Merge pull request #18 from openai/dev/scl/better-ci

better ci

.github/actions/run-rust-python-tests/action.yml

@@ -0,0 +1,69 @@
+name: Run Rust and Python tests
+
+description: Format, clippy, Rust tests (incl. doctests), build Python extension and run pytest
+
+inputs:
+  python-version:
+    description: Python version to use
+    required: false
+    default: "3.11"
+  rust-toolchain:
+    description: Rust toolchain channel
+    required: false
+    default: stable
+
+runs:
+  using: composite
+  steps:
+    - name: Setup Rust toolchain
+      uses: dtolnay/rust-toolchain@stable
+      with:
+        toolchain: ${{ inputs.rust-toolchain }}
+        components: clippy,rustfmt
+
+    - name: Setup Python
+      uses: actions/setup-python@v5
+      with:
+        python-version: ${{ inputs.python-version }}
+
+    - name: Upgrade pip
+      shell: bash
+      run: |
+        python -m pip install --upgrade pip
+
+    - name: Install test deps
+      shell: bash
+      run: |
+        python -m pip install pytest
+
+    - name: Check rustfmt
+      shell: bash
+      run: |
+        cargo fmt --all --check
+
+    - name: Run clippy
+      shell: bash
+      run: |
+        cargo clippy --all-targets --all-features -- -D warnings
+
+    - name: Run Rust tests (unit/integration)
+      shell: bash
+      run: |
+        cargo test --all-targets --all-features
+
+    - name: Run Rust doctests
+      shell: bash
+      run: |
+        cargo test --doc
+
+    - name: Build and install Python package
+      shell: bash
+      run: |
+        pip install .
+
+    - name: Run pytest
+      shell: bash
+      env:
+        PYTHONUTF8: "1"
+      run: |
+        pytest -q

.github/workflows/CI.yml

@@ -19,7 +19,23 @@ permissions:
   contents: read
 
 jobs:
+  tests:
+    name: Tests (fmt, clippy, cargo test, doctest, pytest)
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, macos-14, windows-latest]
+    steps:
+      - uses: actions/checkout@v4
+      - name: Run composite test suite
+        uses: ./.github/actions/run-rust-python-tests
+        with:
+          python-version: "3.11"
+          rust-toolchain: stable
+
   linux:
+    needs: [tests]
     runs-on: ${{ matrix.platform.runner }}
     strategy:
       matrix:
@@ -54,6 +70,7 @@ jobs:
           path: dist
 
   musllinux:
+    needs: [tests]
     runs-on: ${{ matrix.platform.runner }}
     strategy:
       matrix:
@@ -88,6 +105,7 @@ jobs:
           path: dist
 
   windows:
+    needs: [tests]
     runs-on: ${{ matrix.platform.runner }}
     strategy:
       matrix:
@@ -115,6 +133,7 @@ jobs:
           path: dist
 
   macos:
+    needs: [tests]
     runs-on: ${{ matrix.platform.runner }}
     strategy:
       matrix:
@@ -141,6 +160,7 @@ jobs:
           path: dist
 
   sdist:
+    needs: [tests]
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4

src/py_module.rs

@@ -12,9 +12,6 @@
 //! A thin, typed, user-facing Python wrapper around these low-level bindings is
 //! provided in `harmony/__init__.py`.
 
-// Only compile when the `python-binding` feature is enabled.
-#![cfg(feature = "python-binding")]
-
 use pyo3::prelude::*;
 
 // We need the `Python` type later on.
@@ -34,8 +31,6 @@ use crate::{
     load_harmony_encoding, HarmonyEncodingName,
 };
 
-use serde_json;
-
 /// A thin PyO3 wrapper around the Rust `HarmonyEncoding` struct.
 #[pyclass]
 struct PyHarmonyEncoding {
@@ -393,8 +388,7 @@ fn openai_harmony(_py: Python<'_>, m: &Bound<'_, PyModule>) -> PyResult<()> {
             "python" => ToolNamespaceConfig::python(),
             _ => {
                 return Err(PyErr::new::<pyo3::exceptions::PyValueError, _>(format!(
-                    "Unknown tool namespace: {}",
-                    tool
+                    "Unknown tool namespace: {tool}"
                 )));
             }
         };

src/tests.rs

@@ -1,3 +1,5 @@
+use std::path::Path;
+
 use crate::{
     chat::{
         Author, Conversation, DeveloperContent, Message, ReasoningEffort, Role, SystemContent,
@@ -10,12 +12,25 @@ use crate::{
 use pretty_assertions::{assert_eq, Comparison};
 use serde_json::json;
 
