Overview
Domain 6 covers Oracle AI Database Private Agent Factory, a no-code and low-code platform for building, deploying and running AI agents next to your own data. It is delivered as a container you run in your own environment, on premises or in OCI, and it uses Oracle AI Database 26ai for its metadata and vector storage.
The visual Agent Builder lets you assemble agents and workflows from nodes, and prebuilt agents cover common jobs such as answering questions from a document knowledge base or analysing database tables. You can connect to LLMs from OCI Generative AI, other hosted providers, or private models served through Ollama or vLLM. The governance side asks you to choose deployment models, control which data sources and models agents may use, manage users and roles, and monitor agent activity. Interoperability through MCP is also part of the picture.
How it's tested
Questions describe an organisation's constraints, such as no data leaving the network, a need for business users to build agents, or audit requirements, and ask which Private Agent Factory capability or deployment choice satisfies them. Some ask you to distinguish it from Select AI Agent.
-
Choose the deployment
Data residency and model constraints lead to a private deployment with a local model or a cloud deployment with OCI Generative AI.
-
No-code or SQL-based
The builder's skills and the workflow's complexity decide between Private Agent Factory and Select AI Agent.
-
Apply a governance control
A risk such as unapproved data sources or unknown agent behaviour maps to roles, approved connections or monitoring.
A defence contractor wants business analysts to build document Q&A agents, and no data or prompts may leave its data centre. Private Agent Factory runs as a container on its own servers, the knowledge agent stores vectors in Oracle AI Database 26ai locally, and a model served through vLLM or Ollama keeps inference private. A hosted LLM endpoint would break the stated constraint.
Topics 4
Official topics from the Oracle outline, in outline order.
6.1 Describe Private Agent Factory architecture and deploymentContainer deployment, database repository and model connections.
- What it means
- Private Agent Factory runs as a containerised application that you install in your own environment, such as an on-premises server or an OCI compute instance. It connects to Oracle AI Database 26ai for storing configuration, knowledge-base vectors and agent data, and to one or more LLM endpoints that you configure.
- What to understand
- Because everything runs inside your boundary, you decide where data lives and which models see it. Hosted model providers are optional; local serving through Ollama or vLLM keeps inference private. Sizing depends on the model choice, since local LLMs need GPU or significant CPU resources while hosted ones shift that load elsewhere.
- How it shows up
- A scenario lists residency or network rules and asks which deployment and model combination meets them.
- Common confusion
- Assuming Private Agent Factory is a multi-tenant SaaS service run by Oracle; it is deployed in your own environment.
6.2 Build agents with Agent Builder and prebuilt agentsVisual workflows, knowledge agents and data analysis agents.
- What it means
- Agent Builder is a visual canvas where you connect nodes such as LLM steps, tools, data sources and conditions to form an agent or workflow. Prebuilt agents give ready starting points, including a knowledge agent that answers from uploaded documents and a data analysis agent that queries database tables in natural language.
- What to understand
- No-code building lets domain experts create agents without PL/SQL, while developers can extend flows with custom tools and MCP connections. Knowledge agents handle chunking, embedding and retrieval for you, which is the same RAG pattern covered earlier, packaged for non-specialists.
- How it shows up
- Expect a question asking which prebuilt agent suits a document Q&A or table analysis request.
- Common confusion
- Thinking Agent Builder requires writing DBMS_CLOUD_AI_AGENT code; it is a visual, no-code builder.
6.3 Govern data sources, models and user accessRoles, approved connections and data boundaries.
- What it means
- Governance in Private Agent Factory means controlling who can build or use agents, which data sources and models are connected, and what each agent may reach. Administrators register database connections, document stores and LLM endpoints, and users receive roles that separate building and publishing from simply using agents.
- What to understand
- Restricting connections to approved sources stops builders attaching sensitive systems on their own. Database connections should use least-privilege accounts, because the agent inherits those rights. Choosing private models for sensitive workloads and hosted models for public content is a common policy split.
- How it shows up
- Questions describe an unapproved data source or model and ask which control prevents it.
- Common confusion
- Believing agent builders should manage their own credentials; administrators register approved connections.
