# Kinenix

> Kinenix is an open-source RPA framework in Python that runs on the user's own machines. An automation is a flow project: the task described in plain language in `requirements.md`, the flow written in Markdown in `flow.md` (usually by an AI assistant), settings in `config/config.json`, and secrets in `.env`. `kinenix run` runs it; unattended workers run it on a schedule and report to a self-hosted Hub.

How an AI assistant builds a flow:

1. Read `requirements.md` and the project's `AGENTS.md` (created by `kinenix init`; it holds the rules below in full).
2. Write `flow.md` as Markdown steps: `### N. Step Name (`action.name`)` followed by `- **parameter:** value` bullets. There is no `flow.json`.
3. Put changeable values in `config/config.json` (read as `${config.name}`) and secrets only in `.env` (read as `${env.NAME}`).
4. Run `kinenix actions` for the exact parameters of every action; do not guess parameter names.
5. Run `kinenix validate .` and fix every problem, then `kinenix run .` and read the log.

Expressions in `${...}` are a fixed, safe set of functions (no code runs). Business data stays local: do not send page content or flow data to cloud services unless the task asks for it.

## Build flows

- [Flow Markdown specification](https://www.kinenix.com/flow_markdown_spec/): the `flow.md` format: steps, parameters, loops, conditions, expressions, `on_error` (`continue`, `next_item`), `## On Error` and `## Finally`, `${error}` and `${run}`
- [Actions reference](https://www.kinenix.com/actions_reference/): every action (web, Excel and CSV, files, logic, dates, HTTP, subflows, e-mail, local AI) with its parameters and a `flow.md` example
- [Project bundles](https://www.kinenix.com/project_bundles/): the project folder, `config.json`, `.env`, subflows with `flow.call`, and testing a flow against saved pages with `kinenix run --config`
- [CLI guide](https://www.kinenix.com/cli_guide/): `kinenix init`, `run`, `validate`, `actions`, `list`, and their options

## Start

- [Quickstart](https://www.kinenix.com/quickstart/): install, create a project for a task, build it with an AI assistant, run it, and connect a worker
- [Tutorial](https://www.kinenix.com/tutorial/): a first flow from `requirements.md` to a run, step by step (`kinenix init my-first-flow --example first_flow`), with a solution in the repository to compare afterwards
- [Examples](https://www.kinenix.com/examples/): complete projects, each with the `requirements.md` it implements, and the patterns each shows; `kinenix init --list` shows them and `kinenix init NAME --example hello` copies one

## Run and operate

- [Hub](https://www.kinenix.com/hub/): the Hub, connecting workers (`kinenix connect`), the dashboard, and why a run does not appear there
- [Hub data](https://www.kinenix.com/hub_data/): tables and views of run records, for answering questions about runs with `kinenix hub query`
- [Logging](https://www.kinenix.com/logging/): the execution log format, failure causes, and redaction of secrets and personal data

## Optional

- [Architecture](https://www.kinenix.com/architecture/): modules and design principles
- [Roadmap](https://www.kinenix.com/roadmap/): what is built, what is planned, and the release milestones
- [Changelog](https://github.com/arttopix/kinenix/blob/main/CHANGELOG.md): changes and upgrade notes of each release