-fn parse_tokens(text: &str) -> Vec<Rank> {
-    text.split_whitespace()
+fn parse_tokens(text: impl AsRef<str>) -> Vec<Rank> {
+    text.as_ref()
+        .split_whitespace()
         .map(|s| s.parse().unwrap())
         .collect()
 }
 
+fn load_test_data(path: impl AsRef<Path>) -> String {
+    // on windows, we need to replace \r\n with \n
+    let cargo_manifest_dir = Path::new(env!("CARGO_MANIFEST_DIR"));
+    let src_dir = cargo_manifest_dir.join("src");
+    let path = src_dir.join(path);
+    std::fs::read_to_string(path)
+        .unwrap()
+        .replace("\r\n", "\n")
+        .trim_end()
+        .to_string()
+}
+
 const ENCODINGS: [HarmonyEncodingName; 1] = [HarmonyEncodingName::HarmonyGptOss];
 
 #[test]
@@ -25,7 +40,7 @@ fn test_simple_convo() {
         let expected_tokens = encoding
             .tokenizer
             .encode(
-                include_str!("../test-data/test_simple_convo.txt").trim_end(),
+                load_test_data("../test-data/test_simple_convo.txt").as_str(),
                 &encoding.tokenizer.special_tokens(),
             )
             .0;
@@ -50,45 +65,42 @@ fn test_simple_convo_with_effort() {
     let test_cases = [
         (
             ReasoningEffort::Low,
-            include_str!("../test-data/test_simple_convo_low_effort.txt"),
+            load_test_data("../test-data/test_simple_convo_low_effort.txt"),
             true,
         ),
         (
             ReasoningEffort::Medium,
-            include_str!("../test-data/test_simple_convo_medium_effort.txt"),
+            load_test_data("../test-data/test_simple_convo_medium_effort.txt"),
             true,
         ),
         (
             ReasoningEffort::High,
-            include_str!("../test-data/test_simple_convo_high_effort.txt"),
+            load_test_data("../test-data/test_simple_convo_high_effort.txt"),
             true,
         ),
         (
             ReasoningEffort::Low,
-            include_str!("../test-data/test_simple_convo_low_effort_no_instruction.txt"),
+            load_test_data("../test-data/test_simple_convo_low_effort_no_instruction.txt"),
             false,
         ),
         (
             ReasoningEffort::Medium,
-            include_str!("../test-data/test_simple_convo_medium_effort_no_instruction.txt"),
+            load_test_data("../test-data/test_simple_convo_medium_effort_no_instruction.txt"),
             false,
         ),
         (
             ReasoningEffort::High,
-            include_str!("../test-data/test_simple_convo_high_effort_no_instruction.txt"),
+            load_test_data("../test-data/test_simple_convo_high_effort_no_instruction.txt"),
             false,
         ),
     ];
 
     for encoding_name in ENCODINGS {
         let encoding = load_harmony_encoding(encoding_name).unwrap();
-        for (effort, expected_text, use_instruction) in test_cases {
+        for &(effort, ref expected_text, use_instruction) in &test_cases {
             let expected_tokens = encoding
                 .tokenizer
-                .encode(
-                    expected_text.trim_end(),
-                    &encoding.tokenizer.special_tokens(),
-                )
+                .encode(expected_text.as_str(), &encoding.tokenizer.special_tokens())
                 .0;
             let sys = SystemContent::new()
                 .with_model_identity("You are ChatGPT, a large language model trained by OpenAI.")
@@ -123,8 +135,8 @@ fn test_simple_convo_with_effort() {
 
