build: share language checks and staged pre-commit validation

.claude/settings.json

@@ -1 +1,8 @@
-{}
+{
+  "hooks": {
+    "PostToolUse": [{
+      "matcher": "Edit|Write",
+      "hooks": [{"type": "command", "command": "python3 \"$CLAUDE_PROJECT_DIR/scripts/agent_check.py\"", "timeout": 45}]
+    }]
+  }
+}

.codex/hooks.json

@@ -0,0 +1,8 @@
+{
+  "hooks": {
+    "PostToolUse": [{
+      "matcher": "apply_patch|Edit|Write",
+      "hooks": [{"type": "command", "command": "python3 \"$(git rev-parse --show-toplevel)/scripts/agent_check.py\"", "timeout": 45}]
+    }]
+  }
+}

.credo.exs

@@ -0,0 +1,17 @@
+%{
+  configs: [
+    %{
+      name: "default",
+      files: %{included: ["lib/", "test/"], excluded: [~r"/fixtures/", ~r"/package/"]},
+      checks: %{
+        enabled: [
+          {Credo.Check.Warning.IExPry, []},
+          {Credo.Check.Warning.IoInspect, []},
+          {Credo.Check.Warning.UnusedEnumOperation, []},
+          {Credo.Check.Consistency.TabsOrSpaces, []},
+          {Credo.Check.Readability.TrailingBlankLine, []}
+        ]
+      }
+    }
+  ]
+}

.formatter.exs

@@ -1,4 +1,7 @@
 # Used by "mix format"
 [
-  inputs: ["{mix,.formatter}.exs", "{config,lib,test}/**/*.{ex,exs}"]
+  inputs: [
+    "{mix,.formatter,.credo}.exs",
+    "{config,lib,test,scripts,bench,examples}/**/*.{ex,exs}"
+  ]
 ]

Makefile

@@ -1,9 +1,10 @@
 .DEFAULT_GOAL := help
 PYTHON ?= python3
+export PATH := $(CURDIR)/.venv/bin:$(CURDIR)/.venv/Scripts:$(PATH)
 export MIX_ENV := test
 export ORT_DYLIB_PATH ?= $(shell $(PYTHON) scripts/install_runtime.py path 2>/dev/null)
 
-.PHONY: help setup deps doctor test docs
+.PHONY: help setup deps doctor test docs hooks fmt lint check
 
 help: ## Show development commands
 	@awk 'BEGIN {FS = ":.*## "} /^[a-zA-Z0-9_.-]+:.*## / {printf "  %-22s %s\n", $$1, $$2}' $(MAKEFILE_LIST)
@@ -13,6 +14,8 @@ setup: ## Fetch Mix dependencies and the development runtime
 	mix local.rebar --force
 	$(MAKE) deps
 	$(PYTHON) scripts/install_runtime.py install
+	$(PYTHON) -m venv .venv
+	.venv/bin/python -m pip install -r scripts/requirements.txt
 
 deps: ## Fetch the locked test and documentation dependencies
 	mix deps.get --check-locked
@@ -26,3 +29,17 @@ test: ## Run library tests and doctests
 
 docs: ## Build documentation with warnings treated as errors
 	MIX_ENV=docs mix docs --warnings-as-errors
+
+hooks: ## Install the pre-commit launcher (refuses to overwrite an unrelated hook)
+	$(PYTHON) scripts/hooks.py
+
+fmt: ## Apply the language formatters
+	mix format
+	cd native/ortex && cargo fmt
+	ruff format .
+	shfmt -w scripts/precommit
+
+lint: ## Check formatting, static analysis, and compiler warnings
+	$(PYTHON) scripts/check.py --full
+
+check: lint test ## Run the ordinary local checks

docs/development.md

@@ -23,3 +23,17 @@ normal ExUnit discovery. Additional suites are introduced by later stack PRs.
 
