ENGINEERING
Shadow-traffic A/B bench two Replicate versions before cutover
On a schedule, replays a sampled slice of recent production inputs through both the live and candidate Replicate versions, compares quality and cost side by side.
How it runs
The automated pipeline, trigger to output.
- TriggerScheduled run pulls sampled production inputsBigQuery
- ActionReplay slice through live + candidate versionsReplicate
- LogicCompare quality, latency, cost head to head
- ActionWrite A/B comparison reportBigQuery
- OutputPost cutover recommendation to SlackSlack
What it does
Runs a shadow A/B comparison so you can judge a candidate Replicate version on real recent traffic, not just a synthetic bench. It replays a sampled slice of production inputs through both the live and candidate versions, scores them head to head on quality, latency, and cost, and produces a cutover recommendation.
When to use it
Use it before swapping production traffic to a new version when you need confidence on representative inputs and want to weigh quality against inference cost. Ideal for cost-sensitive endpoints where a marginally better model may not justify a price increase.
How it works
- 1A schedule trigger pulls a sampled slice of recent production inputs from BigQuery.
- 2The flow replays each input through the live Replicate version and the candidate version.
- 3It scores both outputs head to head and tallies latency and per-call cost.
- 4A logic step decides recommend-cutover or hold based on the quality-versus-cost trade-off.
- 5It writes the full A/B comparison report to BigQuery for the record.
- 6It posts the side-by-side summary and recommendation to Slack.
Set it up
What you configure once, before turning it on.
- 1Connect BigQueryDatasets, queries, schemas.
- 2Connect ReplicateImage, video, and model inference.
- 3Connect SlackChannels, DMs, threads, mentions.
- 4Set each agent's modelWe leave models unset so you pick the tier — fast + cheap, or top-quality.
- 5Tune it to your dataEdit the prompts, filters, and field mappings so it matches how your team works.
- 6Test, then turn it onRun once against a sample, confirm the output, then enable the trigger.
More Engineering workflows
Upgrade Impact Router to Module Code Owners
Maps a dependency-bump PR's affected modules to their CODEOWNERS, then DMs each owner on Slack with only the changelog slice that touches code they own.
Re-Voice IVR Prompts on Phone-Tree Config Merge
When a phone-tree config change merges in GitHub, regenerates the ElevenLabs audio for any prompt whose script changed in the diff and opens a follow-up PR adding the new audio…
Agent reviews model-license fit and suggests compliant swaps on the PR
When a PR adds a Hugging Face model, an agent reads the model card and license, judges fit against your commercial-use policy.
Scan for deprecated endpoints and email consumers a weekly sunset countdown
On a weekly schedule, scans the OpenAPI spec for endpoints marked deprecated with a sunset date, and emails each consuming team a countdown of how many days remain before removal.
Publish a versioned API changelog to Confluence on each release tag
On a new semver release tag, gathers the contract changes since the last release and writes a clean.
Gate breaking API PRs behind downstream consumer acknowledgement
When a PR introduces a breaking contract change, comments the impact summary back on the PR, applies a blocking label.
Run it inside a business
This workflow drops into a full company template. Import the org, and this is one of the playbooks its agents run.
Run this workflow in your colony.
14-day trial. No DevOps. No Sales call. Provisioned in under a minute.