The AI service provides a unified access and invocation capability for large models on OceanBase Cloud. With the AI service, you can configure models, manage API keys, invoke models, and monitor usage from a single console. It also allows you to quickly integrate large model capabilities into your application systems or database processes by supporting interfaces compatible with OpenAI, Anthropic, and Vertex AI.
Note
Currently, this feature is available only to allowlisted users. If you want to use it, contact OceanBase Technical Support.
What is the AI service?
The AI service is designed for developers, data engineers, and application teams who need to integrate large model capabilities. It provides a unified access layer and invocation entry for models. The AI service helps you overcome the differences in interfaces among different model providers by integrating model configuration, invocation credentials, database-side invocation, and invocation statistics into a single workflow.
With the AI service, you can invoke models through HTTP APIs in your applications or directly use model capabilities in OceanBase Database through SQL and AI functions. This allows you to more efficiently build capabilities such as question-answering, generation, vector retrieval, re-ranking, and intelligent analysis.
The AI service provides capabilities in the following areas:
Model configuration: Configure default models for different AI tasks for easier subsequent invocation.
API key management: Create and manage API keys for applications, security services, or integration tasks to access the AI service.
Multiple invocation methods: Supports invoking models through APIs and using model capabilities in database scenarios through SQL and AI functions.
Database registration of models: Generate SQL statements required for registering and enabling AI models in the database for easier model registration and activation.
Compatibility with multiple interface protocols: Provides APIs compatible with OpenAI, Anthropic, and Vertex AI for convenient reuse of existing application access methods and SDK usage habits.
Usage observability: Provides statistics on token usage, model dimensions, and API key dimensions to help you analyze invocation situations and resource consumption.
Use cases
The AI service is suitable for the following scenarios:
Application access to large model services: Applications can quickly implement capabilities such as text generation, summarization, question-answering, and classification by invoking models through API keys.
AI processing within the database: By combining OceanBase AI functions with SQL, you can directly invoke models in the database to implement capabilities such as data analysis, extraction, summarization, vectorization, and re-ranking.
Vector retrieval and hybrid search: By combining vector capabilities with large model capabilities, you can build applications such as semantic retrieval, RAG, and intelligent recommendations.
Unified management of model invocations: Reduces redundant integration work among different model providers by using unified model configurations and compatible interfaces.
Usage analysis and cost management: Continuously optimize model selection, invocation frequency, and resource investment by analyzing model and API key statistics.
Advantages
Compared with directly integrating with multiple model providers, the AI service offers the following advantages:
Unified entry point, shortening the integration cycle: Concentrates model configuration, invocation credentials, invocation methods, and statistical analysis in a single entry point, eliminating the need to adapt to multiple model platforms and enabling faster integration from testing to production.
Good protocol compatibility, reducing development costs: Compatible with APIs of OpenAI, Anthropic, and Vertex AI, facilitating quick migration and reuse of existing code and reducing redundant development work in interface adaptation, authentication, and invocation processes.
Dual access paths for applications and databases, enhancing business implementation efficiency: Suitable for both application-level integration and direct invocation within the database, allowing application development, database processing, and vector retrieval to quickly combine around the same set of service capabilities.
More efficient model selection and easier expansion: Supports setting default models and viewing model lists and invocation examples from different providers, reducing selection and trial-and-error costs; when you need to integrate more models, scenarios, or invocation parties in the future, you can smoothly expand by reusing existing management and invocation methods.
Clearer operations and governance, enhancing resource controllability: Easier identification of invocation sources, resource consumption, and abnormal fluctuations through API key and model dimension statistics, facilitating permission isolation, cost analysis, and resource optimization.