God AI / AURA 1.1.0

Hybrid terminal assistant, Python SDK, and Aura runtime for AI-powered developer workflows.

God AI provides a confirmation-gated terminal assistant, a Python SDK, and the god_ai.aura namespace for resilient execution, structured parsing, system inspection, and model-assisted developer workflows.

What's new in 1.1.0

• Expanded Aura exports for tracing, patch suggestions, sandboxed execution, expression evaluation, data conversion, benchmarking, and cost profiling.
• Lightweight core dependencies with opt-in aura, torch, web, cloud, and examples extras.
• Twenty runnable example files, including ten new real-world integration blueprints.
• A GitHub Pages documentation portal under docs/, with generated logo assets.
• Tag-triggered PyPI publishing through GitHub Actions.

Installation

python -m pip install god-ai

Install optional capabilities as needed:

python -m pip install "god-ai[aura]"      # broad Aura/example integrations
python -m pip install "god-ai[torch]"     # NumPy, PyTorch, and torchvision
python -m pip install "god-ai[web]"       # FastAPI, Uvicorn, and Flask
python -m pip install "god-ai[cloud]"     # hosted provider SDKs
python -m pip install "god-ai[examples]"  # dependencies used by examples

The turtle example uses Python's standard-library turtle module and does not require a separate package.

Configuration

Configuration can come from a .env file or environment variables. The provider defaults to local mode when no OpenRouter key is present.

For OpenRouter:

export AURA_PROVIDER=openrouter
export AURA_API_KEY="your-key"
export AURA_MODEL="openrouter/auto"
export AURA_BASE_URL="https://openrouter.ai/api/v1"

For Ollama:

export AURA_PROVIDER=ollama
export AURA_BASE_URL=http://localhost:11434/v1
export AURA_MODEL=llama3

For another OpenAI-compatible local server, use its /v1 endpoint and the exact model ID returned by /v1/models:

export AURA_PROVIDER=local
export AURA_BASE_URL=http://localhost:1234/v1
export AURA_MODEL=exact-model-id

Useful settings include AURA_FREE_ONLY, AURA_REQUEST_TIMEOUT, AURA_HEAL_RETRIES, and AURA_GHOST_TIMEOUT when passed through the Python configuration API. A refused connection means the local server is not running; a 404 commonly means /v1 is missing from the base URL.

CLI

god [--free] [--web] prompt...
aura [--free] [--web] prompt...

Examples:

god "find the likely cause of this test failure"
god --free "draft a README section about installation"
god --web

--web starts the optional local interface at http://127.0.0.1:8000. Shell commands remain confirmation-gated. Generated code is only described as tested when the CLI actually ran a command and captured its result.

Python SDK

from god_ai import aura

settings = aura.configure(
    provider="openrouter",
    api_key="your-key",
    model="openrouter/auto",
)

@aura.heal
def parse_port(value: str) -> int:
    return int(value)

print(parse_port("8000"))

The public Aura surface includes:

• aura.configure() and aura.get_config() for process-wide settings.
• @aura.heal for model diagnosis and bounded retries; it does not silently edit source files.
• aura.trace_exceptions() and @aura.auto_patch for diagnostic and model-suggested repair workflows.
• aura.do() and aura.exec_sandboxed() for constrained generated Python. AST validation reduces risk but is not a complete security boundary; use a container or low-privilege process for untrusted workloads.
• aura.eval_expr(), aura.parse(), aura.to_pydantic(), and aura.to_dataclass() for structured data workflows.
• @aura.optimize, @aura.benchmark, and @aura.cost_profiler for runtime measurement and suggestions.
• aura.agent() for iterative pytest diagnosis, with edits disabled by default.
• aura.system for read-only platform and package-manager detection.

Example structured parsing:

from pydantic import BaseModel
from god_ai import aura

class Person(BaseModel):
    name: str
    age: int

person = aura.parse("Ada Lovelace is 36 years old", Person)
print(person.name, person.age)

Examples

The repository contains 20 runnable example files. The original compatibility examples are:

examples/01_pytorch_healer.py
examples/02_fastapi_agent.py
examples/03_resilient_microservice.py
examples/04_terminal_scaffolder.py

The 1.1.0 release adds:

examples/01_pytorch_neural_healer.py
examples/02_tensorflow_model_optimizer.py
examples/03_numpy_matrix_ghost.py
examples/04_fastapi_unstructured_parser.py
examples/05_flask_resilient_microservice.py
examples/06_turtle_ai_generative_art.py
examples/07_autonomous_pytest_bot.py
examples/08_data_pipeline_transformer.py
examples/09_multimodal_vision_generator.py
examples/10_terminal_fullstack_scaffolder.py

Run an example from the repository root, for example:

python examples/03_numpy_matrix_ghost.py
python examples/08_data_pipeline_transformer.py

Some examples require optional dependencies or a configured model provider. Each file documents its own prerequisites and fallback behavior. The turtle example opens a desktop window and may not run in a headless environment.

Documentation portal

Serve the static portal locally with:

python -m http.server 8080 --directory docs

Then open http://127.0.0.1:8080. The portal contains API and CLI search, provider configuration helpers, interactive examples, and the generated logo assets. It is suitable for GitHub Pages deployment from the docs/ directory.

Development and verification

git clone https://github.com/ayushgiriai21-cmd/God.git
cd God
python3 -m venv .venv
. .venv/bin/activate
python -m pip install -e .
pytest -q
python -m py_compile $(find src examples scripts -name '*.py')
git diff --check

Build and validate distributions:

python -m pip install build twine
python scripts/process_logo.py
python -m build
python -m twine check dist/*

Release process

The workflow .github/workflows/publish.yml runs for tags matching v*, regenerates package logo assets, builds an sdist and wheel, validates both with Twine, and publishes them to PyPI using the PYPI_API_TOKEN repository secret.

For this release:

git add .
git commit -m "feat: release v1.1.0 - Aura SDK expansion and documentation portal"
git tag v1.1.0
git push origin main --tags

Never commit the PyPI token. The package version and tag must be unique on PyPI.

Project structure

God/
├── assets/                 # source logo artwork
├── docs/                   # GitHub Pages portal
├── examples/               # runnable SDK examples
├── scripts/                # release asset generation
├── src/god_ai/             # installable package
├── tests/                  # contract tests
├── pyproject.toml          # package metadata and extras
└── .github/workflows/      # release automation

License

God AI is available under the MIT License. See LICENSE.