About Building Agents in OCI Generative AI

• File Search lets models retrieve relevant information from a Vector Store and generate accurate, relevant responses. By creating vector stores and adding files to them, you enable both semantic and keyword-based search across your data.
• SQL Search (NL2SQL) enables the OCI Responses API to interact with federated enterprise data without moving or copying customer data. Natural language is translated into SQL queries, and a semantic enrichment layer maps business terms to underlying fields and joins.
• Code Interpreter lets models write and run Python code in a sandboxed environment for complex tasks such as data analysis, coding, and mathematics. It includes more than 420 preinstalled Python libraries and runs with no external network access.
• Local Function Calling lets the OCI Responses API invoke functions defined in your application. The API returns control to your application with a JSON payload describing which function to call and which parameters to use, and your application resumes execution after returning the function output.
• Remote MCP Calling lets the OCI Responses API invoke tools hosted on a remote MCP server directly from the platform, removing the extra client-side round trip required by standard function calling.

• Short-term memory uses the Responses API and Conversations API to simplify conversation state management across multiple turns in a conversation.
• Long-term memory provides durable memory across conversations through a unique subject_id within a project.
• Short-term memory optimization condenses chat history into retained memory, helping reduce latency and token usage in long conversations.

• Vector Stores API converts documents into searchable embeddings and stores them in a managed index so agents can retrieve the most relevant passages for a user's question.
• Files API lets you securely upload and manage documents and other data that an agent can use during task execution.
• Connector API lets you use data stored in Object Storage buckets with Vector Stores. The Vector Store Connector can continuously synchronize data from Object Storage into a Vector Store at scale.
• Containers API provides secure, stateful sandboxes for isolated code execution and data processing.
• Container Files API lets agents persist and manage data directly within container execution environments across multiple steps of a task.

A project is an OCI resource that organizes agent-specific artifacts created through the Generative AI service, including responses, conversations, files, and containers. Projects are isolated from each other, so artifacts in one project aren’t accessible from another project. A project is required to call the OCI Responses API for agent-related tasks.

A project setting can include:

• Data retention for responses and conversations
• Long-term memory settings
• Conversation history compaction (short-term memory compaction) settings