moonclaude 2.0.0__tar.gz

moonclaude-2.0.0/PKG-INFO

@@ -0,0 +1,110 @@
+Metadata-Version: 2.4
+Name: moonclaude
+Version: 2.0.0
+Summary: Moon for Claude: run Claude Code on external LLMs via LiteLLM
+License: MIT
+Keywords: moonclaude,claude,llm,openrouter,litellm,cli
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+Requires-Dist: pyyaml>=6.0
+Requires-Dist: litellm[proxy]>=1.40.0
+Requires-Dist: rich>=13.0.0
+
+# MoonClaude 🌙
+
+**Moon for Claude**: Run [Claude Code](https://github.com/anthropics/claude-code) fully autonomously on external LLMs (e.g. Qwen, Gemini, MiniMax, OpenAI) through a LiteLLM proxy, while enjoying **zero-downtime hot reloading** and an **Antigravity-grade persistent AI memory** engine.
+
+---
+
+## 🚀 Features
+
+- **Free OpenRouter Models**: Interactively fetch and route your prompts through any of the 25+ completely free models on OpenRouter, or configure your own premium API keys (Gemini, Groq, OpenAI).
+- **True Zero-Downtime Hot Reloading**: Press `Alt+M` directly in the proxy terminal to instantaneously swap the active model. MoonClaude dynamically intercepts the routing under the hood—no proxy restarts, no lost context in the CLI, no downtime.
+- **Antigravity-Grade Project Memory**:
+  - Auto-discovers and enforces `MOONCLAUDE.md` global project instructions into Claude Code's system prompts.
+  - Generates cross-session, highly structured summarization context representing your recent AI coding sessions, intelligently managing your AI's working context without maxing out open files.
+  - Easily browse transcripts natively with `moon history`.
+
+---
+
+## 📦 Installation & Setup
+
+### 1. Prerequisites 
+If you haven't already, you need `Node.js` installed to run Anthropic's official Claude Code CLI.
+```bash
+# Install Claude Code globally via npm
+npm install -g @anthropic-ai/claude-code
+```
+
+### 2. Install MoonClaude
+MoonClaude is available directly on PyPI!
+```bash
+pip install moonclaude
+```
+
+### 3. Initialize Settings
+Run the interactive setup wizard to map your API keys (OpenRouter, Gemini, Groq) and select your preferred chat models.
+```bash
+moon setup
+```
+
+---
+
+## 🏃 Quick Start
+
+The easiest way to launch the proxy and jump straight into Claude Code is a single command. Open your project folder and run:
+```bash
+moon launch
+```
+*(If you haven't run setup yet, it will guide you through it first!)*
+
+**Manual Control:**
+If you prefer running the Proxy and Claude independently, open two terminals:
+- **Terminal 1**: Run `moon start` (Starts the LiteLLM routing proxy)
+- **Terminal 2**: Run `moon chat` (Launches Claude Code seamlessly attached to the proxy)
+
+---
+
+## 🎛️ Core Commands
+
+```text
+moon setup                  Interactive wizard for API keys and preferred models
+moon launch                 Quick launch (starts proxy in background, then launches Claude)
+moon chat                   Launch Claude connected to the proxy
+moon start                  Start LiteLLM proxy in the foreground with live token monitoring
+moon switch                 Fast CLI model switcher 
+moon history                Interactive rich browser for your past conversation logs
+moon memory                 Detailed control over the project memory context engine
+moon status                 Display system health, active models, and paths
+moon                        Show CLI help and commands
+```
+
+> **Compatibility**: MoonClaude preserves the old `claude-ext ...` legacy command aliases. Your legacy `~/.claude-ext` config will automatically migrate to `~/.moonclaude/` at launch!
+
+---
+
+## ♻️ Zero-Downtime Hot Reloading
+
+When you are deep in a Claude Code debug session, discovering your chosen model is failing can be frustrating.
+
+Instead of dropping context:
+1. Keep the proxy running in the foreground (`moon start`).
+2. Press `Alt+M`. 
+3. Select a new LLM from your interactive UI overlay.
+4. MoonClaude dynamically pre-warms its internal router injection without dropping the proxy connection. Your next prompt in the Claude Code terminal instantly bridges over to the new model!
+
+---
+
+## 🧠 Memory Engine
+
+MoonClaude injects long-term memory into Claude Code.
+
+1. **`MOONCLAUDE.md`**: Place this file in your project root. MoonClaude automatically reads it and appends it to your AI's persistent context across all sessions.
+2. **Context Compression**: When you boot `moon chat`, MoonClaude evaluates your last 20 sessions and constructs an elegant, compressed timeline of files you've modified, goals you've accomplished, and context you shouldn't forget.
+3. **Session Replay**: Type `moon history` to explore Rich-formatted logs of all of your previous AI code iterations.
+
+---
+
+## License
+
+MIT 