 Documentation has its own Mix environment so building docs does not install
 EXLA, Torchx, or example tokenizers. Large examples own their extra dependencies.
+
+## Checks and hooks
+
+Activate `.venv/bin` alongside the mise tools after setup. `make lint` runs
+format checks, Credo, Clippy, and workflow validation. `make fmt` applies
+formatters. `scripts/precommit` runs fast checks against an isolated snapshot of
+the Git index, so partially staged files are checked as committed. It never
+changes the index or working tree. Filenames are passed as arguments, not shell
+code. Full checks require prepared dependencies; hooks never install them.
+
+`make hooks` installs a launcher through `git rev-parse --git-path`; it refuses
+to replace an unrelated hook. Git's hooks directory is shared by linked
+worktrees. Agent hooks are adapters to the same checker and require the agent's
+normal hook review/trust. No global permissions or model settings are changed.

examples/distilbert/README.md

@@ -6,4 +6,4 @@ Run `python export.py` to create the ONNX model for distilbert/distilbert-base-u
 
 ### Labels
 
-When exporting the model from huggingface transformers to ONNX, a `config.json` file is added to the chosen directory. This file has the id to label mappings and you can extract them directly to give a label to the input, as shwon in `distilbert_classification.exs`. 
+When exporting the model from huggingface transformers to ONNX, a `config.json` file is added to the chosen directory. This file has the id to label mappings and you can extract them directly to give a label to the input, as shwon in `distilbert_classification.exs`.

examples/distilbert/export.py

@@ -7,13 +7,15 @@
 
 """
 
-from transformers import DistilBertTokenizer
 from optimum.onnxruntime import ORTModelForSequenceClassification
+from transformers import DistilBertTokenizer
 
 save_directory = "./models/distilbert-onnx/"
 
 tokenizer = DistilBertTokenizer.from_pretrained("distilbert-base-uncased-finetuned-sst-2-english")
-model = ORTModelForSequenceClassification.from_pretrained("distilbert-base-uncased-finetuned-sst-2-english", export=True)
+model = ORTModelForSequenceClassification.from_pretrained(
+    "distilbert-base-uncased-finetuned-sst-2-english", export=True
+)
 print(model)
 
 model.save_pretrained(save_directory)

examples/stablelm/export.py

@@ -1,8 +1,8 @@
+import torch
 from transformers import (
     AutoModelForCausalLM,
     AutoTokenizer,
 )
-import torch
 
 tokenizer = AutoTokenizer.from_pretrained("stabilityai/stablelm-tuned-alpha-3b")
 model = AutoModelForCausalLM.from_pretrained("stabilityai/stablelm-tuned-alpha-3b")

examples/stablelm/stablelm.exs

@@ -29,9 +29,9 @@ defmodule M do
 
     x = output |> Nx.backend_transfer() |> Nx.argmax(axis: 2)
     last = x[[.., -1]] |> Nx.new_axis(0)
-    IO.inspect(last[0][0] |> Nx.to_number)
+    IO.inspect(last[0][0] |> Nx.to_number())
 
-    case Enum.member?([50278, 50279, 50277, 1, 0], last[0][0] |> Nx.to_number) do
+    case Enum.member?([50278, 50279, 50277, 1, 0], last[0][0] |> Nx.to_number()) do
       true ->
         input
 

lib/ortex/serving.ex

@@ -8,7 +8,7 @@ defmodule Ortex.Serving do
 
   ## Examples
 
-  ### Inline/serverless workflow 
+  ### Inline/serverless workflow
 
   To quickly create an `Ortex.Serving` and run it
 

mix.exs

@@ -37,6 +37,7 @@ defmodule Ortex.MixProject do
     [
       {:rustler, "~> 0.29.1"},
       {:nx, "~> 0.6"},
+      {:credo, "~> 1.7.19", only: :test, runtime: false},
       {:ex_doc, "~> 0.40.4", only: :docs, runtime: false}
     ]
   end

