Versioning prompts, models, and retrieval
On This Lesson 4 topics
This lesson is planned. The outline below lists the topics it will cover. The full lesson has not been published yet.
In this lesson
make AI behavior reproducible as versioned configuration.
Planned topics
- Store prompts, model identifiers (pinned, not "latest"), parameters, and retrieval settings in version control as one configuration bundle with an ID.
- Deploy configuration like any release: reviewed, promoted through environments, reversible.
- Log the bundle ID, model version, and retrieved document IDs with every answer, so any answer can be traced.
- Keep the evaluation set versioned alongside it.
Draft outline details
Demonstration
assistant-config: v42
model: <pinned model id; verify provider naming before recording>
temperature: 0.2
prompt: prompts/order-support@v42.md
retrieval: { index: orders-faq@2026-10, top_k: 5 }
eval-set: evals/order-support@v17Check your understanding
What must be logged with each answer to trace it later?
Show answer
Config bundle ID, model version, and retrieved sources.
Practice
Apply this lesson with the module's lab, exercises and quiz.
Marking a lesson complete is optional and you can undo it any time. Your progress stays in this browser.