If your need is limited with API requests, jan’s advise will work perfectly. Anyway, if you want more flexibility, I suggest you to use RabbitMQ. After getting the data that your flow will do the work with it, push messages (items) to queue and process them in another workflow with RabbitMQ trigger. Combine the best options with your case such as limit parallel processing limit to X, ack when flow is done etc.
My typical scenario involves receiving tasks via HTTP from upstream, performing data integration processing, and then needing something like the downstream LLM to handle about 5 concurrent processes. I haven’t seen concurrency configurations in Basic LLM Chain or OpenAI Chat Model scenarios.
The earlier suggestion to manage concurrency through RabbitMQ is interesting. Are there specific examples of this, and how can different tasks be limited to use different levels of concurrency? This approach would also require the introduction of additional external components. Is n8n considering incorporating concurrency management capabilities in the future? This would be very necessary for many application scenarios, such as API calls (for example, llm).
