Locked & Coding
project name: Le Bac à Sable Add a sandbox to Le Chat to supercharge its capabilities.
Core trick
Attaches a stateful Linux container to each Le Chat session so tool calls can edit files and run commands in place.
Description
project name: Le Bac à Sable
Add a sandbox to Le Chat to supercharge its capabilities. Iterate on your files and datasets directly in your sandbox and deploy your product—all from Le Chat.
Le Bac à Sable attaches a Linux container sandbox to every chat, enabling the model to create, modify, and execute files during the conversation in a stateful way. No more document copy jungle, in modify your file in place. While Mistral Medium excels at coding tasks, the native code interpreter tool has limitations (one shot input/output). Le Bac à Sable is especially useful when working with medium to large files or transferring them between tools.
**Examples:**
- Ask Le Chat to download and clean a dataset. Le Bac à Sable’s MCP provides a tool to run commands for that.
- Build a full-stack app, iterate on the generated code, and deploy it—all from Le Chat.
- Convert video or audio files to your desired format.
README excerpt
# Le Bac à Sable
Add a sandbox to **Le Chat** to supercharge its capabilities. Iterate on your files and datasets directly in your sandbox and deploy your product—all from Le Chat.
Le Bac à Sable attaches a Linux container sandbox to every chat, enabling the model to create, modify, and execute files during the conversation. While Mistral Medium excels at coding tasks, the native code interpreter tool has limitations. This is especially useful when working with medium to large files or transferring them between tools.
## Examples
- Ask Le Chat to download and clean a dataset. Le Bac à Sable’s MCP provides a tool to run commands for that.
- Build a full-stack app, iterate on the generated code, and deploy it—all from Le Chat.
- Convert video or audio files to your desired format.
- Expose your workspace to the internet with ngrok for quick sharing or testing.
- Push your sandbox files directly to a new GitHub repository.
- …and more 🚀
---
## Installation
```bash
uv python install
uv sync --locked
```
## Usage
Start the server on port 3000:
```bash
uv run main.py
```
---
## Available Tools
### Command & File Tools
1. **`run_command`**
Execute arbitrary shell commands with structured output inside the sandbox container.
- Parameters: `command`, `stdin`, `timeout`, `shell`, `max_output_bytes`.
- Response: `segments`, `exit_code`, `truncated`, `timeout`, `is_error`.
Example:
```python
run_command("ls -1")
```
2. **`write_t