 #[test]
 fn test_simple_reasoning_response() {
-    let expected_tokens = parse_tokens(include_str!(
-        "../test-data/test_simple_reasoning_response.txt"
+    let expected_tokens = parse_tokens(load_test_data(
+        "../test-data/test_simple_reasoning_response.txt",
     ));
     for encoding_name in ENCODINGS {
         let encoding = load_harmony_encoding(encoding_name).unwrap();
@@ -180,7 +192,7 @@ fn test_reasoning_system_message() {
         let expected = encoding
             .tokenizer
             .encode(
-                include_str!("../test-data/test_reasoning_system_message.txt").trim_end(),
+                load_test_data("../test-data/test_reasoning_system_message.txt").as_str(),
                 &encoding.tokenizer.special_tokens(),
             )
             .0;
@@ -211,8 +223,8 @@ fn test_reasoning_system_message_no_instruction() {
         let expected = encoding
             .tokenizer
             .encode(
-                include_str!("../test-data/test_reasoning_system_message_no_instruction.txt")
-                    .trim_end(),
+                load_test_data("../test-data/test_reasoning_system_message_no_instruction.txt")
+                    .as_str(),
                 &encoding.tokenizer.special_tokens(),
             )
             .0;
@@ -245,8 +257,8 @@ fn test_reasoning_system_message_with_dates() {
         let expected = encoding
             .tokenizer
             .encode(
-                include_str!("../test-data/test_reasoning_system_message_with_dates.txt")
-                    .trim_end(),
+                load_test_data("../test-data/test_reasoning_system_message_with_dates.txt")
+                    .as_str(),
                 &encoding.tokenizer.special_tokens(),
             )
             .0;
@@ -275,8 +287,7 @@ fn test_reasoning_system_message_with_dates() {
 #[test]
 fn test_render_functions_with_parameters() {
     let encoding = load_harmony_encoding(HarmonyEncodingName::HarmonyGptOss).unwrap();
-    let expected_output =
-        include_str!("../test-data/test_render_functions_with_parameters.txt").trim_end();
+    let expected_output = load_test_data("../test-data/test_render_functions_with_parameters.txt");
 
     let sys = SystemContent::new()
         .with_reasoning_effort(ReasoningEffort::High)
@@ -382,7 +393,7 @@ fn test_render_functions_with_parameters() {
 #[test]
 fn test_browser_and_python_tool() {
     let encoding = load_harmony_encoding(HarmonyEncodingName::HarmonyGptOss).unwrap();
-    let expected_output = include_str!("../test-data/test_browser_and_python_tool.txt").trim_end();
+    let expected_output = load_test_data("../test-data/test_browser_and_python_tool.txt");
 
     let convo = Conversation::from_messages([Message::from_role_and_content(
         Role::System,
@@ -403,7 +414,7 @@ fn test_browser_and_python_tool() {
 #[test]
 fn test_dropping_cot_by_default() {
     let encoding = load_harmony_encoding(HarmonyEncodingName::HarmonyGptOss).unwrap();
-    let expected_output = include_str!("../test-data/test_dropping_cot_by_default.txt").trim_end();
+    let expected_output = load_test_data("../test-data/test_dropping_cot_by_default.txt");
 
     let convo = Conversation::from_messages([
         Message::from_role_and_content(Role::User, "What is 2 + 2?"),
@@ -433,8 +444,7 @@ fn test_dropping_cot_by_default() {
 #[test]
 fn test_does_not_drop_if_ongoing_analysis() {
     let encoding = load_harmony_encoding(HarmonyEncodingName::HarmonyGptOss).unwrap();
-    let expected_output =
-        include_str!("../test-data/test_does_not_drop_if_ongoing_analysis.txt").trim_end();
+    let expected_output = load_test_data("../test-data/test_does_not_drop_if_ongoing_analysis.txt");
 
     let convo = Conversation::from_messages([
         Message::from_role_and_content(Role::User, "What is the weather in SF?"),
@@ -470,7 +480,7 @@ fn test_does_not_drop_if_ongoing_analysis() {
 #[test]
 fn test_preserve_cot() {
     let encoding = load_harmony_encoding(HarmonyEncodingName::HarmonyGptOss).unwrap();
-    let expected_output = include_str!("../test-data/test_preserve_cot.txt").trim_end();
+    let expected_output = load_test_data("../test-data/test_preserve_cot.txt");
 
