tiny-7coder

A return to form for 7coder — going back to its origins of being a simple tool.

"Not a coding platform. Just a model with hands." --ChatGPT

tiny-7coder is a small, dependency-free Unix-like systems agent designed to do one thing well: take a user's request, reason through the task with an AI model, and perform actions directly on the local system through a minimal set of tools.

Instead of being a large framework, plugin system, or complicated automation platform, tiny-7coder focuses on simplicity. It is a lightweight command-line companion that connects an LLM to a handful of practical system operations.

Features

• 🪶 Tiny footprint — built with only Bash and Python's standard library

• 🤖 AI-powered task execution — uses a local or remote model endpoint

• 🛠️ Simple tool interface

  • Read files
  • Append to files
  • Replace file contents
  • Run shell commands
  • List directories
  • Reply
  • Get website HTML
  • Do a web search

• 🔌 Model agnostic — works with any Ollama-compatible API endpoint

• ⚙️ Minimal configuration — no databases, no services, no dependencies

• 🐧 Unix-oriented workflow — designed around the command line and filesystem

Philosophy

tiny-7coder is intentionally small.

The goal is not to create a massive coding environment or an all-in-one development platform. The goal is to return to the original spirit of 7coder: a simple tool that gets out of the way and helps you accomplish tasks quickly.

A user gives a request. The model decides the next action. The tool executes it. The process repeats until the task is complete.

Requirements

• Python 3.9+
• An Ollama-compatible AI model endpoint
• A Unix-like environment (Linux/macOS recommended)

Installation

Follow instructions to create .env

Clone the repository:

git clone <repository-url>
cd /your/project/path

Run it:

python3 /path/to/t7c.py "Create a hello world script"

Configuration

tiny-7coder stores its configuration in:

~/.config/nodemixaholic-software/tiny-7coder/.env

The configuration file is automatically created on first run.

Example configuration:

MODEL=deepseek-v4-flash:cloud
HOST=localhost:11434
VERIFY_SSL=False

Configuration Options

Variable	Description	Default
MODEL	Model name to use	deepseek-v4-flash:cloud
HOST	Model API host	http://127.0.0.1:11434
SEARCH_PREFIX	Search query prefix	https://searx.nodemixaholic.com/search?q=
VERIFY_SSL	Enable SSL certificate verification	false

How It Works

1. The user provides a task:

tiny-7coder "Update the README with installation instructions"

2. tiny-7coder sends the request and current context to the configured model.

3. The model responds with a single executable tool command.

4. tiny-7coder runs that command.

5. The result is returned to the model.

6. The loop continues until the model responds with:

DONE

The agent limits execution to 200 steps per request to prevent runaway operations.

Security Notes

tiny-7coder can execute shell commands on the machine where it runs.

Use it responsibly:

• Review the model being used
• Avoid running it with unnecessary privileges
• Use isolated environments for untrusted tasks
• Keep backups of important files

Why "tiny"?

Because useful tools do not need to be complicated.

tiny-7coder intentionally avoids:

• Large dependency trees
• Complex configuration systems
• Heavy user interfaces
• Unnecessary abstractions

It is a small bridge between an AI model and the operating system.

License

MIT