API v2.0: Simpler, Faster AI Development with Machbase
The previous API focused on connecting Machbase Neo with Python. API v2.0 expands that focus to the functions needed for AI development in Python, making code more concise and development faster. From data analysis to model building, it provides tools to improve productivity and streamline AI workflows.
Here are the main features.
API v2.0 organizes its functions into five categories aligned with the AI development process. Each category is designed to make development easier and faster.
1) EDA (Exploratory Data Analysis)
Explore the characteristics of your data through general information, visualizations, and statistical analysis. A better understanding of the data can reduce analysis time and improve prediction accuracy.
2) Model
Load and use a range of AI models through a straightforward interface. Flexible options simplify model implementation and help you work more efficiently.
3) Preprocessing
Prepare data for modeling. Use the characteristics identified during EDA to guide preprocessing tasks.
4) Tql
This category retains the integration capabilities from API v1.0, connecting Machbase Neo with Python for efficient data exchange.
5) Util
Utility functions support tasks such as loading data and visualizing results. They automate or simplify supporting work that developers would otherwise implement themselves.
Changes Required When Updating
Existing functionality remains available, but API calls replace functions used in the previous version. You will need to update Python code you have already written or are currently developing.
The endpoint structure has also changed:
api/v1 → api/v2/{category}
How to Use the New API
The following steps walk through Python code for training an AI model.
1. Import Libraries
In addition to the existing libraries, add the new API imports shown in the screenshot. If you move folders or files, update the corresponding path added to sys as well.
2. Set Basic Parameters
Define the basic information about the data, based on the findings from exploratory data analysis.
3. Configure the Model
Set the initial parameters for the SparseTSF model loaded through the new API.
Details of the input parameters for each model are available at https://www.yotahub.com/api/ .
4. Configure the Dataset and Data Loader
Structure the data so the model can use past observations to predict future values.
5. Train and Test the Model
Machbase’s training approach loads only the required amount of data from Neo in sequence, then immediately preprocesses it and trains the model. This reduces the need to hold the entire dataset in memory and supports training and testing on large datasets with limited local resources.
In the screenshot below, the red box marks data loading, the blue box preprocessing, and the yellow box model training.
6. Evaluate Model Performance
Evaluate the trained model using the test results.
The complete notebook is available at https://github.com/machbase/datahub/blob/main/dataset/2024/17.Bitcoin/code/17.Bitcoin_Buffered_Fetch.ipynb .
Where to Find More Examples and API Documentation
API v2.0 is used starting with the “17. Bitcoin Data” example.
Machbase installation and release information: https://docs.machbase.com/neo/releases/
Machbase Datahub: https://www.yotahub.com/
Detailed API v2.0 documentation: https://www.yotahub.com/api/
Machbase Datahub on GitHub: https://github.com/machbase/datahub
For additional features or support, contact joy.cho@machbase.com .
Build AI models more easily and quickly with Machbase.
API v2.0 is designed to improve your development experience, making AI modeling, data analysis, and real-time processing easier and faster. Try it and share your feedback.
