Input
ExecutionExecution trigger for starting the agent loop
AI/Agents
LLM-driven control loop that repeatedly calls referenced Flow functions as tools until it decides to stop
Scores range from 0 to 10. Higher values mean more impact, exposure, or operational weight.
Execution trigger for starting the agent loop
Bit describing the LLM that powers the agent
Conversation history shared with the agent (used for reasoning context)
Maximum number of internal iterations/tool calls before failing
Enable automatic context window management to prevent overflow
Maximum tokens to retain when truncating (default: 32000)
How to handle context overflow: 'truncate' (drop old messages) or 'summarize' (LLM compresses history)
Triggers whenever the agent streams its final response
Latest streamed agent chunk (final response)
Fires when the agent stops (successfully or due to error)
Final assistant response produced when the agent halts
Conversation history enriched with all agent/tool turns
Token usage, cost, and model statistics