6.4 Monitor and operate deployed agentsActivity tracking, evaluation and lifecycle management.
- What it means
- After agents are published, operators need to see how they behave. Private Agent Factory records agent runs and interactions so teams can review responses, tool calls and errors, refine prompts and workflows, and retire agents that are no longer needed. Updates to the platform container follow the organisation's own change process.
- What to understand
- Monitoring closes the loop between building and governance: it shows whether agents use approved tools, answer accurately and stay within scope. Logs also support audits. Since the platform is self-hosted, patching, backups of the repository database and capacity planning are the customer's job.
- How it shows up
- A scenario asks how to investigate a poor agent answer or prove which data an agent used.
- Common confusion
- Assuming Oracle patches a self-hosted Private Agent Factory automatically like an Autonomous service.
High-yield concepts
The terms, rules and formulas this domain turns into questions most often.
-
Private Agent Factory
No-code platform for building and running agents in your own environment.
- Remember
- Deployed as a container, not SaaS.
- Tested as
- Deployment and fit questions.
- Trap
- Treating it as an Oracle-hosted service.
-
Repository database
Oracle AI Database 26ai storing configuration and vectors.
- Remember
- Back it up like any production database.
- Tested as
- Architecture questions.
- Trap
- Forgetting it holds knowledge-base vectors.
-
Agent Builder
Visual canvas for composing agents and workflows from nodes.
- Remember
- No PL/SQL required.
- Tested as
- Build questions.
- Trap
- Confusing it with DBMS_CLOUD_AI_AGENT.
-
Knowledge agent
Prebuilt agent answering from uploaded documents.
- Remember
- Packages the RAG pattern.
- Tested as
- Document Q&A scenarios.
- Trap
- Building RAG by hand for a simple use case.
-
Data analysis agent
Prebuilt agent querying database tables in natural language.
- Remember
- Uses least-privilege connections.
- Tested as
- Analytics scenarios.
- Trap
- Connecting it as a schema owner.
-
Private model serving
Running LLMs locally through Ollama or vLLM.
- Remember
- Keeps prompts and data inside the boundary.
- Tested as
- Residency scenarios.
- Trap
- Choosing a hosted endpoint under no-egress rules.
-
Approved connections
Administrator-registered data sources and model endpoints.
- Remember
- Stops ad hoc attachment of sensitive systems.
- Tested as
- Governance questions.
- Trap
- Letting builders add credentials freely.
-
Role separation
Different rights for administrators, builders and users.
- Remember
- Limits who can publish agents.
- Tested as
- Access control questions.
- Trap
- Giving every user builder rights.
-
Agent monitoring
Review of runs, tool calls and errors after deployment.
- Remember
- Supports improvement and audit.
- Tested as
- Operations questions.
- Trap
- Assuming agents need no oversight once live.
Side by side
Concepts candidates mix up, compared on the point that decides the answer.
Private Agent Factory vs. Select AI Agent
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Private Agent Factory | Visual, container-deployed agent platform | No-code builders, any environment you run | Assuming it needs PL/SQL skills |
| Select AI Agent | PL/SQL agent framework in Autonomous AI Database | SQL-native, managed with the database | Expecting a drag-and-drop canvas |
Model choices for agents
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| OCI Generative AI | Oracle-hosted models in OCI | Managed, data goes to the OCI service | Using it under strict no-egress rules |
| Other hosted providers | Third-party model APIs | Wide choice, external data flow | Ignoring contract and residency terms |
| Local Ollama or vLLM | Models served on your own hardware | Full privacy, you manage capacity | Underestimating GPU needs |
Exam traps
Each trap is an answer that looks right. Learn the clue that should change your mind.
-
It is SaaS
- What candidates pick
- Assume Oracle hosts and patches Private Agent Factory for you.
- Why it looks right
- Many Oracle AI services are managed.
- Why it's wrong
- Private Agent Factory is a container you deploy and operate, so patching, backups and capacity are your responsibility.
Clue to watch for The scenario mentions on-premises or self-managed infrastructure.
-
Hosted model is private enough
- What candidates pick
- Use a hosted LLM when no data may leave the network.
- Why it looks right
- Cloud providers offer strong security.
- Why it's wrong
- Any hosted endpoint receives prompts and context; only locally served models meet a strict no-egress rule.
Clue to watch for Phrases such as air-gapped or no data leaves the data centre.
