GPT Chat Task (LLM Task)

Failsafes

Due to the nature of AI chat assistants, there is a risk of infinitely repeating conversations when two auto-responding bots find themselves talking to one another. To mitigate this risk, two failsafe mechanisms are in place:

# Message Count Failsafe

The stop_on_high_messages failsafe prevents a chat from exceeding 200 messages in a single conversation. This counts all message types (user, system, and assistant) within the current chat context.

This message count can be reset by using the message_operation: "set" parameter, which clears previous messages. The failsafe can be disabled by setting stop_on_high_messages: false, or adjusted to a different limit by setting a specific number.

# Interaction Count Failsafe

The stop_on_high_interactions failsafe tracks how many times a specific chat (identified by chat_name) has interacted with a model, with a default limit of 1000 interactions.

Unlike the message count, this counter cannot be reset and persists even when messages are cleared. This provides protection against scenarios where a process might clear messages but still be caught in an infinite bot-to-bot conversation loop. The failsafe can be disabled by setting stop_on_high_interactions: false, or adjusted to a different limit by setting a specific number.

Both failsafe limits are enabled by default to prevent runaway processes.