Metadata-Version: 2.5
Name: rio
Version: 0.7.1
Summary: rio.ai: a ported multi-provider LLM streaming SDK. rio.agent: a Context Language Model (CLM) agent runtime (arXiv:2609.37725). rio.coding: a CLM coding agent.
Project-URL: Repository, https://github.com/soasme/rio
Author-email: Ju Lin <soasme@gmail.com>
License-Expression: MIT
License-File: LICENSE
Requires-Python: >=3.12
Requires-Dist: anyio>=4.0
Requires-Dist: httpx[socks]>=0.27
Requires-Dist: ipykernel>=7.4.0
Requires-Dist: jsonpatch>=1.33
Requires-Dist: nbclient>=0.11.0
Requires-Dist: nbformat>=5.11.1
Requires-Dist: packaging>=24.0
Requires-Dist: pillow>=11.0
Requires-Dist: pydantic>=2.11
Requires-Dist: pygments>=2.18
Requires-Dist: pyte>=0.8.2
Description-Content-Type: text/markdown

# rio

Rio is a fully autonomous, one-shot coding agent.
It has no built-in interactive TUI, no conversation history, no human steer in the middle and no follow-up turns.
Start a task and Rio runs until the task completes or aborts.

## Getting Started

Install via `uv`:

```bash
uv tool install rio
```

## Usage

Login to your preferred providers:

* `rio login anthropic`
* `rio login github-copilot`
* `rio login codex`
* ...

Run a adhoc task:
```
rio run "Fix gh issue 123."
```

Rio works in a Jupyter notebook: it acts by adding Python, shell, and `%%edit`
cells, which run in one live kernel in the project directory.

Run a predefined workflow:

```
rio run task.md
```

The workflow is to write the task in a Markdown file, then run Rio
against it.


You can add a short instruction when invoking it:

```bash
uv run rio run "Follow feature.md, run the tests, and finish the implementation"
```

Use `--provider NAME --model MODEL` to choose a provider and a model.

Rio prints a compact human transcript by default. Use `--output json` for one JSON
event per line. Each human-mode run prints a session id; continue it with
`rio run --resume SESSION_ID "Next task"`.

## Documentation

See the [documentation](docs/README.md) for tutorials, guides, command-line
reference, and an explanation of Rio's architecture.

## Development

```bash
uv run pytest
uv run ruff check .
```

To run the Lean 4 verification, install [elan](https://github.com/leanprover/elan)
and build the pinned toolchain from the repository root:

```bash
cd tests/lean && lake build
```

See [the Lean behavior models](tests/lean/README.md) for their scope and limits.

* `rio.ai` is a multi-provider LLM streaming SDK.
* `rio.agent` is the [Context Language Model][CLM] runtime: the context is a
  Jupyter notebook the model patches, and the runtime runs the changed cells.
* `rio.coding` supplies the autonomous coding skill.
* `rio.cli` is the public one-shot command-line entry point.

[CLM]: https://arxiv.org/abs/2609.37725
