# `myMeans`: Double-Entry Accounting

myMeans automates a plain-text, double-entry accounting workflow built on top of [ledger-cli](https://ledger-cli.org).
It imports downloaded bank `.csv` statements or stages read-only Plaid changes, keeps remote source and local Ledger mutation behind separate review boundaries, categorizes transactions, and generates tabular reports.

It is especially useful if you want:

- a **read-only provider sync** whose fetched source, immutable plan, and generated sidecar remain
  separate and recoverable;
- a visible **manual fallback** for accounts that Plaid does not cover;
- regex templates, routes, and interactive tools for **uncategorized transactions**;
- exact-cent **conditional planning** from explicit cashflow and debt assumptions, without reading
  or mutating a Ledger;
- an optional, explicitly remote **LLM template suggester** whose proposals still require local
  confirmation; or
- typed, validated accounting primitives for accounts, posts, balanced transactions, and ledgers.

## Install

Install ledger-cli through your operating system, then install the published Python package:

```sh
ledger --version
uv tool install my-means
means-bank --help
```

Python API users can run `uv add my-means` instead.
The hosted LLM path is an optional extra:

```sh
uv tool install 'my-means[llms]'
```

## First workflow

Every data command accepts one configured means directory.
Status is read-only and is the safest first probe:

```sh
export MY_MEANS=/path/to/your/means-directory
means-bank --directory "$MY_MEANS" status --json
```

From there, the provider workflow is `fetch` → inspect `accounts` → `bind` → `plan` → reviewed `apply`.
The deterministic categorizer stays local; the optional LLM command sends unknown memos, templates, and the account tree to Anthropic and writes only after confirmation:

```sh
categorize --directory "$MY_MEANS" --auto

export ANTHROPIC_API_KEY=your-api-key
suggest --directory "$MY_MEANS" --dry-run
```

`--dry-run` prevents a local write but still makes the remote request.
See the command-line guide for the complete security and recovery contract.

The documentation separates user workflows from the Python API and shared infrastructure:

```{toctree}
:maxdepth: 2
1. Conditional Planning <planning>
2. Scenario Lab <scenario_lab>
3. Interactive Interfaces <interfaces>
4. Bank Synchronization <bank_sync>
5. Command-Line Tools <cli>
6. Core Data Model <types>
7. Shared Infrastructure <base>
```