Routine steps
Classification, extraction, validation, routing, and repair do not always need the most expensive model in the stack.
Sample Output
What a Flev DevOps pilot should hand back.
This is a sanitized example of the shape of the deliverable. The real report changes with the repo, deployment path, and evidence available.
Model Cost And Privacy
Use frontier models only where they matter.
Flev workflows can route routine, structured steps to local or private small models while keeping stronger models available for complex reasoning.
Reviewable routing
Teams should be able to see which model handled which step, why fallback exists, and who can approve changes.
Small-model ready
Better Call evidence shows tool-call accuracy improving from 73.4% to 83.8% on 3,625 granite4.1:3b BFCL v4 cases.
A release pipeline fails after a dependency and Docker build change. The team needs to know whether to retry, patch, rollback, or change the runbook.
01
Diagnosis brief
Confirmed
The failure occurs after dependency install succeeds and before the image is pushed. The failing step is the Docker build stage.
Likely cause
The build context no longer includes the generated runtime artifact expected by the Dockerfile.
Missing evidence
The last successful artifact list and the exact Dockerfile diff should be checked before approving a patch.
Recommended next action
Add an explicit build-artifact check before Docker build, then update the runbook with the verification command.
02
Evidence table
| Source | What it showed | Decision impact |
|---|---|---|
| CI step log | Install completed, Docker build failed on missing runtime artifact. | Rules out package install as the primary failure. |
| Repository diff | Build script changed output directory without matching Dockerfile update. | Supports a patch plan instead of blind retry. |
| Previous successful run | Artifact existed at the old path before image build. | Shows the runbook should verify artifact location. |
03
Runbook patch
- Before Docker build, list the expected runtime artifact path.
- If the artifact is missing, stop the release and inspect build output changes.
- Only retry the pipeline after the artifact path and Dockerfile agree.
- Record whether the fix was script, Dockerfile, or release configuration.
04
Approval boundary
Read-only
Inspect CI logs, diffs, run history, Kubernetes events, and existing runbooks.
Needs approval
Open a PR, push a patch, publish a package, deploy, rollback, or mutate cluster state.
Never automatic
Production deploys, credential changes, destructive cluster commands, or customer-visible communication.
Flev DevOps
Start with your failure path
Send one failing engineering path. We will scope the smallest useful paid pilot.
Use this checklist before the first call. The goal is not a perfect brief; the goal is enough context to decide whether a 7-day diagnostic sprint can produce a useful result.