Hello. I am running performance load tests on our new web messaging deployment before we migrate completely to Genesys Cloud. I have a custom script generating 500 concurrent web messaging sessions, all hitting a Dialog Engine Bot Flow. During the load test, the bot engine starts dropping messages. Some sessions receive the initial greeting, but when the script sends the first intent utterance, the bot simply never replies. The flow execution history shows the bot triggered, but it fails to process the input. Are there specific API rate limits or concurrency caps for Dialog Engine Bot Flows that I am violating with this load test?
Genesys Cloud dynamically scales its microservices, including the Natural Language Understanding engine. However, you are likely encountering the platform API rate limits if your custom script is initializing all 500 sessions simultaneously via the Guest Chat API or Web Messaging Guest API. When you create sessions too rapidly, the edge gateways throttle the requests.
Furthermore, bot flows have internal timeouts. If the message delivery is delayed by rate limiting, the bot flow session may expire before it receives the user utterance.
You must implement a staggered ramp-up in your load testing script to accurately simulate organic traffic.
This exact issue drove me crazy during our PureConnect migration! PureConnect handled massive bursts locally, but you cannot just hammer the Genesys Cloud APIs like that! I spent a week troubleshooting dropped bot sessions. Check your Data Actions inside the bot flow! If your bot is querying an external system to validate the user or look up an order status as soon as the session starts, your external system might be the bottleneck, not the Genesys bot engine! The bot will just hang and drop the session if the Data Action times out. You absolutely must mock your external APIs during a bot load test!