Overview
This topic describes the model service in seekdb. The model service allows you to register and manage AI models and endpoints. It serves as the foundation for using AI functions.
What is the model service
The model service in seekdb allows you to register, modify, and delete AI models and endpoints. It is implemented through the DBMS_AI_SERVICE system package.
The following table describes the core concepts of the model service:
| Concept | Description |
|---|---|
| AI Model | An AI model is an abstract object in seekdb that identifies and references a specific AI model within the database. After creating a model object using CREATE_AI_MODEL, you can reference the model in AI functions using model_key. |
| AI Model Endpoint | An AI model endpoint is a specific configuration that connects to an external AI model service. It contains information such as the URL, API Key, model name, and provider. The endpoint defines how to communicate with the external model service. |
| Model Type | Model types classify the functionalities of AI models. seekdb supports the following types:
|
| Provider | A provider is a service that offers AI model services, such as OpenAI or DeepSeek. Different providers may have different API formats and parameters. seekdb adapts to different providers' interface specifications through the provider parameter. |
Relationship with the AI function service
Hybrid search relies on AI model management and the AI_EMBED function. When deleting an AI model, check if it is referenced by hybrid search to avoid potential issues.
The AI model management service is the foundation of the AI function service. In seekdb, you can only call various AI functions (such as AI_COMPLETE, AI_EMBED, and AI_RERANK) at the SQL layer after successfully registering and managing AI models and their endpoints. This enables built-in intelligent processing capabilities like text generation, vector embedding, and retrieval reranking. For more information about AI functions, see AI Functions Overview.