-
Builders own connections
- What candidates pick
- Let each builder attach any database they can log into.
- Why it looks right
- It speeds up agent building.
- Why it's wrong
- Governance relies on administrators registering approved, least-privilege connections so sensitive systems are not attached ad hoc.
Clue to watch for The question is about preventing unapproved data access.
-
Agent Builder is PL/SQL
- What candidates pick
- Pick DBMS_CLOUD_AI_AGENT code for business analysts.
- Why it looks right
- Both build Oracle agents.
- Why it's wrong
- Agent Builder is a visual canvas aimed at non-developers; PL/SQL agent objects suit database developers.
Clue to watch for The builders are analysts without SQL skills.
-
Live agents need no review
- What candidates pick
- Publish agents and rely on initial testing alone.
- Why it looks right
- Testing seemed thorough.
- Why it's wrong
- Behaviour drifts with new data and prompts; monitoring runs and tool calls is how teams catch errors and prove compliance.
Clue to watch for The scenario involves audit or quality complaints after launch.
Answer framework
Use these steps for Private Agent Factory deployment and governance scenarios.
- List the constraints first: data residency, network egress, who will build agents and what systems they must reach.
- Pick the deployment location and model serving option, choosing local Ollama or vLLM when nothing may leave the boundary.
- Choose a prebuilt agent or a custom Agent Builder flow that matches the job, such as document Q&A or table analysis, and only build a custom flow when no prebuilt agent covers the requirement.
- Register approved data sources and models centrally with least-privilege accounts and assign builder and user roles.
- Monitor agent runs and tool calls, refine flows from the evidence, and own patching and backups of the platform, since the customer manages the platform rather than Oracle.
Practice question
Want more like this? The 1Z0-1159-26 practice test has questions for every domain.
Checklist
0 of 10
Tick each one you can do without notes. Your ticks stay in this browser.
Study strategy
Read the Private Agent Factory documentation overview first and draw the architecture: container, repository database, data sources and model endpoints. That diagram answers most deployment questions.
If you can, install it on a lab VM or OCI compute instance by following Oracle's installation guide. Connect an Oracle AI Database 26ai instance as the repository, add one hosted model and one local model through Ollama, and build a knowledge agent over a handful of documents. Then try the data analysis agent against a small schema using a read-only user. Open the Agent Builder canvas and extend a flow with an extra step so you can describe nodes in your own words.
After the hands-on work, write a governance checklist covering who registers connections, which roles exist, which models are allowed for which data, and where agent activity is reviewed. Compare Private Agent Factory with Select AI Agent in a two-column table, since fit questions ask you to choose between them. Finish with scenario practice that stresses residency and access constraints. Because the product changes often, check the release notes for renamed features shortly before your exam date.
FAQ
How much of 1Z0-1159-26 is Domain 6, Govern Private Agent Factory Deployments?
Oracle publishes no weights. Third-party practice material gives this Private Agent Factory section a smaller share, mostly deployment and governance scenarios.
Is Private Agent Factory a cloud service?
No. It is a containerised platform you install and run in your own environment, on premises or on OCI compute, so you manage its operation.
Which models can it use?
Oracle documents hosted options such as OCI Generative AI and private serving through tools like Ollama or vLLM. Pick local serving when data must not leave your network.
What does the knowledge agent do?
It answers questions from uploaded documents by chunking, embedding and retrieving them, packaging the RAG pattern for non-developers.
How is it different from Select AI Agent?
Private Agent Factory is a visual builder you deploy yourself. Select AI Agent is a PL/SQL framework inside Autonomous AI Database aimed at database developers.
Who should register data connections?
Administrators should register approved sources with least-privilege accounts, and builders should use only those connections.
Other Artificial Intelligence & Machine Learning certifications
Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.
Test your Domain 6 knowledge
You have covered Private Agent Factory deployment, Agent Builder, prebuilt agents, governance of data and models, and monitoring. Check how you handle deployment and governance scenarios.
Keep going
- Training Training Built on your Weak areas
- Practice test Questions by domain, with explanations
- Mock exam Full length, timed like the real exam
- Readiness test Find your weakest domain in minutes
- Cheat sheet Key terms and formulas on one page
- Ask the 1Z0-1159-26 AI tutor Get any concept explained your way