livae-sc 0.2.3__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
livae_sc-0.2.3/LICENSE ADDED
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2025 Zeyu Fu
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
@@ -0,0 +1,30 @@
1
+ # Include essential files
2
+ include README.md
3
+ include LICENSE
4
+ include pyproject.toml
5
+ include requirements.txt
6
+
7
+ # Include package data
8
+ recursive-include livae *.py
9
+
10
+ # Include tests
11
+ recursive-include tests *.py
12
+
13
+ # Exclude everything else
14
+ global-exclude __pycache__
15
+ global-exclude *.py[cod]
16
+ global-exclude .DS_Store
17
+ global-exclude *.so
18
+ global-exclude .git*
19
+
20
+ # Explicitly exclude frontend, api, and other project files
21
+ prune frontend
22
+ prune api
23
+ prune scripts
24
+ prune .github
25
+ prune dist
26
+ prune build
27
+ prune downloads
28
+ exclude start_services.py
29
+ exclude uvicorn_pid.txt
30
+ exclude review-points.md
@@ -0,0 +1,288 @@
1
+ Metadata-Version: 2.4
2
+ Name: livae-sc
3
+ Version: 0.2.3
4
+ Summary: Lorentzian Interpretable Variational Autoencoder for single-cell data
5
+ Author: LiVAE Contributors
6
+ License: MIT License
7
+
8
+ Copyright (c) 2025 Zeyu Fu
9
+
10
+ Permission is hereby granted, free of charge, to any person obtaining a copy
11
+ of this software and associated documentation files (the "Software"), to deal
12
+ in the Software without restriction, including without limitation the rights
13
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
14
+ copies of the Software, and to permit persons to whom the Software is
15
+ furnished to do so, subject to the following conditions:
16
+
17
+ The above copyright notice and this permission notice shall be included in all
18
+ copies or substantial portions of the Software.
19
+
20
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
21
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
22
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
23
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
24
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
25
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
26
+ SOFTWARE.
27
+
28
+ Project-URL: Homepage, https://github.com/PeterPonyu/LiVAE
29
+ Project-URL: Source, https://github.com/PeterPonyu/LiVAE
30
+ Project-URL: Issues, https://github.com/PeterPonyu/LiVAE/issues
31
+ Keywords: single-cell,variational-autoencoder,hyperbolic,lorentz,bioinformatics
32
+ Classifier: Programming Language :: Python :: 3
33
+ Classifier: Programming Language :: Python :: 3.10
34
+ Classifier: License :: OSI Approved :: MIT License
35
+ Classifier: Intended Audience :: Science/Research
36
+ Classifier: Topic :: Scientific/Engineering :: Bio-Informatics
37
+ Requires-Python: >=3.10
38
+ Description-Content-Type: text/markdown
39
+ License-File: LICENSE
40
+ Requires-Dist: torch<2.5.0,>=2.3.0
41
+ Requires-Dist: torchvision<0.20.0,>=0.18.0
42
+ Requires-Dist: scanpy<1.11.0,>=1.10.0
43
+ Requires-Dist: scvi-tools<1.2.0,>=1.1.0
44
+ Requires-Dist: anndata<0.11.0,>=0.10.0
45
+ Requires-Dist: scib>=1.0.0
46
+ Requires-Dist: numpy<1.27.0,>=1.26.0
47
+ Requires-Dist: pandas<2.3.0,>=2.2.0
48
+ Requires-Dist: scipy<1.13.0,>=1.12.0
49
+ Requires-Dist: scikit-learn<1.6.0,>=1.5.0
50
+ Requires-Dist: tqdm<5.0.0,>=4.66.0
51
+ Requires-Dist: fastapi<0.118.0,>=0.117.0
52
+ Requires-Dist: uvicorn[standard]<0.37.0,>=0.36.0
53
+ Requires-Dist: python-multipart>=0.0.6
54
+ Dynamic: license-file
55
+
56
+
57
+ # LiVAE: Lorentz-regularized Interpretable Variational Autoencoder
58
+
59
+ [![PyPI version](https://img.shields.io/pypi/v/livae-sc.svg)](https://pypi.org/project/livae-sc/)
60
+ [![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
61
+ [![PyTorch](https://img.shields.io/badge/PyTorch-2.3+-red.svg)](https://pytorch.org/)
62
+ [![CI](https://github.com/PeterPonyu/LiVAE/actions/workflows/ci.yml/badge.svg)](https://github.com/PeterPonyu/LiVAE/actions/workflows/ci.yml)
63
+ [![License](https://img.shields.io/badge/license-MIT-green.svg)](LICENSE)
64
+
65
+ LiVAE (Lorentz-regularized Interpretable Variational Autoencoder) is a dual-pathway architecture for learning interpretable latent representations from single-cell transcriptomic and epigenomic data. It applies hyperbolic geometry as soft regularization over standard Euclidean latent spaces to balance local fidelity with global coherence.
