AI Agent V2 Node
warning
This node is in beta and may not be fully tested or documented. Use with caution.
Predicts text using a Language Model
Properties
| Label | Property | Type | Description | Default Value | Is Required |
|---|---|---|---|---|---|
| Function Configuration | functionConfig | botdojo/types/kitchen_sink | Object | ||
| Max Iterations | maxIterations | botdojo/types/number | Agent will stop after this many iterations. Each iteration the agent can invoke one or more tools. | 10 | |
| Warning Threshold Iteration | warningThresholdIteration | botdojo/types/number | If the number of iterations exceeds this threshold, the warning message will be passed as an assitant message to the LLM. This is used to nudge the LLM to finish the response. | null | ✅ |
| Warning Threshold Message | warningThresholdMessage | botdojo/types/string | Optional. When the number of iterations exceeds the warning threshold, the warning message will be passed as an assitant message to the LLM. | null | ✅ |
| Use Parallel Tool Calling | useParallelToolCalling | botdojo/types/boolean | When enabled, the agent will call tools in parallel if the model supports it. This can speed up the response time of the agent. | true | |
| Show Steps | showStepsToUser | botdojo/types/boolean | When enabled, the tool calls will be shown to the user. | true | |
| Hide Step Details | hideStepDetails | botdojo/types/boolean | When enabled, the inputs and outputs of tool calls will be hidden from the user. The step name will still be shown. | false | ✅ |
| Disable Final Output Message | disableFinalOutputMessage | botdojo/types/boolean | By default, if the LLM finishes without calling the final_output tool, it will be prompted to call it. When enabled, this setting disables that prompt and results in an error instead. Only applies when the agent needs to return JSON output rather than a single text response. | false | ✅ |
| Custom Final Output Message | customFinalOutputMessage | botdojo/types/string | Override the default final output message prompt. default : 'You must call the final_output tool to complete the response.' | null | ✅ |
| Final Output Prompt Tries | finalOutputPromptTries | botdojo/types/number | Defaults to 3. When the Output is not a single string, The agent needs to call the final_output tool to return the JSON. This setting controls how many times the agent will attempt to call the final_output tool before giving up. | 3 | ✅ |
| Validate Output with Tool | enableValidatorTool | botdojo/types/boolean | This will add a 'Validator Tool' input to this agent that determins if the agent is done. The tool must take the output of this agent as is inputs and return { success: true, message: 'The output is valid' } if it is valid, or { success: false, message: 'The output is invalid' } if it is invalid. | null | ✅ |
| Validate Output Tries | validateOutputTries | botdojo/types/number | Number of unsuccesful calls to the 'Validator Tool' before throwing an error. Only applies when Validate Output with Tool is enabled. Defaults to 3. | 3 | ✅ |
Output
| Label | Property | Type | Description | Default Value |
|---|---|---|---|---|
| Output | output | botdojo/types/data | Output | null |