Follow a run, answer it, or run it again
Each time an action runs on an object, it’s a run. The Runs tab lists them, newest first, with what each did and what it cost. A run that can’t go on without a person waits there for an answer.
Anyone who can open the Ontology platform can see runs. To answer, run again or skip, you need Respond to runs. Ask your Flowstate admin.
See your runs
Go to Organization → Modules → Ontology platform and open the Runs tab. The tab shows a count when runs need you.
Each run shows its number, the Action, the object it ran On, its Status, when it Started, how long it Took and its AI cost. A try from the action editor shows as Try. Filter by Status or Action.
| Status | What it means |
|---|---|
| Queued | Waiting to start. Only so many of your runs run at once, so others wait their turn. |
| Running | Working now |
| Needs you | Waiting for a person to answer a question |
| Failed | Stopped before it finished. It says what went wrong. |
| Done | Finished |
| Skipped | Didn’t run, or was stopped, and says why, for example Skipped: the object was deleted |
Open a run
Select a run to open it.
- What it did lists each step as it happens: reading the object, asking the model, using a tool, calling an app, changing properties, linking. Each step shows when it happened.
- What it cost shows the Model, Model calls, the tokens it Read and Wrote, its AI cost and its Compute. See Usage, billing and model rates.
Every change a run makes also shows in the type’s History, naming the run.
Answer a question
A run asks when an action can’t decide on its own, like an agent that finds two pieces of Work that fit a deal.
- Filter Status to Needs you, and open the run.
- Read the question under It needs you. What it did shows how it got there.
- Answer. Depending on the question, you choose an option, choose a person, say whether to keep what’s there or use the new value, confirm, or type a number or some text.
- Select Answer and carry on.
The run picks up where it stopped. If the question shouldn’t be answered, select Skip this one.
If two people answer at once, the first answer counts. The second sees “Someone has already answered that one.”
Run a failed run again, or skip it
Open the run. What went wrong says why it failed, and for app problems, which try it was on and when it tries again on its own.
- Run again starts it from the beginning.
- Skip marks it skipped, so it no longer shows as failed.
- Reconnect HubSpot, or the app it needs, appears when the app has signed Flowstate out. Sign in again, then run it again.
To deal with many at once, filter Status to Failed, tick the runs, and select Run again or Skip. You can also apply it to every run that matches the filter.
What went wrong
| Message | What to do |
|---|---|
| The app needs signing in to again | Select Reconnect, sign in, then Run again |
| The app hasn’t allowed this | Allow Flowstate to change things in the app. See Connect an app. |
| The app asked us to slow down or The app didn’t answer | Flowstate tries again on its own. If it still fails, Run again later. |
| The record is gone from the app | It was deleted in the app. Skip it. |
| The object changed while it ran | If the change started a newer run, Skip this one. If not, Run again. |
| The action isn’t allowed to do this | Add what it needs on its Access tab, then Run again |
| Flowstate refused the change | Open the run to see which change, then check the action’s Access and the property’s values. |
| Reached its AI cost limit, time limit or memory limit | Raise the limit on the action’s Limits tab, then Run again |
| Reached the action’s limit for the month | Raise the action’s monthly limit, or wait for next month |
| Reached the organization’s quota for the month | See Monthly limits |
| The code stopped with an error | Fix the code. The steps and log lines show where it stopped. |
| The model didn’t answer | Run again |
| The agent gave up | The agent couldn’t do the job with what it had. Read its last steps, then improve the instructions or the meanings it reads. |
| Something went wrong in Flowstate | Run again. If it keeps happening, contact Flowstate support. |
If something’s not right
A run has been Queued for a long time. Other runs are using your organization’s places. It starts when one finishes.
A run with an Anthropic model fails straight away. If it fails at its first Asked the model step with “The model didn’t answer”, the most likely cause is that the model hasn’t been turned on for your organization yet. Contact your Flowstate account manager to turn it on, then select Run again.
You can’t see Answer and carry on. You need Respond to runs. The run says “Someone with Respond to runs answers this.”