66
+
67
+
68
+ ## Installation
69
+
70
+ ```bash
71
+ # Install from PyPI (recommended)
72
+ pip install livae-sc
73
+
74
+ # Or install from source for development
75
+ git clone https://github.com/PeterPonyu/LiVAE.git
76
+ cd LiVAE
77
+ pip install -e .
78
+ ```
79
+
80
+ From version 0.2.3 the package is published on PyPI as `livae-sc`; the import name is still `livae`. The earlier `livae` project on PyPI stays at 0.2.1 and is no longer updated.
81
+
82
+ ### Core Requirements
83
+
84
+ ```text
85
+ torch>=2.3.0,<2.5.0
86
+ torchvision>=0.18.0,<0.20.0
87
+ scanpy>=1.10.0,<1.11.0
88
+ scvi-tools>=1.1.0,<1.2.0
89
+ anndata>=0.10.0,<0.11.0
90
+ scib>=1.0.0
91
+ numpy>=1.26.0,<1.27.0
92
+ pandas>=2.2.0,<2.3.0
93
+ scipy>=1.12.0,<1.13.0
94
+ scikit-learn>=1.5.0,<1.6.0
95
+ tqdm>=4.66.0,<5.0.0
96
+ fastapi>=0.117.0,<0.118.0
97
+ uvicorn[standard]>=0.36.0,<0.37.0
98
+ python-multipart>=0.0.6
99
+ ```
100
+
101
+ ## Python Quick Start
102
+
103
+ ```python
104
+ import scanpy as sc
105
+ from livae import agent
106
+
107
+ # Load your single-cell data
108
+ adata = sc.read_h5ad('your_data.h5ad')
109
+
110
+ # Initialize LiVAE agent
111
+ model = agent(
112
+ adata=adata,
113
+ layer='counts', # Data layer to use
114
+ latent_dim=10, # Primary latent dimension
115
+ i_dim=2, # Interpretable latent dimension
116
+ hidden_dim=128, # Hidden layer dimension
117
+ lr=1e-4, # Learning rate
118
+ # Regularization weights
119
+ beta=1.0, # β-VAE weight
120
+ lorentz=1.0, # Lorentzian regularization
121
+ irecon=1.0, # Interpretable reconstruction
122
+ )
123
+
124
+ # Train the model
125
+ model.fit(epochs=100)
126
+
127
+ # Extract embeddings
128
+ latent = model.get_latent() # Primary latent representation
129
+ interpretable = model.get_iembed() # Interpretable embedding (i_dim)
130
+
131
+ # Access results
132
+ print(f"Latent shape: {latent.shape}")
133
+ print(f"Interpretable embedding shape: {interpretable.shape}")
134
+ ```
135
+
136
+ ## Web Interface
137
+
138
+ LiVAE includes an integrated web interface for interactive training and visualization.
139
+
140
+ ### Launch the Application
141
+
142
+ ```bash
143
+ # Start the FastAPI server (serves both API and frontend)
144
+ uvicorn api.main:app --host 0.0.0.0 --port 8000
145
+ ```
146
+
147
+ Access the web interface at `http://localhost:8000`
148
+
149
+ ### Features
150
+
151
+ - **Data Upload**: Upload single-cell datasets (H5AD format)
152
+ - **Training Configuration**: Configure model parameters and hyperparameters
153
+ - **Real-time Monitoring**: Track loss curves and metrics during training
154
+ - **Results Visualization**: Download embeddings and view training summaries
155
+
156
+ ### API Endpoints
157
+
158
+ - `POST /upload` - Upload dataset
159
+ - `POST /train/start` - Start training
160
+ - `GET /train/metrics` - Get training metrics
161
+ - `GET /embeddings/interpretable` - Get interpretable embeddings
162
+ - `GET /embeddings/latent` - Get latent embeddings
163
+ - `GET /download/embeddings/{type}` - Download embeddings as CSV
164
+
165
+
166
+ ## Architecture Overview
167
+
168
+ LiVAE consists of three main components working in concert:
169
+
170
+ ### 1. **Encoder Network**
171
+
172
+ ```text
173
+ Input (gene expression) → Hidden → Hidden → μ, σ (latent parameters)
174
+ ```
175
+
176
+ - Maps high-dimensional gene expression to latent distribution parameters
177
+ - Uses reparameterization trick for differentiable sampling
178
+
179
+ ### 2. **Hyperbolic Transformation**
180
+
181
+ ```text
182
+ Latent Sample → Tangent Space → Exponential Map → Lorentzian Manifold
183
+ ```
184
+
185
+ - Projects latent vectors onto hyperbolic manifold
186
+ - Enables natural representation of hierarchical relationships
187
+
188
+ ### 3. **Dual Decoder Pathway**
189
+
190
+ ```text
191
+ Primary: Latent → Decoder → Reconstruction
192
