As Oracle continues to expand its AI ecosystem, a powerful new addition has emerged for organizations looking to build enterprise-grade AI agents directly on top of their Oracle databases: Oracle AI Database Private Agent Factory (PAF).
This platform enables developers, architects, and data professionals to create sophisticated AI agents and multi-step workflows through a visual drag-and-drop interface while leveraging the native AI capabilities already embedded within Oracle Autonomous Database.
Unlike standalone AI orchestration frameworks, Private Agent Factory tightly integrates with Oracle Database AI capabilities, including the DBMS Cloud AI framework, allowing organizations to build secure, private, and database-centric AI solutions within their OCI environment.
In this article, I'll walk through the complete deployment process, from infrastructure setup to installing and configuring the PAF platform.
What is Oracle AI Database Private Agent Factory?
- Build AI agents visually
- Create complex agent workflows
- Connect to Oracle Autonomous AI Database
- Integrate multiple LLM providers
- Utilize Select AI and MCP capabilities
- Create Retrieval-Augmented Generation (RAG) solutions
- Develop knowledge-based assistants
- Configure custom tools and integrations
- Orchestrate multi-agent processes
The platform can be deployed directly from OCI Marketplace and is designed for rapid implementation.
Architecture Overview
Before beginning the installation, let's understand the major components involved.
OCI Components
- OCI Compartment
- IAM Policies
- Autonomous AI Database
- Virtual Cloud Network (VCN)
- OCI Marketplace Stack
- Compute Instance (PAF Application)
- Oracle AI Database Private Agent Factory
- Oracle Generative AI Service
- Embedding Models
- Chat Models
- Knowledge Bases
- Agent Workflows
Prerequisites
- Permissions to create OCI resources
- Access to OCI Marketplace
- Ability to create IAM policies
Recommended Resources
- Autonomous Database
- Compute Shape : E5 Flex or E6 Flex
- VCN : Dedicated Network
- LLM Provider : OCI Generative AI
Step 1: Create a Dedicated OCI Compartment
Navigate to:
Identity & Security -> Compartments
Create Compartment
Example:
Name: PrivateAgentFactoryCompartment
Description: PrivateAgentFactoryCompartment
Using a dedicated compartment simplifies:
- Resource management
- Cost tracking
- Security governance
- Environment isolation
Step 2: Configure IAM Policies
Before deploying resources, appropriate permissions must be granted.
Let's navigate to: Identity & Security -> Policies
Create a policy at the root tenancy level.
For a Sandbox or testing environments, you can create a broad policy allowing users to manage resources within the target compartment.
In production environments, Oracle recommends implementing more granular least-privilege policies.
Example:
Allow group Administrators to manage all-resources in compartment PrivateAgentFactoryCompartment
This policy enables users to:
- Create databases
- Deploy marketplace stacks
- Manage networking
- Provision compute resources
Step 3: Create an Autonomous AI Database
Let's Navigate to:
Oracle Database -> Autonomous AI Database
Select your compartment and click: Create Autonomous Database
Configure Database Details
Database Name
Workload Type
Choose one of the following:
- Data Warehouse
- Transaction Processing
For this deployment, I've chosen ATP (Transaction processing).
Compute Resources: Increase CPU allocation beyond the default configuration to provide adequate capacity for AI workloads.
Example: 4 ECPUs
Specify the Administrator Password
Create Autonomous Database
Database provisioning typically takes: 3 - 4 minutes
Step 4: Create the Database User Required by PAF
After the database becomes available:
Navigate to: Database Actions -> SQL
Open the SQL Worksheet.
GRANT CONNECT, RESOURCE, CREATE TABLE, CREATE SYNONYM, CREATE DATABASE LINK,
CREATE ANY INDEX, INSERT ANY TABLE, CREATE SEQUENCE, CREATE TRIGGER, CREATE USER, DROP USER TO <DB_USER>;
GRANT CREATE SESSION TO <DB_USER> WITH ADMIN OPTION;
GRANT READ, WRITE ON DIRECTORY DATA_PUMP_DIR TO <DB_USER>;
GRANT SELECT ON SYS.V$PARAMETER TO <DB_USER>;
What does this script do ?
- Creates the application user
- Grants privileges
- Enables required database capabilities
Modify this script and replace the placeholder username with your desired value like PAF_USER
Step 5: Download the Autonomous Database Wallet
Navigate to: Database Connection
Click: Download Wallet
Configure Wallet Password: Enter a wallet password.
Important: The wallet password does not need to match the database password.
Download the wallet ZIP file and save it securely.
Example: Wallet_AJATP.zip
This file will be uploaded during application installation.
Step 6: Create a Virtual Cloud Network (VCN)
Navigate to: Networking -> Virtual Cloud Networks
Let's create the VCN using Actions -> VCN Wizard
Select Create VCN with Internet Connectivity
Proceed through the wizard using default values and create the VCN.
Provisioning generally completes within seconds.
