({"role": "user", "content": results})

Implementing Modern Mainstream AI Algorithms in Zero-Dependency Single File.This is a teaching project designed specifically for learning AI algorithms, including 30 zero-dependency, single-file, directly runnable Python implementations covering from basic GPT to fine-tuning (LoRA, PPO) and inference optimization (Flash Attention), etc. Each algorithm is implemented with easy-to-understand code, accompanied by corresponding Manim animations for easy comprehension and learning.

Phenomenal Personal AI Assistant.This is an open-source personal AI assistant developed with TypeScript, which can be quickly deployed on macOS, Windows, and Linux systems, and supports interaction through instant messaging apps like WhatsApp, Telegram, and Slack. As long as your token quota is sufficient, it can work continuously 24/7 to serve you

Minimalist AI Agent Toolkit.This is a TypeScript-based AI Agent toolkit. The popular OpenClaw is developed based on this project. It provides fundamental functions for AI Agent development, including unified multi-LLM service interfaces, Agent state management, tool invocation, interactive command-line interface, WebUI, and Slack bot integration, etc.

Intelligent Knowledge Base Search Tool for Local Operation.This is a fully locally-operated intelligent search engine that can be used to retrieve personal documents, knowledge bases, meeting minutes, and Markdown files. It integrates functions such as locally-run lightweight models, BM25 full-text search, vector semantic search, and re-ranking. It is ready to use out of the box, doesn't require internet access, supports the MCP protocol, and can be used as a knowledge search tool in AI assistant and Agent workflows.

AI-Driven Cultivation World Simulator.This is an LLM-based cultivation simulation game. Different from traditional RPGs, all NPCs in the game are AI agents with independent personalities, memories, and behavioral logics. Players play the role of 'Heaven's Will' in the game, observing and intervening in AI cultivators and the rules of the fairy world from a god-like perspective, and witnessing the rise and fall of sects and the emergence of outstanding talents

Controlling Claude Code and Codex via Mobile Phone.This is a tool that enables remote operation of Claude Code or Codex, allowing you to view and remotely control the AI coding assistant anytime and anywhere, with iOS, Android, and Web clients provided.

Generate 3D Scenes in Less Than a Second.This project is the companion code for Apple's open-source monocular view synthesis technique, capable of generating high-quality 3D scenes from a single image in a short time. It regresses 3D Gaussian parameters from a single photo using a neural network and outputs a ply file for the 3DGS renderer.

Lightweight and Efficient ONNX Model Optimization Tool.This is a pure Python implementation of an ONNX model pruning and structure optimization tool with no extra compilation dependencies. By analyzing and rewriting the computational graph, it automatically removes redundant nodes, invalid branches, and excess parameters, reducing the model size and improving inference speed while maintaining model accuracy, suitable for model publishing, inference deployment, and engineering scenarios

import onnx
import onnxslim

model = onnx.load("model.onnx")
slimmed_model = onnxslim.slim(model)

if slimmed_model:
    onnx.save(slimmed_model, "slimmed_model.onnx")