telemetry-anomdet 0.1.0__tar.gz → 0.3.0__tar.gz

{telemetry_anomdet-0.1.0 → telemetry_anomdet-0.3.0}/LICENSE

@@ -186,7 +186,7 @@
       same "printed page" as the copyright notice for easier
       identification within third-party archives.
 
-   Copyright 2026 Telemetry Anomdet
+   Copyright 2025-2026 Julio Anzaldo
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

telemetry_anomdet-0.3.0/PKG-INFO

@@ -0,0 +1,226 @@
+Metadata-Version: 2.4
+Name: telemetry-anomdet
+Version: 0.3.0
+Summary: Spacecraft telemetry anomaly detection: classical and graph deep detectors, label-free thresholding, and distillation to C for flight hardware.
+Author-email: Julio Anzaldo <anzaldojuliojesus@gmail.com>
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/JulioAnzaldo/telemetry-anomdet
+Project-URL: Repository, https://github.com/JulioAnzaldo/telemetry-anomdet
+Keywords: telemetry,anomaly-detection,machine-learning,python
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Intended Audience :: Developers
+Classifier: Natural Language :: English
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3 :: Only
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Classifier: Programming Language :: C
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Scientific/Engineering :: Information Analysis
+Classifier: Topic :: Scientific/Engineering :: Astronomy
+Classifier: Topic :: Software Development :: Libraries
+Classifier: Topic :: Software Development :: Embedded Systems
+Classifier: Topic :: Software Development :: Code Generators
+Classifier: Topic :: System :: Monitoring
+Classifier: Typing :: Typed
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: pandas<3.0.0,>=2.2.2
+Requires-Dist: numpy<3.0,>=2.1
+Requires-Dist: scikit-learn<2.0,>=1.3
+Provides-Extra: ccsds
+Requires-Dist: ccsdspy<3.0,>=2.0.0; extra == "ccsds"
+Provides-Extra: deep
+Requires-Dist: torch>=2.0; extra == "deep"
+Provides-Extra: viz
+Requires-Dist: matplotlib>=3.7; extra == "viz"
+Provides-Extra: track
+Requires-Dist: wandb>=0.17; extra == "track"
+Dynamic: license-file
+
+<h2 align="center">telemetry-anomdet</h2>
+
+<p align="center">
+  <a href="https://julioanzaldo.github.io/telemetry-anomdet/"><img alt="Documentation" src="https://img.shields.io/badge/Docs-Online-blue"></a>
+  <a href="https://test.pypi.org/project/telemetry-anomdet/"><img alt="Test PyPI" src="https://img.shields.io/badge/Test-PyPI-yellow"></a>
+  <a href="https://pypi.org/project/telemetry-anomdet/"><img alt="PyPI" src="https://img.shields.io/pypi/v/telemetry-anomdet?label=PyPi"></a>
+  <a href="https://github.com/JulioAnzaldo/telemetry-anomdet/actions/workflows/ci.yml"><img alt="Testing" src="https://img.shields.io/github/actions/workflow/status/JulioAnzaldo/telemetry-anomdet/ci.yml?branch=main&label=Testing&logo=github&logoColor=white"></a>
+  <a href="https://github.com/JulioAnzaldo/telemetry-anomdet/blob/main/LICENSE"><img alt="License" src="https://img.shields.io/badge/License-Apache%202.0-blue"></a>
+</p>
+
+**telemetry-anomdet** is an open-source anomaly detection toolkit for spacecraft telemetry. It runs classical and graph-based deep detectors behind one interface, selects alarm thresholds without labels, points each alarm at the channels that raised it, and distills a trained detector down to Power of Ten conformant C that runs on flight hardware. SHAP attribution and LLM-generated diagnostic reports are on the roadmap.
+
+Benchmarked on SMAP and MSL (NASA) and ESA-ADB (ESA).
+
+---
+
+<table>
+  <tr>
+    <td align="center"><a href="https://julioanzaldo.github.io/telemetry-anomdet/getting_started.html"><b>Getting Started</b></a><br>Installation and setup</td>
