OpenRastr Evolve

OpenRastr-Evolve is a Python-based AI workflow runtime built around three core modules:

1. Goal Interpretation
2. Capability Registration
3. Agent Spawning

It is designed to work with Ollama-served LLMs and SLMs and to act as an orchestration layer for agent-based systems.

What It Does

OpenRastr-Evolve helps turn a natural-language goal into an executable AI workflow by:

• interpreting the goal into a structured schema
• matching or generating the capabilities needed to fulfill it
• spawning agents to carry out the work

Who This Is For

OpenRastr Evolve is useful for teams and builders who already use tools such as:

• Claude Code
• Codex
• OpenClaw
• Ollama-backed local agents
• custom coding or execution agents

It is not intended to replace those tools.

Instead, OpenRastr Evolve sits as a workflow layer that helps:

• interpret high-level goals
• map goals to reusable capabilities
• register missing capabilities
• spawn and coordinate agents around those capabilities

In other words, if Claude Code, Codex, or OpenClaw are the workers, OpenRastr Evolve is the orchestration layer that helps decide what work should be done and how it should be structured.

Prerequisites

Before you start, make sure you have:

• Python 3.10 or newer
• pip
• Ollama installed and running
• at least one Ollama model available locally
• for example, qwen2.5:7b

Quick Start: macOS and Linux

git clone https://github.com/Rastrai/Openrastr-evolve.git
cd Openrastr-evolve
./scripts/bootstrap.sh
openrastr-evolve doctor

The bootstrap script will:

• create a local virtual environment in .venv
• install required packaging tools
• install the project in editable mode
• launch the OpenRastr Evolve onboarding wizard

Quick Start: Windows

Use PowerShell for the steps below.

git clone https://github.com/Rastrai/Openrastr-evolve.git
cd Openrastr-evolve

py -m venv .venv
.\.venv\Scripts\Activate.ps1

python -m pip install --upgrade pip
python -m pip install setuptools wheel
python -m pip install --no-build-isolation -e .
python -m openrastr_evolve.cli onboard
python -m openrastr_evolve.cli doctor
python -m openrastr_evolve.cli run --help

If PowerShell blocks virtual environment activation, run:

Set-ExecutionPolicy -Scope Process Bypass
.\.venv\Scripts\Activate.ps1

If openrastr-evolve is not recognized on Windows, use:

python -m openrastr_evolve.cli onboard
python -m openrastr_evolve.cli doctor
python -m openrastr_evolve.cli run --help

Manual Setup

If you prefer to install manually on macOS or Linux:

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install setuptools wheel
pip install --no-build-isolation -e .
openrastr-evolve onboard
openrastr-evolve doctor

CLI Commands

On macOS and Linux:

openrastr-evolve onboard
openrastr-evolve doctor
openrastr-evolve run --goal-file path/to/goal.txt
openrastr-evolve register-skill path/to/skill.md

On Windows, if the CLI command is not available in PATH:

python -m openrastr_evolve.cli onboard
python -m openrastr_evolve.cli doctor
python -m openrastr_evolve.cli run --goal-file path\to\goal.txt
python -m openrastr_evolve.cli register-skill path\to\skill.md

Configuration

The onboarding wizard writes configuration to:

~/.openrastr_evolve/config.json

On Windows, ~ refers to your user home directory.

Supported settings include:

• workspace_root
• ollama_base_url
• goal_interpreter_model
• goal_interpreter_fallback_model
• pipeline_capability_model
• pipeline_agent_model

Recommended Install Mode

For a GitHub source checkout, the recommended installation method is:

pip install --no-build-isolation -e .

This keeps the repository editable while exposing the openrastr-evolve CLI command.

If you want the fastest setup path on macOS or Linux, use:

./scripts/bootstrap.sh

First Run

Typical first-run flow on macOS and Linux:

openrastr-evolve onboard
openrastr-evolve doctor
openrastr-evolve run --help

Typical first-run flow on Windows:

python -m openrastr_evolve.cli onboard
python -m openrastr_evolve.cli doctor
python -m openrastr_evolve.cli run --help

License

This project is licensed under the GNU General Public License v3.0.

This product includes software developed by RastrAI Pvt Ltd.