mix.lock

@@ -1,7 +1,10 @@
 %{
+  "bunt": {:hex, :bunt, "1.0.0", "081c2c665f086849e6d57900292b3a161727ab40431219529f13c4ddcf3e7a44", [:mix], [], "hexpm", "dc5f86aa08a5f6fa6b8096f0735c4e76d54ae5c9fa2c143e5a1fc7c1cd9bb6b5"},
   "complex": {:hex, :complex, "0.7.0", "695632ef9487517aa5d57edd1697801079d622414cb2e1a7cf538b1f9a50f205", [:mix], [], "hexpm", "0ee39c0803129f546e7f3f640da8f021c9e659402bf59da6f7f2c4848f068f8d"},
+  "credo": {:hex, :credo, "1.7.19", "cc52129665fc7c15143d47838fda0f9cd6dac9ceced7bf4da6f85fcbfe64b12a", [:mix], [{:bunt, "~> 0.2.1 or ~> 1.0", [hex: :bunt, repo: "hexpm", optional: false]}, {:file_system, "~> 0.2 or ~> 1.0", [hex: :file_system, repo: "hexpm", optional: false]}, {:jason, "~> 1.0", [hex: :jason, repo: "hexpm", optional: false]}], "hexpm", "2d8bc95d5a7bb99dd2613621d4f08c6a3575c3fd4b62e6a2b48a100352a557b8"},
   "earmark_parser": {:hex, :earmark_parser, "1.4.46", "67607a0532e810c6f630a515c548d0b24949643f168cc556303bee4cf96105c7", [:mix], [], "hexpm", "9c44636e8a1c68c62f526b2dcd85d941dbbcee7ab82cf64ba06ce28bef8e89f5"},
   "ex_doc": {:hex, :ex_doc, "0.40.4", "66f2e42bf588594d5a8aab31cad87f2ddad09d0da1b1a2f379340ec2c2e497cb", [:mix], [{:earmark_parser, "~> 1.4.46", [hex: :earmark_parser, repo: "hexpm", optional: false]}, {:makeup_c, ">= 0.1.0", [hex: :makeup_c, repo: "hexpm", optional: true]}, {:makeup_elixir, "~> 0.14 or ~> 1.0", [hex: :makeup_elixir, repo: "hexpm", optional: false]}, {:makeup_erlang, "~> 0.1 or ~> 1.0", [hex: :makeup_erlang, repo: "hexpm", optional: false]}, {:makeup_html, ">= 0.1.0", [hex: :makeup_html, repo: "hexpm", optional: true]}], "hexpm", "6222b9e423d76584ee34df2c82a5ed72c2d53dc153f7f483ad28b378694186cc"},
+  "file_system": {:hex, :file_system, "1.1.1", "31864f4685b0148f25bd3fbef2b1228457c0c89024ad67f7a81a3ffbc0bbad3a", [:mix], [], "hexpm", "7a15ff97dfe526aeefb090a7a9d3d03aa907e100e262a0f8f7746b78f8f87a5d"},
   "jason": {:hex, :jason, "1.4.5", "2e3a008590b0b8d7388c20293e9dcc9cf3e5d642fd2a114e4cbbb52e595d940a", [:mix], [{:decimal, "~> 1.0 or ~> 2.0 or ~> 3.0", [hex: :decimal, repo: "hexpm", optional: true]}], "hexpm", "b0c823996102bcd0239b3c2444eb00409b72f6a140c1950bc8b457d836b30684"},
   "makeup": {:hex, :makeup, "1.2.3", "c3ef0ff83acf305384acba0f70bff87d8732502baa1eb3eb081d8c154fbe18f5", [:mix], [{:nimble_parsec, "~> 1.4", [hex: :nimble_parsec, repo: "hexpm", optional: false]}], "hexpm", "15970751b7021c2bcd7ed56a54b3def23affd3969fb7c214dffdf28e97ce6f76"},
   "makeup_elixir": {:hex, :makeup_elixir, "1.0.1", "e928a4f984e795e41e3abd27bfc09f51db16ab8ba1aebdba2b3a575437efafc2", [:mix], [{:makeup, "~> 1.0", [hex: :makeup, repo: "hexpm", optional: false]}, {:nimble_parsec, "~> 1.2.3 or ~> 1.3", [hex: :nimble_parsec, repo: "hexpm", optional: false]}], "hexpm", "7284900d412a3e5cfd97fdaed4f5ed389b8f2b4cb49efc0eb3bd10e2febf9507"},

native/ortex/src/lib.rs

@@ -30,20 +30,15 @@ fn init(
 }
 
 #[rustler::nif]
-fn show_session(
-    model: ResourceArc<model::OrtexModel>,
-) -> NifResult<(
-    Vec<(String, String, Option<Vec<i64>>)>,
-    Vec<(String, String, Option<Vec<i64>>)>,
-)> {
+fn show_session(model: ResourceArc<model::OrtexModel>) -> NifResult<model::SessionDescription> {
     Ok(model::show(model))
 }
 
