Chapter 4 of 7 · Databricks Context Engineer Associate
Memory Architecture with Lakebase and MLflow
Session vs long-term memory, Lakebase state, memory retrieval, intent and MLflow 3 comparison.
- 3topics
- 16 minfast track
- 1diagrams and exhibits
- 18%of the exam
Topics in this chapter
Take them in order, or jump to the one you need.
- 1Memory Types and Persistent Memory with LakebaseDatabricks Context Engineer: match session, long-term and task memory to the need, use Delta state over a scratchpad, and resume…Diagram6 min
- 2Retrieving Memories from LakebaseDatabricks Context Engineer: choose AI Search or SQL for memory lookups, weigh static against dynamic retrieval, and diagnose…5 min
- 3Intent Resolution and Comparing Configurations in MLflow 3Databricks Context Engineer: resolve user intent before retrieval to stop mismatched answers, and pick the most reliable context…5 min
About this chapter
What this chapter covers
Memory Architecture with Lakebase and MLflow is 18% of the Databricks Context Engineer Associate exam, the second largest chapter. It covers what an agent remembers, where that memory lives, how it comes back into the context window, and how you prove which memory setup works best.
The chapter has three parts: choosing the right memory type and store (session memory, long-term memory, Delta-backed task state and Lakebase checkpointers), retrieving memories well (AI Search or structured queries, static or dynamic loading, too much or too little), and resolving user intent before retrieval while comparing context configurations with MLflow 3.
Questions are single-answer and multiple-selection ("Which TWO ...?") multiple choice. Almost all are scenarios that describe an agent, a trace, a configuration or MLflow evaluation results and ask for the diagnosis or the remediation that meets a stated constraint. There are no exhibits.
How the topics fit together
An agent first works out what the user actually means, then decides which memory holds the answer, then pulls only what the step needs into the window. MLflow 3 tells you whether the whole arrangement is reliable.
- Memory type questions ask whether the need is in the current session, across sessions or in durable task state.
- Store questions ask when a Delta table or Lakebase is required instead of an in-context scratchpad.
- Retrieval questions ask whether a memory lookup is semantic or exact, and whether to load it up front or on demand.
- Volume questions ask you to tell over-retrieval from under-retrieval in a trace.
- Pipeline questions ask where user intent must be resolved and which MLflow 3 run is the reliable one.
Before reading the options, name the scope of the missing information: this turn, this conversation, every conversation for this user, or this multi-step task. The scope picks the memory type, and most distractors use the right technique on the wrong scope.
How to use these pages
Each topic page has the same shape, so you can skim for what you need.
- What it is: the idea in plain words.
- How it works: the mechanism, with tables and examples.
- What to remember for the exam: the facts to memorize and how questions are phrased.
- Common mistakes: the classic traps.
- Check yourself: three questions to test your understanding.
Ready for more practice?
The full Databricks Context Engineer Associate question bank has 600 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Databricks. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.