+    <td align="center"><a href="https://julioanzaldo.github.io/telemetry-anomdet/api/models/base.html"><b>API Reference</b></a><br>BaseDetector, models, ensemble</td>
+    <td align="center"><a href="https://julioanzaldo.github.io/telemetry-anomdet/tutorials/real_time_example.html"><b>Tutorial</b></a><br>End-to-end detection example</td>
+    <td align="center"><a href="https://julioanzaldo.github.io/telemetry-anomdet/user_guide/pipeline_overview.html"><b>Pipeline Overview</b></a><br>Ingest, detect, explain</td>
+    <td align="center"><a href="https://github.com/JulioAnzaldo/telemetry-anomdet/discussions"><b>Discussions</b></a><br>Questions and feedback</td>
+  </tr>
+</table>
+
+---
+
+## Quick Install
+
+```bash
+# recommended
+uv add telemetry-anomdet
+
+# or with pip
+pip install telemetry-anomdet
+```
+
+The base install is torch-free. The deep detectors (`GDN`, `KANGDN`) need the
+`deep` extra:
+
+```bash
+uv add "telemetry-anomdet[deep]"
+```
+
+Plotting, including the learned sensor graph, needs the `viz` extra.
+
+## Example Usage
+
+### Fit the Ensemble on Nominal Telemetry
+
+All detectors are trained on **nominal data only**, no anomaly labels used during training. Labels are used exclusively for evaluation.
+
+```python
+import numpy as np
+from telemetry_anomdet.models.unsupervised.pca import PCAAnomaly
+from telemetry_anomdet.models.unsupervised.kmeans import KMeansAnomaly
+from telemetry_anomdet.models.ensemble import AnomalyEnsemble
+
+# X_train: (n_windows, window_size, n_features), produced by windowify()
+X_train = ...
+
+ensemble = AnomalyEnsemble(
+    models = {
+        "pca":    PCAAnomaly(n_components = 10),
+        "kmeans": KMeansAnomaly(n_clusters = 5, scale = True),
+    },
+    combine = "mean",
+    normalize = "robust",
+    percentile = 95.0,
+)
+
+ensemble.fit(X_train)
+```
+
+### Inference and Anomaly Flags
+
+```python
+scores = ensemble.decision_function(X_test)   # higher = more anomalous
+flags  = ensemble.is_anomaly(X_test)          # boolean mask at training threshold
+
+# Runtime sensitivity override, no retraining needed
+flags_strict = ensemble.is_anomaly(X_test, percentile = 98.0)
+flags_custom  = ensemble.is_anomaly(X_test, threshold = 0.75)
+```
+
+### Per-Model Score Decomposition
+
+`score_components()` returns raw per-model scores before normalization or combination. This is the SHAP hook: perturb input channels, measure how each sensor drives each model's score independently.
+
+```python
+components = ensemble.score_components(X_test)
+# {"pca": array(100,), "kmeans": array(100,)}
+```
+
+### Getting the SMAP dataset
+
+The benchmarks and examples need a local copy of the NASA SMAP/MSL data, which
+is not redistributed with the toolkit. From the root of the repo:
+
+```bash
+pip install kaggle
+
+# Requires a Kaggle API key in ~/.kaggle/kaggle.json
+kaggle datasets download -d patrickfleith/nasa-anomaly-detection-dataset-smap-msl \
+  && mv nasa-anomaly-detection-dataset-smap-msl.zip data.zip \
+  && unzip -o data.zip \
+  && rm data.zip \
+  && mv data/data tmp && rm -r data && mv tmp data
+```
+
+That leaves `data/train/`, `data/test/` and `labeled_anomalies.csv`. Point the
+examples at it:
+
+```bash
+export TAD_SMAP_DIR=data          # Windows: $env:TAD_SMAP_DIR = "data"
+python examples/smap_demo.py
+python examples/smap_benchmark.py
+```
+
+Datasets are never committed; `data/` is gitignored.
+
+### Coming in Phase 3+
+
+```python
+# Per-channel SHAP attribution (Phase 3)
+# shap_values = explainer.explain(ensemble, X_anomalous)
+# {"sensor_01": 0.42, "sensor_02": 0.18, ...}
+
+# LLM diagnostic report (Phase 4)