 #[rustler::nif(schedule = "DirtyIo")]
 fn run(
     model: ResourceArc<model::OrtexModel>,
     inputs: Vec<ResourceArc<OrtexTensor>>,
-) -> NifResult<Vec<(ResourceArc<OrtexTensor>, Vec<usize>, Atom, usize)>> {
+) -> NifResult<Vec<model::TensorOutput>> {
     model::run(model, inputs).map_err(|e| rustler::Error::Term(Box::new(e.to_string())))
 }
 
@@ -71,7 +66,7 @@ fn to_binary<'a>(
 }
 
 #[rustler::nif]
-pub fn slice<'a>(
+pub fn slice(
     tensor: ResourceArc<OrtexTensor>,
     start_indicies: Vec<isize>,
     lengths: Vec<isize>,
@@ -85,15 +80,15 @@ pub fn slice<'a>(
 }
 
 #[rustler::nif]
-pub fn reshape<'a>(
+pub fn reshape(
     tensor: ResourceArc<OrtexTensor>,
     shape: Vec<usize>,
 ) -> NifResult<ResourceArc<OrtexTensor>> {
     Ok(ResourceArc::new(tensor.reshape(shape)?))
 }
 
 #[rustler::nif]
-pub fn concatenate<'a>(
+pub fn concatenate(
     tensors: Vec<ResourceArc<OrtexTensor>>,
     dtype: Term,
     axis: i32,
@@ -116,9 +111,14 @@ rustler::init!(
         reshape,
         concatenate
     ],
-    load = |env: Env, _| {
-        rustler::resource!(OrtexModel, env);
-        rustler::resource!(OrtexTensor, env);
-        true
-    }
+    load = load
 );
+
+// Rustler 0.29 expands resource registration into impls inside this callback.
+// Remove this compatibility allowance with the separate Rustler migration.
+#[allow(non_local_definitions)]
+fn load(env: Env, _info: Term) -> bool {
+    rustler::resource!(OrtexModel, env);
+    rustler::resource!(OrtexTensor, env);
+    true
+}

native/ortex/src/model.rs

@@ -19,6 +19,10 @@ use rustler::Atom;
 
 /// Holds the model state which include onnxruntime session and environment. All
 /// are threadsafe so this can be called concurrently from the beam.
+pub type Descriptor = (String, String, Option<Vec<i64>>);
+pub type SessionDescription = (Vec<Descriptor>, Vec<Descriptor>);
+pub type TensorOutput = (ResourceArc<OrtexTensor>, Vec<usize>, Atom, usize);
+
 pub struct OrtexModel {
     pub session: ort::Session,
 }
@@ -51,13 +55,8 @@ pub fn init(
 /// Returns input/output information about a model. The result is a Tuple of
 /// `inputs` and `outputs` with elements of `(Name, Type, Dimension)` where
 /// `Dimension` elements of -1 are dynamic.
-pub fn show(
-    model: ResourceArc<OrtexModel>,
-) -> (
-    Vec<(String, String, Option<Vec<i64>>)>,
-    Vec<(String, String, Option<Vec<i64>>)>,
-) {
-    let model: &OrtexModel = &*model;
+pub fn show(model: ResourceArc<OrtexModel>) -> SessionDescription {
+    let model: &OrtexModel = &model;
 
     let mut inputs = Vec::new();
     for input in model.session.inputs.iter() {
@@ -83,7 +82,7 @@ pub fn show(
 pub fn run(
     model: ResourceArc<OrtexModel>,
     inputs: Vec<ResourceArc<OrtexTensor>>,
-) -> Result<Vec<(ResourceArc<OrtexTensor>, Vec<usize>, Atom, usize)>, Error> {
+) -> Result<Vec<TensorOutput>, Error> {
     // Grab the session and run a forward pass with it
     let session: &ort::Session = &model.session;
 
@@ -94,7 +93,7 @@ pub fn run(
         if is_bool_input(&onnx_input.input_type) {
             // this assumes that the boolean input isn't huge -- we're cloning it twice;
             // once below, once in the try_into()
-            let boolified_input: &OrtexTensor = &derefed_input.clone().to_bool();
+            let boolified_input: &OrtexTensor = &derefed_input.clone().into_bool();
             let v: ort::SessionInputValue = boolified_input.try_into()?;
             ortified_inputs.push(v);
         } else {
@@ -109,12 +108,12 @@ pub fn run(
 
     for output_descriptor in &session.outputs {
         let output_name: &str = &output_descriptor.name;
-        let val = outputs.get(output_name).expect(
-            &format!(
+        let val = outputs.get(output_name).unwrap_or_else(|| {
+            panic!(
                 "Expected {} to be in the outputs, but didn't find it",
                 output_name
-            )[..],
-        );
+            )
+        });
 
