Flows That Don't Trigger Themselves
Stop a flow retriggering on its own writes with a trigger condition, set concurrency deliberately, and filter at the source.
به چه درد میخورد
جریانی که روی همان منبعی مینویسد که تماشایش میکند خودش را دوباره صدا میزند و تا وقتی کانکتور جواب ندهد میدود. اینجا میگوید حلقه را در شرط تریگر ببند نه در شرط داخل جریان، همزمانی را آگاهانه تنظیم کن، و فیلتر را به خود سرویس بسپار. آزمون اصلی همان است که همه ردش میکنند: تغییری بده که خود جریان میداد و ببین هیچ اجرایی شروع نشود.
خود اسکیل
Before you save any flow that writes back to the source it watches, do these three things. A self-triggering flow is the most common way a tenant gets throttled. 1. Break the loop at the trigger, not in the logic. Add a trigger condition in the trigger's settings so the flow never starts on its own writes - typically a check on the modifying account, or on a status column your flow sets. A Condition action inside the flow does not help: by then the run has already started and already cost you a call. 2. Set concurrency control deliberately. Left off, runs go parallel without limit. That is right for independent items and wrong for anything that reads shared state and then writes it, because two runs read the same value and one write is lost. Set the degree of parallelism to 1 when order matters. Know the cost: once the queue is full, further trigger events can be dropped rather than queued. 3. Filter at the source, not in the flow. Use the connector's own filter query to fetch only the rows you want. Fetching five thousand rows and discarding them in a Filter array burns API calls against your limit and makes every run slower than it needs to be. Then test it deliberately: - Make one change as a normal user. Confirm exactly one run started. - Make the change your flow itself would make. Confirm zero runs started. That is the test everyone skips. - Open the run history ten minutes later. A column of runs one second apart is the loop. Turn the flow off before you debug it. The failure this prevents: a flow that stamps a field on the item that triggered it, retriggers, and runs until the connector starts refusing calls - usually noticed because somebody else's flow broke.
نصبش چطور است
1. Open the tool and start a new conversation, project or template. 2. Paste the text below wherever the tool keeps standing instructions — usually called system prompt, instructions, or persona. 3. Save it so it applies to every run rather than a single message.