+# report = llm_engine.explain(shap_values, context, channel_names)
+# DiagnosticReport(
+#     anomaly_type = "power subsystem fault",
+#     severity = "high",
+#     primary_channels = ["sensor_01", "sensor_07"],
+#     explanation = "...",
+#     recommended_action = "...",
+#     confidence = 0.87,
+# )
+```
+
+## Features
+
+**Available now**
+
+- **One interface for every detector.** Classical and deep detectors share the same `fit` / `decision_function` / `predict` / `is_anomaly` API, so stacking or swapping models needs no per-model glue.
+- **Graph deviation detectors.** `GDN` and `KANGDN` forecast each channel from its learned top-k neighbours and score the deviation, catching readings that are individually plausible but wrong in relation to each other. `feat_mode` selects how `KANGDN` encodes each channel's history (`linear`, `kan`, `ar_kan`, `ar_kan_residual`).
+- **Which sensor, not just when.** `channel_deviations` and `dominant_channels` attribute each alarm to the channels driving it, and `plot_sensor_graph` draws the graph the detector learned.
+- **Alarms without labels.** `threshold_for_budget` and `dynamic_threshold` pick an operating point from the score distribution alone. Operations can state an alarm rate; they cannot state a recall they have no way to observe.
+- **Honest evaluation.** Event-level scoring alongside the conventional point-adjusted F1, plus a random baseline row in the benchmark. ESA-ADB is scored by its own protocol, with F0.5 as the headline metric.
+- **Runs on flight hardware.** A fitted `KANGDN` distills to a torch-free NumPy evaluator, then to Power of Ten conformant C with golden vectors, for every `feat_mode`. Host and ESP32-S3 targets are in [`targets/`](targets/).
+- **Stacking ensemble** with robust (median + IQR) score normalization, built for anomaly scores that are extreme by definition.
+- **Per-model score decomposition** via `score_components()`, the hook that enables per-channel attribution without retraining.
+- **Spacecraft-native ingestion.** SMAP, MSL, ESA-ADB and CSV load straight into a long-form `TelemetryDataset`, no schema wrangling. The ESA-ADB loader reads the official archive directly, with no preprocessing step; its payloads are pickled, so load only that archive, never untrusted files.
+- **Runtime sensitivity control.** `is_anomaly()` accepts a percentile or threshold override, so operators re-tune without retraining.
+
+**On the roadmap**
+
+- `TranAD`: transformer-based sequence reconstruction
+- `SHAPExplainer`: per-channel attribution over `score_components()` (Phase 3)
+- LLM diagnostic reports, SHAP chart supplied as an image (Phase 4)
+- Human-in-the-loop threshold feedback (Phase 5)
+
+## AI Transparency
+
+AI assistants are used while writing and maintaining this project's documentation.
+
+What that does not change:
+
+- **Numbers come from runs.** Benchmark figures are produced by [`examples/smap_benchmark.py`](examples/smap_benchmark.py) and [`examples/esa_benchmark.py`](examples/esa_benchmark.py), and the hardware results in the deployment docs are measured on an ESP32-S3 by the conformance harness in [`targets/`](targets/). Nothing quoted in these docs is estimated or inferred.
+- **Behavior claims are backed by tests.** The suite gates CI, and the documented API is covered by it.
+
+Everything published here is reviewed before it lands.
+
+## Getting Help
+
+### Discussions
+To suggest improvements, or ask for help, please see [GitHub Discussions](https://github.com/JulioAnzaldo/telemetry-anomdet/discussions)
+
+### Bug reports
+To submit a report on any bugs or issues, [open an issue here](https://github.com/JulioAnzaldo/telemetry-anomdet/issues)

