中文

Code Atlas

Overview

This project is a lightweight interpreter developed in C++, inspired by Open Interpreter. It supports the execution of multi-language code snippets and partial Markdown rendering. Compared to Open Interpreter, this project has significantly lower resource consumption, making it more suitable for resource-limited devices. Additionally, it leverages llama.cpp for local large-model inference, providing an efficient and secure execution environment. It also supports cloud-based large-model APIs, configurable via config.json.

---

Features

1. Code Execution:

   • Supports scripting languages like Python, Batch, PowerShell, and Shell/Bash.
   • Executable code blocks are highlighted in purple, while others are shown in yellow.

2. Markdown Rendering:

   • Supports the following Markdown elements:
     • Headings (levels 1–4)
     • Bold and Italic
     • Code blocks
   • Output is color-coded for enhanced readability.

3. Local AI Inference:

   • Uses llama.cpp for efficient local large-language model inference.
   • Compatible with gguf models, such as the Qwen2.5-Interpreter, fine-tuned for this task.
   • Configuration: Set API to False in config.json.

4. Cloud-Based Large Models:

   • Supports HTTPS API-based cloud models like OpenAI.
   • Configuration: Set API to True in config.json and provide the base URL and API key.

5. Resource Efficiency:

   • Significantly reduces memory and computational resource requirements compared to Open Interpreter.

---

Usage

1. Download the project source code and compile it using a compatible C++ compiler, or download the precompiled executable (Releases).

g++ code-atlas.cpp -o code-atlas.exe -std=c++17 -lcurl -lws2_32

2. Download or compile llama.cpp, ensuring the following files are available:

   • llama.cli.exe
   • Necessary .dll files
   • A compatible gguf model (e.g., Qwen2.5-Interpreter) or an accessible API.

3. Place the following files in the same directory:

   • The project's executable
   • llama.cli.exe
   • Required .dll files
   • Your chosen gguf model
   • config.json

4. Run the program. Note that administrator permissions might be required on some systems.

---

To-Do List

• Support more platforms (e.g., Linux, macOS).
• Develop a graphical user interface (GUI) for enhanced usability.
• Allow users to choose the programming language for code execution.
• Expand supported programming languages beyond Python, Batch, PowerShell, and Shell/Bash.
• Add plugin support to extend functionality.
• Support additional AI model APIs, such as OpenAI.
• Enhance sandboxing mechanisms for improved execution security.
• Extend Markdown rendering to support more elements.
• Support voice mode for hands-free operation.

---

Inspiration

This project is inspired by Open Interpreter, which we admire for its innovative approach to integrating natural language and programming. By optimizing performance, this project is implemented in C++ with specific enhancements for Windows systems while retaining the intuitive and powerful core design.

Additionally, the Qwen2.5-Interpreter model, fine-tuned from Qwen2.5-0.5B, showcases the project's potential for task optimization and performance enhancement.

---