+ Interpretable: Latent → Compress → Decompress → Decoder → Reconstruction
193
+ ```
194
+
195
+ - Dual reconstruction paths for enhanced representation learning
196
+ - Compression bottleneck encourages essential feature extraction
197
+
198
+ ## Loss Function
199
+
200
+ LiVAE optimizes a composite loss function:
201
+
202
+ ```text
203
+ L_total = L_recon + L_irecon + L_lorentz + L_KL
204
+ ```
205
+
206
+ Where:
207
+
208
+ - **L_recon**: Negative binomial reconstruction loss
209
+ - **L_irecon**: Interpretable reconstruction loss
210
+ - **L_lorentz**: Lorentz distance regularization
211
+ - **L_KL**: KL divergence
212
+
213
+
214
+ ## Evaluation Metrics
215
+
216
+ ### Clustering Quality
217
+
218
+ - **ARI** (Adjusted Rand Index): Clustering agreement with ground truth
219
+ - **NMI** (Normalized Mutual Information): Information-theoretic clustering measure
220
+ - **ASW** (Average Silhouette Width): Cluster cohesion and separation
221
+
222
+ ### Cluster Validity
223
+
224
+ - **C_H** (Calinski-Harabasz): Ratio of between to within cluster variance
225
+ - **D_B** (Davies-Bouldin): Average similarity between clusters
226
+ - **P_C** (Graph Connectivity): Connectivity within clusters
227
+
228
+ ### Batch Integration
229
+
230
+ - **cLISI**: Cluster-specific Local Inverse Simpson Index
231
+ - **iLISI**: Integration LISI for batch mixing
232
+ - **bASW**: Batch-corrected Average Silhouette Width
233
+
234
+ ## Performance Tips
235
+
236
+ ### Memory Optimization
237
+
238
+ ```python
239
+ # For large datasets, reduce batch size
240
+ model = agent(adata, percent=0.005) # Use 0.5% of data per batch (default is 1%)
241
+
242
+ # Use CPU if GPU memory is limited
243
+ import torch
244
+ model = agent(adata, device=torch.device('cpu'))
245
+ ```
246
+
247
+ ### Training Efficiency
248
+
249
+ ```python
250
+ # Progressive training with increasing regularization
251
+ model = agent(adata, beta=0.1, lorentz=0.0)
252
+ model.fit(epochs=500)
253
+
254
+ # Increase regularization for fine-tuning
255
+ model.beta = 1.0
256
+ model.lorentz = 0.1
257
+ model.fit(epochs=500)
258
+ ```
259
+
260
+ ## Citation
261
+
262
+ If you use LiVAE in your research, please cite the following paper:
263
+
264
+ ```bibtex
265
+ @ARTICLE{10.3389/fgene.2025.1713727,
266
+ AUTHOR={Fu, Zeyu and Fu, Jiawei and Chen, Chunlin and Zhang, Keyang and Wang, Song},
267
+ TITLE={Lorentz-regularized interpretable VAE for multi-scale single-cell transcriptomic and epigenomic embeddings},
268
+ JOURNAL={Frontiers in Genetics},
269
+ VOLUME={16},
270
+ PAGES={1713727},
271
+ YEAR={2026},
272
+ URL={https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2025.1713727},
273
+ DOI={10.3389/fgene.2025.1713727},
274
+ ISSN={1664-8021}
275
+ }
276
+ ```
277
+
278
+ ## Contributing
279
+
280
+ Contributions are welcome! Please ensure tests pass before submitting PRs:
281
+
282
+ ```bash
283
+ pytest -q
284
+ ```
285
+
286
+ ## License
287
+
288
+ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
@@ -0,0 +1,233 @@
1
+
2
+ # LiVAE: Lorentz-regularized Interpretable Variational Autoencoder
3
+
4
+ [![PyPI version](https://img.shields.io/pypi/v/livae-sc.svg)](https://pypi.org/project/livae-sc/)
5
+ [![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
6
+ [![PyTorch](https://img.shields.io/badge/PyTorch-2.3+-red.svg)](https://pytorch.org/)
7
+ [![CI](https://github.com/PeterPonyu/LiVAE/actions/workflows/ci.yml/badge.svg)](https://github.com/PeterPonyu/LiVAE/actions/workflows/ci.yml)
8
+ [![License](https://img.shields.io/badge/license-MIT-green.svg)](LICENSE)
9
+
10
+ LiVAE (Lorentz-regularized Interpretable Variational Autoencoder) is a dual-pathway architecture for learning interpretable latent representations from single-cell transcriptomic and epigenomic data. It applies hyperbolic geometry as soft regularization over standard Euclidean latent spaces to balance local fidelity with global coherence.