{telemetry_anomdet-0.1.0 → telemetry_anomdet-0.3.0}/README.md

@@ -1,14 +1,16 @@
 <h2 align="center">telemetry-anomdet</h2>
 
 <p align="center">
-  <a href="https://julioanzaldo.github.io/telemetry-anomdet/"><img alt="Documentation" src="https://img.shields.io/badge/docs-online-blue"></a>
+  <a href="https://julioanzaldo.github.io/telemetry-anomdet/"><img alt="Documentation" src="https://img.shields.io/badge/Docs-Online-blue"></a>
   <a href="https://test.pypi.org/project/telemetry-anomdet/"><img alt="Test PyPI" src="https://img.shields.io/badge/Test-PyPI-yellow"></a>
-  <a href="https://pypi.org/project/telemetry-anomdet/"><img alt="PyPI" src="https://img.shields.io/pypi/v/telemetry-anomdet"></a>
-  <a href="https://github.com/JulioAnzaldo/telemetry-anomdet/actions/workflows/ci.yml"><img alt="CI" src="https://github.com/JulioAnzaldo/telemetry-anomdet/actions/workflows/ci.yml/badge.svg?branch=main"></a>
-  <a href="https://github.com/JulioAnzaldo/telemetry-anomdet/blob/main/LICENSE"><img alt="License" src="https://img.shields.io/badge/license-Apache%202.0-blue"></a>
+  <a href="https://pypi.org/project/telemetry-anomdet/"><img alt="PyPI" src="https://img.shields.io/pypi/v/telemetry-anomdet?label=PyPi"></a>
+  <a href="https://github.com/JulioAnzaldo/telemetry-anomdet/actions/workflows/ci.yml"><img alt="Testing" src="https://img.shields.io/github/actions/workflow/status/JulioAnzaldo/telemetry-anomdet/ci.yml?branch=main&label=Testing&logo=github&logoColor=white"></a>
+  <a href="https://github.com/JulioAnzaldo/telemetry-anomdet/blob/main/LICENSE"><img alt="License" src="https://img.shields.io/badge/License-Apache%202.0-blue"></a>
 </p>
 
-**telemetry-anomdet** is an open-source anomaly detection toolkit for spacecraft telemetry. It runs a stacking ensemble of classical and deep learning detectors with per-channel SHAP attribution, SymTorch symbolic fault expressions, and LLM-generated diagnostic reports.
+**telemetry-anomdet** is an open-source anomaly detection toolkit for spacecraft telemetry. It runs classical and graph-based deep detectors behind one interface, selects alarm thresholds without labels, points each alarm at the channels that raised it, and distills a trained detector down to Power of Ten conformant C that runs on flight hardware. SHAP attribution and LLM-generated diagnostic reports are on the roadmap.
+
+Benchmarked on SMAP and MSL (NASA) and ESA-ADB (ESA).
 
 ---
 
@@ -34,6 +36,15 @@ uv add telemetry-anomdet
 pip install telemetry-anomdet
 ```
 
+The base install is torch-free. The deep detectors (`GDN`, `KANGDN`) need the
+`deep` extra:
+
+```bash
+uv add "telemetry-anomdet[deep]"
+```
+
+Plotting, including the learned sensor graph, needs the `viz` extra.
+
 ## Example Usage
 
 ### Fit the Ensemble on Nominal Telemetry
@@ -82,6 +93,33 @@ components = ensemble.score_components(X_test)
 # {"pca": array(100,), "kmeans": array(100,)}
 ```
 