moonclaude-2.0.0/README.md

@@ -0,0 +1,98 @@
+# MoonClaude 🌙
+
+**Moon for Claude**: Run [Claude Code](https://github.com/anthropics/claude-code) fully autonomously on external LLMs (e.g. Qwen, Gemini, MiniMax, OpenAI) through a LiteLLM proxy, while enjoying **zero-downtime hot reloading** and an **Antigravity-grade persistent AI memory** engine.
+
+---
+
+## 🚀 Features
+
+- **Free OpenRouter Models**: Interactively fetch and route your prompts through any of the 25+ completely free models on OpenRouter, or configure your own premium API keys (Gemini, Groq, OpenAI).
+- **True Zero-Downtime Hot Reloading**: Press `Alt+M` directly in the proxy terminal to instantaneously swap the active model. MoonClaude dynamically intercepts the routing under the hood—no proxy restarts, no lost context in the CLI, no downtime.
+- **Antigravity-Grade Project Memory**:
+  - Auto-discovers and enforces `MOONCLAUDE.md` global project instructions into Claude Code's system prompts.
+  - Generates cross-session, highly structured summarization context representing your recent AI coding sessions, intelligently managing your AI's working context without maxing out open files.
+  - Easily browse transcripts natively with `moon history`.
+
+---
+
+## 📦 Installation & Setup
+
+### 1. Prerequisites 
+If you haven't already, you need `Node.js` installed to run Anthropic's official Claude Code CLI.
+```bash
+# Install Claude Code globally via npm
+npm install -g @anthropic-ai/claude-code
+```
+
+### 2. Install MoonClaude
+MoonClaude is available directly on PyPI!
+```bash
+pip install moonclaude
+```
+
+### 3. Initialize Settings
+Run the interactive setup wizard to map your API keys (OpenRouter, Gemini, Groq) and select your preferred chat models.
+```bash
+moon setup
+```
+
+---
+
+## 🏃 Quick Start
+
+The easiest way to launch the proxy and jump straight into Claude Code is a single command. Open your project folder and run:
+```bash
+moon launch
+```
+*(If you haven't run setup yet, it will guide you through it first!)*
+
+**Manual Control:**
+If you prefer running the Proxy and Claude independently, open two terminals:
+- **Terminal 1**: Run `moon start` (Starts the LiteLLM routing proxy)
+- **Terminal 2**: Run `moon chat` (Launches Claude Code seamlessly attached to the proxy)
+
+---
+
+## 🎛️ Core Commands
+
+```text
+moon setup                  Interactive wizard for API keys and preferred models
+moon launch                 Quick launch (starts proxy in background, then launches Claude)
+moon chat                   Launch Claude connected to the proxy
+moon start                  Start LiteLLM proxy in the foreground with live token monitoring
+moon switch                 Fast CLI model switcher 
+moon history                Interactive rich browser for your past conversation logs
+moon memory                 Detailed control over the project memory context engine
+moon status                 Display system health, active models, and paths
+moon                        Show CLI help and commands
+```
+
+> **Compatibility**: MoonClaude preserves the old `claude-ext ...` legacy command aliases. Your legacy `~/.claude-ext` config will automatically migrate to `~/.moonclaude/` at launch!
+
+---
+
+## ♻️ Zero-Downtime Hot Reloading
+
+When you are deep in a Claude Code debug session, discovering your chosen model is failing can be frustrating.
+
+Instead of dropping context:
+1. Keep the proxy running in the foreground (`moon start`).
+2. Press `Alt+M`. 
+3. Select a new LLM from your interactive UI overlay.
+4. MoonClaude dynamically pre-warms its internal router injection without dropping the proxy connection. Your next prompt in the Claude Code terminal instantly bridges over to the new model!
+
+---
+
+## 🧠 Memory Engine
+
+MoonClaude injects long-term memory into Claude Code.
+
+1. **`MOONCLAUDE.md`**: Place this file in your project root. MoonClaude automatically reads it and appends it to your AI's persistent context across all sessions.
+2. **Context Compression**: When you boot `moon chat`, MoonClaude evaluates your last 20 sessions and constructs an elegant, compressed timeline of files you've modified, goals you've accomplished, and context you shouldn't forget.
+3. **Session Replay**: Type `moon history` to explore Rich-formatted logs of all of your previous AI code iterations.
+
+---
+
+## License
+
+MIT 

moonclaude-2.0.0/moonclaude/__init__.py

@@ -0,0 +1,4 @@
+__all__ = ["__version__"]
+
+__version__ = "2.0.0"
+

moonclaude-2.0.0/moonclaude/__main__.py

@@ -0,0 +1,6 @@
+from .main import main
+
+
+if __name__ == "__main__":
+    main()
+

moonclaude-2.0.0/moonclaude/branding.py

@@ -0,0 +1,18 @@
+"""
+Branding constants shared across MoonClaude modules.
+"""
+
+import os
+from pathlib import Path
+
+
+APP_NAME = "moonclaude"
+APP_TITLE = "MoonClaude"
+CLI_NAME = "moonclaude"
+CLI_ALIAS = "moon"
+LEGACY_CLI_NAME = "claude-ext"
+
+CONFIG_DIR = Path(os.environ.get("MOONCLAUDE_HOME", Path.home() / ".moonclaude")).expanduser()
+LEGACY_CONFIG_DIR = Path(os.environ.get("MOONCLAUDE_LEGACY_HOME", Path.home() / ".claude-ext")).expanduser()
+
+DEFAULT_PORT = 4000