# SAIS Prism SDK
[](https://pypi.org/project/sais-prism-sdk/)
[](http://unlicense.org/)
## Unified Interface for ML Lifecycle Management
### Features
- 🚀 Centralized Configuration Management
- 🔄 Auto MLflow Integration
- 📦 Extensible Data Access Layer
- 🧩 Declarative Experiment Tracking
- 📚 Hierarchical Dependency Management
### Installation
```bash
pip install sais-prism-sdk
```
### Configuration
Create `sais_foundation.yaml`:
```yaml
ml:
model_repo:
name: "production"
version: "1.0.0"
tags:
framework: "pytorch"
data_access:
enabled: true
token: ${ENV_TOKEN}
```
### User Manual
#### Experiment Tracking
```python
@sais_foundation(auto_log=True)
def train(data):
# Auto-log params/metrics
accuracy = model.fit()
return {"accuracy": accuracy}
```
#### Model Deployment
```bash
sais-prism deploy --model-uri models:/prod/1 --env kubernetes
```
#### Data Access
```python
client = DataClient()
data = client.load("mysql.sales_data")
```
### API Reference
| Class | Description |
|-------|-------------|
| `ConfigManager` | Central config access |
| `MLflowService` | Model registry & tracking |
| `DataClient` | Unified data interface |
### Contributing
1. Install dev deps:
```bash
pip install -r requirements/dev.txt
```
2. Run checks:
```bash
black . && flake8
```
## License
[The Unlicense](https://unlicense.org)
## Architecture Design
### Core Components
- `ConfigManager`: Centralized configuration management
- `MLflowService`: Model registry and tracking
- `DataClient`: Unified data access interface
### Service Locator
```python
class ServiceLocator:
_ml_instance: Optional[MLflowManager] = None
_data_client_instance: Optional[DataAccessClient] = None
@classmethod
def get_ml_manager(cls) -> MLflowManager:
if not cls._ml_instance:
cls._ml_instance = MLflowManager(config.ml)
return cls._ml_instance
```
### MLflow Configuration
```yaml
ml:
enabled: true # Global switch
auto_log: true # Auto-log params/metrics
model_repo:
name: "production_models"
version: "1.0.0" # Semver required
tags: # Custom metadata
framework: "pytorch"
task_type: "classification"
metrics: # Metric collection config
training: ["accuracy", "f1_score"]
validation: ["auc_roc"]
```
### Data Access Configuration
```yaml
data_access:
enabled: true
cached: true # Enable local caching
token: ${ENV_API_TOKEN} # Env var injection
data_access:
mysql:
- "sensor_data"
- "user_behavior"
mongodb:
- "log_records"
```
### Advanced Usage
#### Custom Model Tags
```python
# Extend ModelRepoConfig to add custom tags
class CustomModelConfig(ModelRepoConfig):
deployment_env: str = "staging"
business_unit: str = "recommendation"
# Specify custom class in config
ml:
model_repo:
__class__: "module.path.CustomModelConfig"
deployment_env: "production"
```
#### Extending Data Clients
1. Implement base client interface:
```python
class CustomDataClient(DataAccessClient):
def connect(self, config):
# Init logic
@retry_policy(max_retries=3)
def fetch_data(self, source):
# Data fetch implementation
```
2. Register with service locator:
```python
ServiceLocator.set_data_client(CustomDataClient())
```
### Troubleshooting
#### Config Load Failure
**Symptoms**: `ConfigurationError: sais_foundation.yaml not found`
✅ Solution:
```bash
# Generate default config file
python -m sais_prism.config generate > sais_foundation.yaml
```
#### MLflow Connection Issues
**Symptoms**: `MLFlowIntegrationError: Failed to connect tracking server`
✅ Check steps:
1. Verify MLflow server address
2. Check network connectivity
3. Verify server certificate (if HTTPS enabled)
```python
# Manual connection test
from mlflow.tracking import MlflowClient
client = MlflowClient(config.ml.tracking_uri)
print(client.search_experiments())
Raw data
{
"_id": null,
"home_page": "http://gitlab-paas.internal.sais.com.cn/data_intelligence_platform/sais-prism",
"name": "sais-prism-sdk",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.10",
"maintainer_email": null,
"keywords": null,
"author": "Shepard",
"author_email": "Shepard <zhaoxun@sais.com.cn>",
"download_url": "https://files.pythonhosted.org/packages/88/b9/8ce8e81534cbeb5264c5114fa511868e42e97c36ecf9f8433812ab5a93fd/sais_prism_sdk-0.1.0b0.tar.gz",
"platform": null,
"description": "# SAIS Prism SDK\n\n[](https://pypi.org/project/sais-prism-sdk/)\n[](http://unlicense.org/)\n\n## Unified Interface for ML Lifecycle Management\n\n### Features\n\n- \ud83d\ude80 Centralized Configuration Management\n- \ud83d\udd04 Auto MLflow Integration\n- \ud83d\udce6 Extensible Data Access Layer\n- \ud83e\udde9 Declarative Experiment Tracking\n- \ud83d\udcda Hierarchical Dependency Management\n\n### Installation\n\n```bash\npip install sais-prism-sdk\n```\n\n### Configuration\nCreate `sais_foundation.yaml`:\n\n```yaml\nml:\n model_repo:\n name: \"production\"\n version: \"1.0.0\"\n tags:\n framework: \"pytorch\"\n\ndata_access:\n enabled: true\n token: ${ENV_TOKEN}\n```\n\n### User Manual\n\n#### Experiment Tracking\n```python\n@sais_foundation(auto_log=True)\ndef train(data):\n # Auto-log params/metrics\n accuracy = model.fit()\n return {\"accuracy\": accuracy}\n```\n\n#### Model Deployment\n```bash\nsais-prism deploy --model-uri models:/prod/1 --env kubernetes\n```\n\n#### Data Access\n```python\nclient = DataClient()\ndata = client.load(\"mysql.sales_data\")\n```\n\n### API Reference\n\n| Class | Description |\n|-------|-------------|\n| `ConfigManager` | Central config access |\n| `MLflowService` | Model registry & tracking |\n| `DataClient` | Unified data interface |\n\n### Contributing\n\n1. Install dev deps:\n```bash\npip install -r requirements/dev.txt\n```\n2. Run checks:\n```bash\nblack . && flake8\n```\n\n## License\n[The Unlicense](https://unlicense.org)\n\n## Architecture Design\n\n\n\n### Core Components\n\n- `ConfigManager`: Centralized configuration management\n- `MLflowService`: Model registry and tracking\n- `DataClient`: Unified data access interface\n\n### Service Locator\n\n```python\nclass ServiceLocator:\n _ml_instance: Optional[MLflowManager] = None\n _data_client_instance: Optional[DataAccessClient] = None\n\n @classmethod\n def get_ml_manager(cls) -> MLflowManager:\n if not cls._ml_instance:\n cls._ml_instance = MLflowManager(config.ml)\n return cls._ml_instance\n```\n\n### MLflow Configuration\n\n```yaml\nml:\n enabled: true # Global switch\n auto_log: true # Auto-log params/metrics\n model_repo:\n name: \"production_models\"\n version: \"1.0.0\" # Semver required\n tags: # Custom metadata\n framework: \"pytorch\"\n task_type: \"classification\"\n metrics: # Metric collection config\n training: [\"accuracy\", \"f1_score\"]\n validation: [\"auc_roc\"]\n```\n\n### Data Access Configuration\n\n```yaml\ndata_access:\n enabled: true\n cached: true # Enable local caching\n token: ${ENV_API_TOKEN} # Env var injection\n data_access:\n mysql:\n - \"sensor_data\"\n - \"user_behavior\"\n mongodb:\n - \"log_records\"\n```\n\n### Advanced Usage\n\n#### Custom Model Tags\n\n```python\n# Extend ModelRepoConfig to add custom tags\nclass CustomModelConfig(ModelRepoConfig):\n deployment_env: str = \"staging\"\n business_unit: str = \"recommendation\"\n\n# Specify custom class in config\nml:\n model_repo:\n __class__: \"module.path.CustomModelConfig\"\n deployment_env: \"production\"\n```\n\n#### Extending Data Clients\n\n1. Implement base client interface:\n```python\nclass CustomDataClient(DataAccessClient):\n def connect(self, config):\n # Init logic\n\n @retry_policy(max_retries=3)\n def fetch_data(self, source):\n # Data fetch implementation\n```\n2. Register with service locator:\n```python\nServiceLocator.set_data_client(CustomDataClient())\n```\n\n### Troubleshooting\n\n#### Config Load Failure\n**Symptoms**: `ConfigurationError: sais_foundation.yaml not found`\n\n\u2705 Solution:\n```bash\n# Generate default config file\npython -m sais_prism.config generate > sais_foundation.yaml\n```\n\n#### MLflow Connection Issues\n**Symptoms**: `MLFlowIntegrationError: Failed to connect tracking server`\n\n\u2705 Check steps:\n1. Verify MLflow server address\n2. Check network connectivity\n3. Verify server certificate (if HTTPS enabled)\n\n```python\n# Manual connection test\nfrom mlflow.tracking import MlflowClient\nclient = MlflowClient(config.ml.tracking_uri)\nprint(client.search_experiments())\n",
"bugtrack_url": null,
"license": "The Unlicense",
"summary": "SAIS Prism: A unified interface for ML data access and lifecycle management",
"version": "0.1.0b0",
"project_urls": {
"Homepage": "http://gitlab-paas.internal.sais.com.cn/data_intelligence_platform/sais-prism"
},
"split_keywords": [],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "e4e6c4eb492b1708d553bf51cffc53014ffb551d23e628b9a3bee14b236d540e",
"md5": "40fb20ff84005bade8726723a7f31fcf",
"sha256": "88805fefb3ee6b729342ab4cee3073aa8612ab149733bfc45a60f1e007f0dcdb"
},
"downloads": -1,
"filename": "sais_prism_sdk-0.1.0b0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "40fb20ff84005bade8726723a7f31fcf",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.10",
"size": 21445,
"upload_time": "2025-02-22T16:37:20",
"upload_time_iso_8601": "2025-02-22T16:37:20.312921Z",
"url": "https://files.pythonhosted.org/packages/e4/e6/c4eb492b1708d553bf51cffc53014ffb551d23e628b9a3bee14b236d540e/sais_prism_sdk-0.1.0b0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "88b98ce8e81534cbeb5264c5114fa511868e42e97c36ecf9f8433812ab5a93fd",
"md5": "84d868adc50d132d87e6b6bbf20b7346",
"sha256": "60b6e163e3ad4410f75e1787881b90fd338944cdd164d703d6f449f67b728701"
},
"downloads": -1,
"filename": "sais_prism_sdk-0.1.0b0.tar.gz",
"has_sig": false,
"md5_digest": "84d868adc50d132d87e6b6bbf20b7346",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.10",
"size": 336128,
"upload_time": "2025-02-22T16:37:22",
"upload_time_iso_8601": "2025-02-22T16:37:22.915553Z",
"url": "https://files.pythonhosted.org/packages/88/b9/8ce8e81534cbeb5264c5114fa511868e42e97c36ecf9f8433812ab5a93fd/sais_prism_sdk-0.1.0b0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-02-22 16:37:22",
"github": false,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"lcname": "sais-prism-sdk"
}
JSON