Step 7: Configure Security Rules
Once the VCN is created, navigate to: Security Lists
Add the required Ingress and Egress rules.
Ingress Rules:
Port 8080: Required for application access.
TCP 8080
Database Connectivity Port: Allow outbound traffic to Autonomous Database.
TCP 1521, 1522
Depending on your database configuration, these ports may vary.
Egress Rules:
0.0.0.0/0 - All Protocols
Step 8: Deploy Oracle AI Database Private Agent Factory
- Navigate to OCI Marketplace -> All Applications
- Search for: Oracle AI Database Private Agent Factory
- Launch Stack
- Accept Marketplace terms.
- Leave most options at default values.
Configure Stack Parameters:
- Name: Provide a desired name for the stack
- Region: Provide your tenant region
- Compartment: Select the previously created compartment for PAF.
- VCN: Select the previously created VCN for PAF.
- Network Type: Public
- Virtual Machine Shape: VM.Standard.E5.Flex
- OCPU Count: 8
- Memory: 64GB
- Create Stack
Provisioning typically takes 5 minutes
Step 9: Launch the Installer
- Once the stack creation completes, navigate to the Jobs section.
- Click on the Job
- Once the Job is opened, navigate to Logs tab
- Scroll all the way to the end the log
- Here, we'll find the installation URL for our Private Agent Factory environment
Example:
Agent_Factory_URL = "https://<Host IP>:8080/agentFactory/installation"
Bookmark this URL as this would be actual URL to access Private Agent Factory platform: https://<Host IP>:8080/agentFactory
Step 10: Create the Initial Platform User
- Open the Agent Factory installation URL (obtained in the previous step) in a browser.
- You will be presented with a installation Wizard.
- Create your administrator account.
- This account will become the initial platform administrator.
Step 11: Configure Database Connectivity
- Navigate to the Database configuration tab. Here, we will provide database connection information.
- Upload the AI database wallet (previously downloaded).
- Provide: Database User and Database Password for the dedicated PAF_USER (previously created).
- Click Test Connection
- A successful validation confirms connectivity between the platform and Autonomous Database.
- Click Next and click Install on the next screen
Step 12: Monitor Installation
The installer displays deployment logs showing:
- Database schema creation
- Metadata initialization
- Application configuration
- Component deployment
We need to wait until installation completes.
The Next button becomes available once all tasks succeed.
Step 13: Configure the LLM Provider
The next step is configuring the AI model provider.
- I've selected OCI Generative AI as the provider but you can select ay other option if you have the API keys from those providers.
For OCI Gen AI, we first need to obtain below details:
- Endpoint URL
- Compartment OCID
- User OCID
- Tenancy OCID
- Fingerprint
- Region
- API Signing Key
Step 14: Generate OCI API Keys
- To complete the above step, we first need to create API keys for a given user so that we can use the same in the LLM configuration.
- Navigate to User Profile -> API Keys
- Generate and download the new API keys (Public and Private).
- It will also show the Configuration file that will contain User OCID, Tenancy OCID, Fingerprint, Region.
- Save these details for later use.
Step 15: Configure Chat Model
- Now, let's use these values and populate in the LLM configuration from Step 13
- Endpoint will be the standard Inference API URL based on your region (Chicago in my case)
- Choose the model of your choice. For example "meta.llama-4-maverick-17b-128e-instruct-fp8"
- Enter User OCID, Tenancy OCID, Compartment OCID, Region, Fingerprint based on the values obtained in the previous step above.
- Upload the Private key file
- Test the Connection
- Save the Configuration
Step 16: Configure Embedding Model
- Repeat the same process for embeddings.
- Endpoint will be the standard Inference API URL based on your region (Chicago in my case)
- Choose the model of your choice. For example "cohere-embed-v4.0"
- Enter User OCID, Tenancy OCID, Compartment OCID, Region, Fingerprint based on the values obtained in the previous step above.
- Upload the Private key file
- Test the Connection
- Save the Configuration
- Finish the installation
Step 17: Complete Installation
- Once you finish the installation, the PAF platform is now fully configured.
- Log in using the credentials created earlier.
Once logged in, you will land in the actual Private Agent Factory platform and it's ready for you to explore, build and publish various AI agents.
I will be covering each feature of this platform along with the hands-on examples in my subsequent blogs so stay tuned !
Conclusion
Oracle AI Database Private Agent Factory provides a powerful, enterprise-ready framework for building AI agents directly on top of Oracle Autonomous AI Database. With its visual drag-and-drop interface, built-in workflow orchestration, multi-model support, and deep database integration, organizations can rapidly build intelligent applications without assembling a complex AI infrastructure stack from scratch.
The deployment process can be completed in less than an hour, and it's an excellent platform for experimentation, proof-of-concepts, as well as enterprise AI development.
Whether you're building conversational assistants, RAG solutions, document-processing pipelines, or multi-agent workflows, Oracle AI Database Private Agent Factory offers a streamlined path from idea to implementation.
0 comments:
Post a Comment