     let convo = Conversation::from_messages([
         Message::from_role_and_content(Role::User, "What is 2 + 2?"),
@@ -534,10 +544,10 @@ fn test_decode_utf8_invalid_token() {
 #[test]
 fn test_tool_response_parsing() {
     let encoding = load_harmony_encoding(HarmonyEncodingName::HarmonyGptOss).unwrap();
-    let text_tokens = include_str!("../test-data/test_tool_response_parsing.txt").trim_end();
+    let text_tokens = load_test_data("../test-data/test_tool_response_parsing.txt");
     let tokens = encoding
         .tokenizer
-        .encode(text_tokens, &encoding.tokenizer.special_tokens())
+        .encode(&text_tokens, &encoding.tokenizer.special_tokens())
         .0;
 
     let expected_message = Message::from_author_and_content(
@@ -616,10 +626,10 @@ fn test_invalid_utf8_decoding() {
 #[test]
 fn test_streamable_parser() {
     let encoding = load_harmony_encoding(HarmonyEncodingName::HarmonyGptOss).unwrap();
-    let text = include_str!("../test-data/test_streamable_parser.txt").trim_end();
+    let text = load_test_data("../test-data/test_streamable_parser.txt");
     let tokens = encoding
         .tokenizer
-        .encode(text, &encoding.tokenizer.special_tokens())
+        .encode(&text, &encoding.tokenizer.special_tokens())
         .0;
     let mut parser =
         crate::encoding::StreamableParser::new(encoding.clone(), Some(Role::Assistant)).unwrap();

src/wasm_module.rs

@@ -1,5 +1,3 @@
-#![cfg(feature = "wasm-binding")]
-
 use wasm_bindgen::prelude::*;
 
 use crate::{
@@ -9,8 +7,6 @@ use crate::{
 };
 
 use serde::Deserialize;
-use serde_json;
-use serde_wasm_bindgen;
 
 #[wasm_bindgen]
 extern "C" {
@@ -335,8 +331,7 @@ pub fn get_tool_namespace_config(tool: &str) -> Result<JsValue, JsValue> {
         "python" => ToolNamespaceConfig::python(),
         _ => {
             return Err(JsValue::from_str(&format!(
-                "Unknown tool namespace: {}",
-                tool
+                "Unknown tool namespace: {tool}"
             )))
         }
     };

tests/test_harmony.py

@@ -59,7 +59,7 @@ def _assert_tokens_eq(encoding, expected: List[int], actual: List[int]):  # type
 
 
 def read_expected_tokens(file_path: Path) -> List[int]:
-    with open(file_path, "r") as f:
+    with open(file_path, "r", encoding="utf-8") as f:
         return [int(x) for x in f.read().split()]
 
 
@@ -78,7 +78,9 @@ def test_simple_convo(encoding_name):
     encoding = load_harmony_encoding(encoding_name)
 
     expected_text = (
-        (ROOT_DIR / "test-data" / "test_simple_convo.txt").read_text().rstrip()
+        (ROOT_DIR / "test-data" / "test_simple_convo.txt")
+        .read_text(encoding="utf-8")
+        .rstrip()
     )
     expected_tokens = encoding.encode(expected_text, allowed_special="all")
 
@@ -143,7 +145,7 @@ def test_simple_convo_with_effort(encoding_name):
     ]
 
     for effort, tokens_path, use_instruction in test_cases:
-        expected_text = tokens_path.read_text().rstrip()
+        expected_text = tokens_path.read_text(encoding="utf-8").rstrip()
         expected_tokens = encoding.encode(expected_text, allowed_special="all")
         sys = (
             SystemContent.new()
@@ -299,7 +301,7 @@ def test_reasoning_system_message(encoding_name):
 
     expected_text = (
         (ROOT_DIR / "test-data" / "test_reasoning_system_message.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
     expected = encoding.encode(expected_text, allowed_special="all")
@@ -336,7 +338,7 @@ def test_reasoning_system_message_no_instruction(encoding_name):
 