+### Getting the SMAP dataset
+
+The benchmarks and examples need a local copy of the NASA SMAP/MSL data, which
+is not redistributed with the toolkit. From the root of the repo:
+
+```bash
+pip install kaggle
+
+# Requires a Kaggle API key in ~/.kaggle/kaggle.json
+kaggle datasets download -d patrickfleith/nasa-anomaly-detection-dataset-smap-msl \
+  && mv nasa-anomaly-detection-dataset-smap-msl.zip data.zip \
+  && unzip -o data.zip \
+  && rm data.zip \
+  && mv data/data tmp && rm -r data && mv tmp data
+```
+
+That leaves `data/train/`, `data/test/` and `labeled_anomalies.csv`. Point the
+examples at it:
+
+```bash
+export TAD_SMAP_DIR=data          # Windows: $env:TAD_SMAP_DIR = "data"
+python examples/smap_demo.py
+python examples/smap_benchmark.py
+```
+
+Datasets are never committed; `data/` is gitignored.
+
 ### Coming in Phase 3+
 
 ```python
@@ -106,19 +144,33 @@ components = ensemble.score_components(X_test)
 **Available now**
 
 - **One interface for every detector.** Classical and deep detectors share the same `fit` / `decision_function` / `predict` / `is_anomaly` API, so stacking or swapping models needs no per-model glue.
+- **Graph deviation detectors.** `GDN` and `KANGDN` forecast each channel from its learned top-k neighbours and score the deviation, catching readings that are individually plausible but wrong in relation to each other. `feat_mode` selects how `KANGDN` encodes each channel's history (`linear`, `kan`, `ar_kan`, `ar_kan_residual`).
+- **Which sensor, not just when.** `channel_deviations` and `dominant_channels` attribute each alarm to the channels driving it, and `plot_sensor_graph` draws the graph the detector learned.
+- **Alarms without labels.** `threshold_for_budget` and `dynamic_threshold` pick an operating point from the score distribution alone. Operations can state an alarm rate; they cannot state a recall they have no way to observe.
+- **Honest evaluation.** Event-level scoring alongside the conventional point-adjusted F1, plus a random baseline row in the benchmark. ESA-ADB is scored by its own protocol, with F0.5 as the headline metric.
+- **Runs on flight hardware.** A fitted `KANGDN` distills to a torch-free NumPy evaluator, then to Power of Ten conformant C with golden vectors, for every `feat_mode`. Host and ESP32-S3 targets are in [`targets/`](targets/).
 - **Stacking ensemble** with robust (median + IQR) score normalization, built for anomaly scores that are extreme by definition.
 - **Per-model score decomposition** via `score_components()`, the hook that enables per-channel attribution without retraining.
-- **Spacecraft-native ingestion.** SMAP and CSV load straight into a long-form `TelemetryDataset`, no schema wrangling.
+- **Spacecraft-native ingestion.** SMAP, MSL, ESA-ADB and CSV load straight into a long-form `TelemetryDataset`, no schema wrangling. The ESA-ADB loader reads the official archive directly, with no preprocessing step; its payloads are pickled, so load only that archive, never untrusted files.
 - **Runtime sensitivity control.** `is_anomaly()` accepts a percentile or threshold override, so operators re-tune without retraining.
 
 **On the roadmap**
 
-- `GDN` and `TranAD` deep detectors: inter-sensor relational and transformer-reconstruction faults (Phase 2)
+- `TranAD`: transformer-based sequence reconstruction
 - `SHAPExplainer`: per-channel attribution over `score_components()` (Phase 3)
 - LLM diagnostic reports, SHAP chart supplied as an image (Phase 4)
 - Human-in-the-loop threshold feedback (Phase 5)
-- OPS-SAT cross-dataset generalization (Phase 6)
-- SymTorch symbolic distillation to closed-form fault expressions for edge / microcontroller deployment (stretch)
+
+## AI Transparency
+
+AI assistants are used while writing and maintaining this project's documentation.
+
+What that does not change:
+
+- **Numbers come from runs.** Benchmark figures are produced by [`examples/smap_benchmark.py`](examples/smap_benchmark.py) and [`examples/esa_benchmark.py`](examples/esa_benchmark.py), and the hardware results in the deployment docs are measured on an ESP32-S3 by the conformance harness in [`targets/`](targets/). Nothing quoted in these docs is estimated or inferred.
