Incident response and capstones · Lesson 3 of 12 · 5 min

Diagnose retrieval regression

A retrieval regression should be compared against corpus, embedding, filters and index revisions. Find the changed boundary before tuning unrelated generation settings.

Open the lab Course map

Mastery 0/3Run itExplain itApply it

learner@clairos:~ $

↑ history · Tab complete · Ctrl+C cancel · Esc, Tab to leaveExit 0

Your objectives

Hints are always available. Wrong turns never cost XP.

1
Read the case evidence
Read /opt/ai-lab/cases/aiops-incidents-capstones-003.json. Separate the recorded facts from your assumptions.
Hint & worked example
A retrieval regression should be compared against corpus, embedding, filters and index revisions. Find the changed boundary before tuning unrelated generation settings.
The complete case JSON is read successfully.
```
cat /opt/ai-lab/cases/aiops-incidents-capstones-003.json
```
Common mistake to avoid: Change temperature to make mismatched vector spaces compatible.
2
Isolate the deciding signal
Select the index_reembedded field from the case JSON. What does it establish—and what does it not establish?
Hint & worked example
Re-embed or restore compatible query embeddings because the index was not rebuilt for embed-b.
A successful JSON query isolates index_reembedded.
```
jq .index_reembedded /opt/ai-lab/cases/aiops-incidents-capstones-003.json
```
Common mistake to avoid: Using a different case or an unrelated signal as the justification.
3
Record your operational decision
Choose the justified code: rebuild-compatible-index, change-without-evidence, or ignore-the-signal. Save that one code followed by a newline in /home/learner/lab/aiops-incidents-capstones-003-decision.txt.
Hint & worked example
Re-embed or restore compatible query embeddings because the index was not rebuilt for embed-b.
The decision artifact contains the correct code, not just a matching command in history.
```
printf '%s\n' 'rebuild-compatible-index' > /home/learner/lab/aiops-incidents-capstones-003-decision.txt
```
Common mistake to avoid: Change temperature to make mismatched vector spaces compatible.

Understand the pattern

A retrieval regression should be compared against corpus, embedding, filters and index revisions. Find the changed boundary before tuning unrelated generation settings.

Interpret the evidence

Re-embed or restore compatible query embeddings because the index was not rebuilt for embed-b.

A tempting mistake

Change temperature to make mismatched vector spaces compatible.

All names, measurements, prices, versions, and thresholds in this exercise are fictional training data. The evidence supports this bounded scenario, not a claim about a live provider.

Worked example and expected result

$ cat /opt/ai-lab/cases/aiops-incidents-capstones-003.json

{
  "fixture": "Fictional training data; not a live measurement.",
  "embedding_before": "embed-a",
  "embedding_after": "embed-b",
  "index_reembedded": false
}

Exit status: 0. The command completed successfully in the training fixture.

$ jq .index_reembedded /opt/ai-lab/cases/aiops-incidents-capstones-003.json

false

Exit status: 0. The command completed successfully in the training fixture.

$ printf '%s\n' 'rebuild-compatible-index' > /home/learner/lab/aiops-incidents-capstones-003-decision.txt

(no output; exit 0)

Exit status: 0. The command completed successfully in the training fixture.

Try a variation

Name the assumption that would have to change for “rebuild-compatible-index” to stop being the justified action. Do not silently replace missing evidence with a guess.

Explain it

What does the evidence in “Diagnose retrieval regression” justify?Make the change now and gather the missing evidence afterward.Re-embed or restore compatible query embeddings because the index was not rebuilt for embed-b.Change temperature to make mismatched vector spaces compatible.

Apply it elsewhere

A colleague concludes: “Change temperature to make mismatched vector spaces compatible.” What should you do?Reject that conclusion; return to the case evidence and its stated limits.Accept it because the training command completed.Treat this fixture as a live measurement.

Keep the momentum.

Complete the lab, then explain and apply the pattern.

Next: Review a prompt rollout incident Try an incident drill

Anonymous lesson and mastery progress lasts for this session. Sign in to keep synced lesson XP between sessions; mastery stays device-local. No speed bonus, lost lives, or hint penalties.

Keep learning

Case methodology: evidence, limits, decision