11
+
12
+
13
+ ## Installation
14
+
15
+ ```bash
16
+ # Install from PyPI (recommended)
17
+ pip install livae-sc
18
+
19
+ # Or install from source for development
20
+ git clone https://github.com/PeterPonyu/LiVAE.git
21
+ cd LiVAE
22
+ pip install -e .
23
+ ```
24
+
25
+ From version 0.2.3 the package is published on PyPI as `livae-sc`; the import name is still `livae`. The earlier `livae` project on PyPI stays at 0.2.1 and is no longer updated.
26
+
27
+ ### Core Requirements
28
+
29
+ ```text
30
+ torch>=2.3.0,<2.5.0
31
+ torchvision>=0.18.0,<0.20.0
32
+ scanpy>=1.10.0,<1.11.0
33
+ scvi-tools>=1.1.0,<1.2.0
34
+ anndata>=0.10.0,<0.11.0
35
+ scib>=1.0.0
36
+ numpy>=1.26.0,<1.27.0
37
+ pandas>=2.2.0,<2.3.0
38
+ scipy>=1.12.0,<1.13.0
39
+ scikit-learn>=1.5.0,<1.6.0
40
+ tqdm>=4.66.0,<5.0.0
41
+ fastapi>=0.117.0,<0.118.0
42
+ uvicorn[standard]>=0.36.0,<0.37.0
43
+ python-multipart>=0.0.6
44
+ ```
45
+
46
+ ## Python Quick Start
47
+
48
+ ```python
49
+ import scanpy as sc
50
+ from livae import agent
51
+
52
+ # Load your single-cell data
53
+ adata = sc.read_h5ad('your_data.h5ad')
54
+
55
+ # Initialize LiVAE agent
56
+ model = agent(
57
+ adata=adata,
58
+ layer='counts', # Data layer to use
59
+ latent_dim=10, # Primary latent dimension
60
+ i_dim=2, # Interpretable latent dimension
61
+ hidden_dim=128, # Hidden layer dimension
62
+ lr=1e-4, # Learning rate
63
+ # Regularization weights
64
+ beta=1.0, # β-VAE weight
65
+ lorentz=1.0, # Lorentzian regularization
66
+ irecon=1.0, # Interpretable reconstruction
67
+ )
68
+
69
+ # Train the model
70
+ model.fit(epochs=100)
71
+
72
+ # Extract embeddings
73
+ latent = model.get_latent() # Primary latent representation
74
+ interpretable = model.get_iembed() # Interpretable embedding (i_dim)
75
+
76
+ # Access results
77
+ print(f"Latent shape: {latent.shape}")
78
+ print(f"Interpretable embedding shape: {interpretable.shape}")
79
+ ```
80
+
81
+ ## Web Interface
82
+
83
+ LiVAE includes an integrated web interface for interactive training and visualization.