     expected_text = (
         (ROOT_DIR / "test-data" / "test_reasoning_system_message_no_instruction.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
     expected = encoding.encode(expected_text, allowed_special="all")
@@ -376,7 +378,7 @@ def test_reasoning_system_message_with_dates(encoding_name):
 
     expected_text = (
         (ROOT_DIR / "test-data" / "test_reasoning_system_message_with_dates.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
     expected = encoding.encode(expected_text, allowed_special="all")
@@ -409,7 +411,7 @@ def test_render_functions_with_parameters():
 
     expected_output = (
         (ROOT_DIR / "test-data" / "test_render_functions_with_parameters.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
 
@@ -526,7 +528,9 @@ def test_render_functions_with_parameters():
 def test_no_tools():
     encoding = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)
     expected_output = (
-        (ROOT_DIR / "test-data" / "test_no_tools.txt").read_text().rstrip()
+        (ROOT_DIR / "test-data" / "test_no_tools.txt")
+        .read_text(encoding="utf-8")
+        .rstrip()
     )
 
     convo = Conversation.from_messages(
@@ -546,7 +550,9 @@ def test_no_tools():
 def test_browser_tool_only():
     encoding = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)
     expected_output = (
-        (ROOT_DIR / "test-data" / "test_browser_tool_only.txt").read_text().rstrip()
+        (ROOT_DIR / "test-data" / "test_browser_tool_only.txt")
+        .read_text(encoding="utf-8")
+        .rstrip()
     )
 
     convo = Conversation.from_messages(
@@ -569,7 +575,7 @@ def test_browser_and_function_tool():
     encoding = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)
     expected_output = (
         (ROOT_DIR / "test-data" / "test_browser_and_function_tool.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
 
@@ -611,7 +617,7 @@ def test_browser_and_python_tool():
     encoding = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)
     expected_output = (
         (ROOT_DIR / "test-data" / "test_browser_and_python_tool.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
 
@@ -637,7 +643,7 @@ def test_dropping_cot_by_default():
 
     expected_output = (
         (ROOT_DIR / "test-data" / "test_dropping_cot_by_default.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
 
@@ -667,7 +673,7 @@ def test_does_not_drop_if_ongoing_analysis():
 
     expected_output = (
         (ROOT_DIR / "test-data" / "test_does_not_drop_if_ongoing_analysis.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
 
@@ -702,7 +708,9 @@ def test_preserve_cot():
     encoding = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)
 
     expected_output = (
-        (ROOT_DIR / "test-data" / "test_preserve_cot.txt").read_text().rstrip()
+        (ROOT_DIR / "test-data" / "test_preserve_cot.txt")
+        .read_text(encoding="utf-8")
+        .rstrip()
     )
 
     convo = Conversation.from_messages(
@@ -738,7 +746,7 @@ def test_keep_analysis_between_final_messages():
 
     expected_output = (
         (ROOT_DIR / "test-data" / "test_keep_analysis_between_finals.txt")
-        .read_text()
+        .read_text(encoding="utf-8")
         .rstrip()
     )
 
@@ -880,7 +888,9 @@ def test_tool_response_parsing():
     encoding = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)
 
     text_tokens = (
-        (ROOT_DIR / "test-data" / "test_tool_response_parsing.txt").read_text().rstrip()
+        (ROOT_DIR / "test-data" / "test_tool_response_parsing.txt")
+        .read_text(encoding="utf-8")
+        .rstrip()
     )
 
     expected_message = (
@@ -904,7 +914,9 @@ def test_streamable_parser():
     encoding = load_harmony_encoding(HarmonyEncodingName.HARMONY_GPT_OSS)
 
     text_tokens = (
-        (ROOT_DIR / "test-data" / "test_streamable_parser.txt").read_text().rstrip()
+        (ROOT_DIR / "test-data" / "test_streamable_parser.txt")
+        .read_text(encoding="utf-8")
+        .rstrip()
     )
 
     tokens = encoding.encode(text_tokens, allowed_special="all")