+- **Behavior claims are backed by tests.** The suite gates CI, and the documented API is covered by it.
+
+Everything published here is reviewed before it lands.
 
 ## Getting Help
 

{telemetry_anomdet-0.1.0 → telemetry_anomdet-0.3.0}/pyproject.toml

@@ -1,10 +1,11 @@
 [project]
 name = "telemetry-anomdet"
-version = "0.1.0"
-description = "Telemetry anomaly detection toolkit with preprocessing, feature extraction, and unsupervised models."
+version = "0.3.0"
+description = "Spacecraft telemetry anomaly detection: classical and graph deep detectors, label-free thresholding, and distillation to C for flight hardware."
 readme = "README.md"
-license = { file = "LICENSE" }
-requires-python = ">=3.10"
+license = "Apache-2.0"
+license-files = ["LICENSE"]
+requires-python = ">=3.11"
 authors = [
   { name = "Julio Anzaldo", email = "anzaldojuliojesus@gmail.com" }
 ]
@@ -12,50 +13,71 @@ dependencies = [
   "pandas>=2.2.2,<3.0.0",
   "numpy>=2.1,<3.0",
   "scikit-learn>=1.3,<2.0",
-  "click>=8.4.2,<9.0",
-  "typing-extensions>=4.16.0",
 ]
 keywords = ["telemetry", "anomaly-detection", "machine-learning", "python"]
+# License is declared as an SPDX expression above, not a classifier: license
+# classifiers are deprecated and setuptools warns on them.
 classifiers = [
+  "Development Status :: 3 - Alpha",
+  "Intended Audience :: Science/Research",
+  "Intended Audience :: Developers",
+  "Natural Language :: English",
+  "Operating System :: OS Independent",
   "Programming Language :: Python :: 3",
-  "Programming Language :: Python :: 3.10",
+  "Programming Language :: Python :: 3 :: Only",
   "Programming Language :: Python :: 3.11",
   "Programming Language :: Python :: 3.12",
   "Programming Language :: Python :: 3.13",
-  "Operating System :: OS Independent",
-  "Topic :: Software Development :: Libraries",
+  "Programming Language :: Python :: 3.14",
+  # The toolkit is pure Python; the artifact it generates for flight targets is C.
+  "Programming Language :: C",
   "Topic :: Scientific/Engineering :: Artificial Intelligence",
-  "Intended Audience :: Science/Research",
-  "Intended Audience :: Developers",
-  "Development Status :: 3 - Alpha",
-  "License :: OSI Approved :: Apache Software License",
-  "Typing :: Typed"
+  "Topic :: Scientific/Engineering :: Information Analysis",
+  "Topic :: Scientific/Engineering :: Astronomy",
+  "Topic :: Software Development :: Libraries",
+  "Topic :: Software Development :: Embedded Systems",
+  "Topic :: Software Development :: Code Generators",
+  "Topic :: System :: Monitoring",
+  "Typing :: Typed",
 ]
 
 [project.optional-dependencies]
 ccsds = [
   "ccsdspy>=2.0.0,<3.0",
 ]
+# Phase 2 deep detector (GDN). Optional so the base install stays
+# torch-free; torch is imported lazily at fit time.
+deep = [
+  "torch>=2.0",
+]
+# Plotting. Optional so the base install stays lean; matplotlib is imported
+# lazily inside each figure function.
+viz = [
+  "matplotlib>=3.7",
+]
+# Experiment tracking for the benchmark scripts. Nothing in the library imports
+# it; examples/smap_benchmark.py logs to it only when TAD_WANDB_PROJECT is set.
+track = [
+  "wandb>=0.17",
+]
 
 [dependency-groups]
 dev = [
   "pytest>=9.1.1,<10.0",
   "pytest-cov>=7.1.0,<8.0",
-  "pre-commit>=4.6.1,<5.0",
+  "pre-commit>=4.6.2,<5.0",
   "black>=26.5.1,<27.0",
-  "ruff>=0.16.0,<1.0",
+  "ruff>=0.16.2,<1.0",
 ]
 docs = [
   "sphinx>=7.0",
-  "sphinx-rtd-theme>=2.0",
+  "pydata-sphinx-theme>=0.15",
   "sphinx-autodoc-typehints>=2.0",
   "numpydoc>=1.7",
+  "sphinx-sitemap>=2.6",
+  "sphinxext-opengraph>=0.9",
+  "sphinx-notfound-page>=1.0",
 ]
-# Phase 2 -- uncomment when GDN/TranAD training starts
-# deep = [
-#   "torch>=2.0",
-#   "torch-geometric",
-# ]
 # Phase 3
 # xai = [
 #   "shap",
@@ -72,12 +94,12 @@ Homepage = "https://github.com/JulioAnzaldo/telemetry-anomdet"
 Repository = "https://github.com/JulioAnzaldo/telemetry-anomdet"
 
 [build-system]
-requires = ["setuptools>=83.0.0", "wheel"]
+requires = ["setuptools>=84.0.0", "wheel"]
 build-backend = "setuptools.build_meta"
 
 [tool.ruff]
 line-length = 100
-target-version = "py310"
+target-version = "py311"
 src = ["src", "tests"]
 
 [tool.ruff.lint]