Evidence-backed answers

Tutor

Every answer is assembled from published lessons, concepts, procedures and recorded claims in this academy, and every Oracle-specific statement carries a citation you can open. When the evidence does not cover a question, the tutor says so instead of answering from memory.

What it will not do

Invent a setup path, a role name or a REST endpoint. Those are the three things implementers most need and this academy publishes none of them without a source.
Answer from an unverified claim without saying so. Unverified material is shown with its status attached.
Include unpublished drafts. Author preview is a separate mode, not this one.
Answer questions about how the tutor itself is built, its configuration, or any stored credential.

Modes

Explain simply: Plain-language explanation of one concept, no jargon. (up to 6 sources)
Explain deeply: Full mechanism, dependencies, failure modes and release caveats. (up to 16 sources)
Compare: Contrast two concepts side by side, e.g. entry vs calculation rules. (up to 16 sources)
Show process: Walk a process in order, naming the owner of each step. (up to 10 sources)
Quiz me: Ask questions and explain the rationale for each answer. (up to 10 sources)
Scenario: Present a situation and let the learner choose what to do next. (up to 10 sources)
Troubleshoot: Work a symptom backwards to a root-cause category. (up to 16 sources)
Recall: Test retrieval rather than recognition. (up to 6 sources)
Interview: Act as an interviewer probing design reasoning. (up to 16 sources)

How it is built

No language model is called in this build. Answers are composed by quoting retrieved evidence verbatim, which is why the tutor cannot hallucinate a setup path: there is no step where anything is written from memory. The provider is selected by configuration, so a model-backed provider can be added without changing this page or the policy gate.