WCC Agent Framework

Overview

The WebCenter Content Agent Framework enables users to create, manage, and communicate with intelligent agents designed to operate within Oracle WebCenter Content. These agents support natural-language interactions and leverage Retrieval-Augmented Generation (RAG) to deliver context-aware responses by combining user input with relevant enterprise content using semantic search. Each agent can be configured with a controlled set of tools, limited to IDC (Intradoc) services —such as content metadata updates, and user or workflow operations—ensuring secure, auditable, and compliant automation.

Prerequisites

  1. In order to use or create AI agents with RAG capabilities, SmartContent component must be enabled and Semantic Search must be configured.

Enabling the component

  1. Log in to WebCenter Content as an administrator.
  2. Navigate to Administration → Admin Server → Component Manager → Advanced Component Manager.
  3. Select the WCCAgentFramework component and click Enable.
  4. Restart the server once the component is enabled.
    Enable Component

Configuring the AI Agents

  1. Log in to WebCenter Content as an administrator.
  2. Navigate to Administration, AI Agents, LLM Configuration for Agents.
  3. From the dropdown, select the LLMProvider and ModelName.
  4. Provide the API key for the LLM, stored under the map oracle.wsm.security in the WebLogic Credential Store.
  5. Provide the proxy details to connect to the LLM, if required.
Name Description
LLMProvider LLM provider name. Currently, only OpenAI is supported.
ModelName LLM model name. Currently, GPT-4o-mini and GPT-4o are supported.
AgentAPIKey Credential key which contains the API key of the LLM saved under the map oracle.wsm.security.
ProxyHost The hostname or IP address of the HTTP proxy server used to route requests.
ProxyPort The port number of the HTTP proxy server. Defaults to 80 if not provided.

LLM Configuration Landing Page

  1. When you click Update, the system validates the configuration. If successful, a banner message will appear: LLM Configuration is valid! LLM Configuration Success Page

Agent Studio

  1. Once the configuration of the LLM is complete, Navigate to Administration, AI Agents, Agent Studio.
  2. This shows default system agents that are shipped with the component.
  3. Following are the system agents.
    1. WccDocumentationAgent: An AI agent that answers questions about WebCenter Content using Retrieval-Augmented Generation from user-seeded documentation. The user must upload the latest WebCenter content documentation to a folder and update the folder path of the WccDocumentationAgent in the Agent Studio.
    2. SecurityAgent: An intelligent assistant designed to help the user efficiently manage WCC security resources—including roles, security groups, and privileges. The agent ensures all necessary details are collected before taking action and always seeks user approval for any create, edit, or delete operations.
    3. FrameworkFolderAgent: An AI Agent for managing folders within WCC. It can help user create, delete, move, copy, search, and list child folders. Before performing any actions, the Agent make sures to gather all required information and ask for user approval on any create, delete, or edit operations.
    4. OnboardingAgent: A step-by-step assistant designed to help new users set up Oracle WebCenter Content efficiently and with ease. This agent guides users through the onboarding process, including setting up folders, creating security groups and roles, and assigning privileges. It ensures all necessary inputs are collected and approvals are requested before making any changes. The agent works in coordination with FrameworkFolderAgent and SecurityAgent to perform folder and security operations. It also uses WccDocumentationAgent to answer any questions users may have about WebCenter Content.
      Agent Studio Landing Page

Creating a custom AI Agent

Name Description
Agent Name Name of the agent. It is recommended to have the name of the agent without any spaces
Description Description of the agent
Security Group Security group to which this agent is accessible
Welcome Message Welcome message shown when user starts the conversation with the agent
System Prompt System prompt that is set for the Agent
Enable RAG A check box to enable RAG for this Agent
RAG Type FolderPath:Folder on which RAG should be done. e.g., /user/AgentDocuments
dDocName: List of comma-separated dDocNames on which Agent performs RAG
Note: If neither option is selected, RAG is performed on all documents in the system.
Service Tools Definition Tools available to the Agent in JSON format. See Appendix for details on tool definition.


Create Agent Landing Page

Updating an AI Agent

  1. Click the Edit button for the agent you want to update.
  2. Update the required fields and click Save at the bottom of the page.
    Update Agent Landing Page

Conversing using the Agent created

  1. Log in to the Redwood UI.
  2. Click the red message icon in the bottom right corner of the page.
    RedwoodUI Landing Page
  3. Select the agent you want to converse with.
    RedwoodUI Agent Selection
  4. Send an instruction or question to the agent.
    RedwoodUI Conversation

Appendix

1. Tool Definition

Almost any IDCService in WebCenter Content can be exposed as a Tool in the WCC Agent Framework. This allows an Agent to perform actions within WebCenter Content. Any number of tools can be provided to an Agent as a JSON array.

Sample Tool Definition

{
   "tools":
   [{
      "name": "getFolderDetailsByFolderID",
      "description": "Returns details of the folder specified by folderID.",
      "parameters": [
         {
            "name": "folderID",
            "description": "ID of the folder",
            "required": true
         }
      ],
      "services": [
         {
            "name": "FLD_INFO",
            "mapping": [
               {
                  "name": "fFolderGUID",
                  "value": ""
               }
            ],
            "responseMapping": [
               {
                  "name": "folderPath",
                  "includeFields": ["folderPath"]
               },
               {
                  "name": "folderDetails",
                  "resultSet": "FolderInfo",
                  "includeFields": ["fFolderName", "fParentGUID"]
               }
            ]
         }
      ]
   }]
}

Sample tool response

{
   "result": "true",
   "response": {
      "folderPath": "/docs/lib",
      "folderDetails": {
         "fFolderName": "lib",
         "fParentGUID": "1912796333DA13A628DC5D64714D48FA"
      }
   }
}

or

{
   "result": "false",
   "response": {
      "message": "This item does not exist."
   }
}

A tool definition consists of two major parts:

1. Tool Metadata (Provided to LLM)

This section describes the tool interface that is visible to the LLM:

Field Description
name Name of the tool.
description A detailed description of what the tool does.
parameters Array of parameters accepted by the tool. Each parameter includes: name, description, and required.

2. Service Invocation Details

This section describes how the tool maps to WebCenter Content IDC services. When the LLM calls the tool, the corresponding service(s) are invoked.

Multiple services can be defined under the services field. They are executed in the order provided. The output of one service will be passed as input (along with tool parameters) to the next. Results from execution of last service will be included in tool response.

a. name

Name of the IDC service to invoke.

b. mapping

Defines the mapping of values to IDC service parameters. Can include literals or mustache templates.

Field Description
name Name of the IDC service parameter.
value Value to be passed. Can be a literal (e.g., "folder") or a template (e.g., "folderID:").

You can reference tool parameters or results from previous service executions using template variables.

c. responseMapping (Optional)

Maps the result of IDC service execution back to the tool response.

Field Description
name Name of the field in the final tool response.
resultSet The result set from the IDC service to extract data from. (Optional)
includeFields List of fields to include from the localData or result set.

If responseMapping is not provided, the tool response will just include ‘result’ field indicating success (true) or failure (false) boolean. If included, tool response will also have ‘response’ field containing results from evaluation of response mappings.