         // NOTE: try_into impl here will implicitly map bool outputs to u8 outputs
         let ortextensor: OrtexTensor = val.try_into()?;

native/ortex/src/tensor.rs

@@ -153,8 +153,8 @@ impl OrtexTensor {
         contents
     }
 
-    pub fn slice<'a>(
-        &'a self,
+    pub fn slice(
+        &self,
         start_indicies: Vec<isize>,
         lengths: Vec<isize>,
         strides: Vec<isize>,
@@ -184,7 +184,7 @@ impl OrtexTensor {
         }
     }
 
-    pub fn to_bool(self) -> OrtexTensor {
+    pub fn into_bool(self) -> OrtexTensor {
         match self {
             OrtexTensor::u8(y) => {
                 let bool_tensor = y.to_owned().mapv(|x| match x {
@@ -205,7 +205,7 @@ impl OrtexTensor {
 
 fn slice_array<'a, T, D>(
     array: &'a Array<T, D>,
-    slice_specs: &'a Vec<(isize, Option<isize>, isize)>,
+    slice_specs: &'a [(isize, Option<isize>, isize)],
 ) -> ArrayView<'a, T, D>
 where
     D: Dimension,
@@ -216,7 +216,7 @@ where
     })
 }
 
-fn get_bytes<'a, T>(array: &'a ArrayBase<T, IxDyn>) -> &'a [u8]
+fn get_bytes<T>(array: &ArrayBase<T, IxDyn>) -> &[u8]
 where
     T: Data,
 {
@@ -343,7 +343,7 @@ macro_rules! concatenate {
     // `typ` is the actual datatype, `ort_tensor_kind` is the OrtexTensor variant
     ($tensors:expr, $axis:expr, $typ:ty, $ort_tensor_kind:ident) => {{
         type ArrayType<'a> = ArrayBase<ViewRepr<&'a $typ>, Dim<IxDynImpl>>;
-        fn filter(tensor: &OrtexTensor) -> Option<ArrayType> {
+        fn filter(tensor: &OrtexTensor) -> Option<ArrayType<'_>> {
             match tensor {
                 OrtexTensor::$ort_tensor_kind(x) => Some(x.view()),
                 _ => None,

native/ortex/src/utils.rs

@@ -14,7 +14,7 @@ use rustler::{Atom, Env, NifResult};
 use ort::{ExecutionProviderDispatch, GraphOptimizationLevel};
 
 /// A faster (unsafe) way of creating an Array from an Erlang binary
-fn initialize_from_raw_ptr<T>(ptr: *const T, shape: &[Ix]) -> ArrayViewMut<T, IxDyn> {
+fn initialize_from_raw_ptr<T>(ptr: *const T, shape: &[Ix]) -> ArrayViewMut<'_, T, IxDyn> {
     let array = unsafe { ArrayViewMut::from_shape_ptr(shape, ptr as *mut T) };
     array
 }
@@ -64,7 +64,7 @@ pub fn from_binary(
             initialize_from_raw_ptr(bin.as_ptr() as *const i64, &shape).to_owned(),
         ))),
         ("u", 8) => Ok(ResourceArc::new(OrtexTensor::u8(
-            initialize_from_raw_ptr(bin.as_ptr() as *const u8, &shape).to_owned(),
+            initialize_from_raw_ptr(bin.as_ptr(), &shape).to_owned(),
         ))),
         ("u", 16) => Ok(ResourceArc::new(OrtexTensor::u16(
             initialize_from_raw_ptr(bin.as_ptr() as *const u16, &shape).to_owned(),

pyproject.toml

@@ -0,0 +1,7 @@
+[tool.ruff]
+target-version = "py312"
+line-length = 100
+extend-exclude = ["deps", "_build", "runtime", "dist", ".venv"]
+
+[tool.ruff.lint]
+select = ["E4", "E7", "E9", "F", "I"]