84
+
85
+ ### Launch the Application
86
+
87
+ ```bash
88
+ # Start the FastAPI server (serves both API and frontend)
89
+ uvicorn api.main:app --host 0.0.0.0 --port 8000
90
+ ```
91
+
92
+ Access the web interface at `http://localhost:8000`
93
+
94
+ ### Features
95
+
96
+ - **Data Upload**: Upload single-cell datasets (H5AD format)
97
+ - **Training Configuration**: Configure model parameters and hyperparameters
98
+ - **Real-time Monitoring**: Track loss curves and metrics during training
99
+ - **Results Visualization**: Download embeddings and view training summaries
100
+
101
+ ### API Endpoints
102
+
103
+ - `POST /upload` - Upload dataset
104
+ - `POST /train/start` - Start training
105
+ - `GET /train/metrics` - Get training metrics
106
+ - `GET /embeddings/interpretable` - Get interpretable embeddings
107
+ - `GET /embeddings/latent` - Get latent embeddings
108
+ - `GET /download/embeddings/{type}` - Download embeddings as CSV
109
+
110
+
111
+ ## Architecture Overview
112
+
113
+ LiVAE consists of three main components working in concert:
114
+
115
+ ### 1. **Encoder Network**
116
+
117
+ ```text
118
+ Input (gene expression) → Hidden → Hidden → μ, σ (latent parameters)
119
+ ```
120
+
121
+ - Maps high-dimensional gene expression to latent distribution parameters
122
+ - Uses reparameterization trick for differentiable sampling
123
+
124
+ ### 2. **Hyperbolic Transformation**
125
+
126
+ ```text
127
+ Latent Sample → Tangent Space → Exponential Map → Lorentzian Manifold
128
+ ```
129
+
130
+ - Projects latent vectors onto hyperbolic manifold
131
+ - Enables natural representation of hierarchical relationships
132
+
133
+ ### 3. **Dual Decoder Pathway**
134
+
135
+ ```text
136
+ Primary: Latent → Decoder → Reconstruction
137
+ Interpretable: Latent → Compress → Decompress → Decoder → Reconstruction
138
+ ```
139
+
140
+ - Dual reconstruction paths for enhanced representation learning
141
+ - Compression bottleneck encourages essential feature extraction
142
+
143
+ ## Loss Function
144
+
145
+ LiVAE optimizes a composite loss function:
146
+
147
+ ```text
148
+ L_total = L_recon + L_irecon + L_lorentz + L_KL
149
+ ```
150
+
151
+ Where:
152
+
153
+ - **L_recon**: Negative binomial reconstruction loss
154
+ - **L_irecon**: Interpretable reconstruction loss
155
+ - **L_lorentz**: Lorentz distance regularization
156
+ - **L_KL**: KL divergence
157
+
158
+
159
+ ## Evaluation Metrics
160
+
161
+ ### Clustering Quality
162
+
163
+ - **ARI** (Adjusted Rand Index): Clustering agreement with ground truth
164
+ - **NMI** (Normalized Mutual Information): Information-theoretic clustering measure
165
+ - **ASW** (Average Silhouette Width): Cluster cohesion and separation
166
+
167
+ ### Cluster Validity
168
+
169
+ - **C_H** (Calinski-Harabasz): Ratio of between to within cluster variance
170
+ - **D_B** (Davies-Bouldin): Average similarity between clusters
171
+ - **P_C** (Graph Connectivity): Connectivity within clusters
172
+
173
+ ### Batch Integration
174
+
175
+ - **cLISI**: Cluster-specific Local Inverse Simpson Index
176
+ - **iLISI**: Integration LISI for batch mixing
177
+ - **bASW**: Batch-corrected Average Silhouette Width
178
+
179
+ ## Performance Tips
180
+
181
+ ### Memory Optimization
182
+
183
+ ```python
184
+ # For large datasets, reduce batch size
185
+ model = agent(adata, percent=0.005) # Use 0.5% of data per batch (default is 1%)
186
+
187
+ # Use CPU if GPU memory is limited
188
+ import torch
189
+ model = agent(adata, device=torch.device('cpu'))
190
+ ```
191
+
192
+ ### Training Efficiency
193
+
194
+ ```python
195
+ # Progressive training with increasing regularization
196
+ model = agent(adata, beta=0.1, lorentz=0.0)
197
+ model.fit(epochs=500)
198
+
199
+ # Increase regularization for fine-tuning
200
+ model.beta = 1.0
201
+ model.lorentz = 0.1
202
+ model.fit(epochs=500)
203
+ ```
204
+
205
+ ## Citation
206
+
207
+ If you use LiVAE in your research, please cite the following paper:
208
+
209
+ ```bibtex
210
+ @ARTICLE{10.3389/fgene.2025.1713727,
211
+ AUTHOR={Fu, Zeyu and Fu, Jiawei and Chen, Chunlin and Zhang, Keyang and Wang, Song},
212
+ TITLE={Lorentz-regularized interpretable VAE for multi-scale single-cell transcriptomic and epigenomic embeddings},
213
+ JOURNAL={Frontiers in Genetics},
214
+ VOLUME={16},
215
+ PAGES={1713727},
216
+ YEAR={2026},
217
+ URL={https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2025.1713727},
218
+ DOI={10.3389/fgene.2025.1713727},
219
+ ISSN={1664-8021}
220
+ }
221
+ ```
222
+
223
+ ## Contributing
224
+
225
+ Contributions are welcome! Please ensure tests pass before submitting PRs:
226
+
227
+ ```bash
228
+ pytest -q
229
+ ```
230
+
231
+ ## License
232
+
233
+ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
@@ -0,0 +1,5 @@
1
+ from .agent import agent
2
+
3
+ __all__ = ['agent']
4
+
5
+ __version__ = '0.2.3'