Last verified: September 21, 2026
What You Can Now Do
Context Memo now runs both the scan and feeder steps of your pipeline from a single button click on the Pipeline tab for a brand. Before this, scan and feeder were two separate triggers you had to fire in sequence, waiting for the first to finish before starting the second. Now one click queues the full workflow and the steps hand off to each other. The decision it changes: you no longer have to sit on the Pipeline tab babysitting step one so you can start step two. Kick off the run, close the tab, come back to results.
Where It Is in Context Memo
Open a brand, then go to the Pipeline tab. The new Run Pipeline button sits on that screen alongside the existing per-step controls. Those individual scan and feeder triggers are still there if you want to run one step on its own. Run Pipeline is the combined path.
How to Use It
- Open the brand you want to run, and select the Pipeline tab. You'll see the pipeline steps for that brand and the current state of each.
- Click Run Pipeline. The request hits the new pipeline endpoint, which hands execution to an Inngest function that coordinates the multi-step run. You get confirmation that the run is queued.
- Watch the scan step execute first. Scan runs against your configured models and prompts, and the results land in the brand's scan data as they complete.
- Let the feeder step pick up automatically. When scan finishes, the orchestration triggers feeder without another click from you. No second trigger, no manual handoff.
- Return to the Pipeline tab to confirm both steps completed. Step status on that screen tells you whether the run finished clean or stopped partway.
- Run individual steps when you need to. If you only want a fresh scan, or only want to re-run feeder against existing scan data, use the per-step controls instead. Run Pipeline is for the full sequence.
Why We Built It
Customers told us the same thing repeatedly: running scan and feeder as separate operations was friction they shouldn't have to absorb. The pipeline is one logical job. Splitting it into two manual triggers meant someone had to remember to come back, check whether scan had finished, and then start feeder. When that person got pulled into something else, feeder didn't run at all. Scan data sat there and nothing downstream happened with it. That's a pipeline that looks like it's working and isn't.
The cost was quiet. A skipped feeder step doesn't throw an error. It just means content that should have moved through your pipeline didn't. You find out later, when you're looking at citation coverage and wondering why a set of prompts never produced anything. Orchestrating the two steps in one action removes the window where a human has to intervene for the job to complete.
What It Does Not Do Yet
A few limits worth knowing before you rely on this:
- Run Pipeline is manual, not scheduled. The button starts a run on demand. It does not set a recurring cadence. If you want scan and feeder to run on a schedule, that's still something you trigger yourself or handle outside the Pipeline tab.
- It runs one brand at a time. The button lives on the Pipeline tab for a specific brand. If you manage several brands, you'll click through each one. There's no multi-brand batch run from a single control.
- Orchestration covers scan then feeder. Those are the two steps the new endpoint and Inngest function coordinate. Any step you run after feeder in your own workflow, including review and publishing decisions, stays where it was.
- Partial failures still need a human read. If scan completes and feeder stops, the Pipeline tab shows you that. It does not automatically retry the full sequence or roll back the scan. You decide whether to re-run feeder on its own or start over.
One practical note on sequencing: because the orchestration is genuinely dependent, feeder waits for scan to finish rather than starting in parallel. A run across a wide prompt set and multiple models will take as long as its slowest step. That's expected. You're trading wall-clock time you weren't spending anyway for the guarantee that feeder actually fires.
If you've been running these two steps by hand, switch to Run Pipeline on your next cycle and compare. The output should match what you get from firing the steps separately. The difference is that nothing waits on you in between.
