lemma_engine

# Lemma

Elixir client for the [Lemma](https://github.com/lemma/lemma) rules engine, via precompiled NIFs.

## Requirements

- Elixir >= 1.14

Precompiled NIF binaries are downloaded automatically for macOS (arm64/x86_64), Linux (gnu x86_64 and arm64), and Windows (arm64/x86_64).

## Development

To compile the NIF from source (contributors), set `LEMMA_BUILD_NIF=1` and have Rust stable on `PATH`:

```bash
LEMMA_BUILD_NIF=1 mix compile
```

Pre-commit check (format, locked deps, compile, ExUnit):

```bash
mix precommit
```

Builds the NIF from source when no checksum file is present (typical in this repo).

## Installation

Add to `mix.exs`:

```elixir
def deps do
  [
    {:lemma_engine, "~> 0.9"}
  ]
end
```

Or from git:

```elixir
{:lemma_engine, git: "https://github.com/lemma/lemma", sparse: "engine/packages/hex"}
```

## Usage

```elixir
# Create an engine
{:ok, engine} = Lemma.new()

# Load inline source
:ok = Lemma.load(engine, """
spec pricing
data quantity: number
data price: 10
rule total: quantity * price
rule discount: 0
  unless quantity >= 10 then 5
  unless quantity >= 50 then 15
""")

# Or load labeled paths / dependencies in one planning pass
:ok = Lemma.load(engine, %{"pricing.lemma" => pricing_source})

# Run with data
{:ok, response} =
  Lemma.run(
    engine,
    %{spec: "pricing"},
    %{data: %{"quantity" => "25"}, explain: false}
  )

# List loaded specs (grouped by repository)
{:ok, groups} = Lemma.list(engine)
# [%{repository: nil, specs: [...]}, %{repository: "lemma", specs: [...]}, ...]
{:ok, show} = Lemma.show(engine, nil, "pricing")

# Canonical Lemma source
{:ok, source} = Lemma.source(engine, nil, "pricing")

# Format source code (no engine needed)
{:ok, formatted} = Lemma.format("spec foo\ndata x: 1\nrule y: x + 1")

# Clean up
:ok = Lemma.remove(engine, nil, "pricing", "2025-01-01")
```

## API

| Function | Description |
|----------|-------------|
| `Lemma.new/1` | Create engine (optional limits map) |
| `Lemma.load/2` | Load sources: binary (volatile) or map / list of `{label, code}` (labeled) |
| `Lemma.list/1` | List loaded specs (includes embedded `lemma` / `spec units`) |
| `Lemma.show/4` | Spec interface and temporal window (`repository`, `spec`, `effective`) |
| `Lemma.source/4` | Formatted canonical Lemma source |
| `Lemma.run/3` | Evaluate a spec (`target` + `options` maps; `explain: true` adds per-rule explanation trees) |
| `Lemma.remove/4` | Remove a spec from the engine |
| `Lemma.format/1` | Format Lemma source code (no engine needed) |

See `Lemma` module docs for full typespecs and options.

## Data values

Data values passed to `Lemma.run/3` must be strings or integers. Elixir floats (IEEE 754 doubles) are rejected to prevent silent precision loss — pass decimal values as strings instead:

```elixir
Lemma.run(engine, %{spec: "pricing"}, %{data: %{"rate" => "0.075"}, explain: false})   # correct
Lemma.run(engine, %{spec: "pricing"}, %{data: %{"quantity" => 42}, explain: false})     # correct (integer)
Lemma.run(engine, %{spec: "pricing"}, %{data: %{"rate" => 0.075}, explain: false})      # raises — use string
```

With `explain: true`, each rule result may include an `explanation` object matching [`explanation.v1.json`](../../../documentation/schemas/explanation.v1.json).

## Engine lifecycle

Each `Lemma.new/1` call creates an independent engine. The engine reference is safe to use from a single process. For shared access across processes, wrap it in a GenServer.

## License

Apache-2.0