Construct a scratchpad to let the agent continue its thought process
Decide what to do given some input.
Steps the LLM has taken so far, along with observations from each.
User inputs.
Optional
callbackManager: CallbackManagerCallback manager to use for this call.
Action specifying what tool to use.
Prepare the agent for output, if needed
Return response when agent has been stopped due to max iterations
Optional
callbackManager: CallbackManagerStatic
createCreate prompt in the style of the zero shot agent.
List of tools the agent will have access to, used to format the prompt.
Optional
args: ZeroShotCreatePromptArgsArguments to create the prompt with.
Static
deserializeLoad an agent from a json-like object describing it.
Static
fromLLMAndCreates a ZeroShotAgent from a Large Language Model and a set of tools.
The Large Language Model to use.
The tools for the agent to use.
Optional
args: ZeroShotCreatePromptArgs & AgentArgsOptional arguments for creating the agent.
A new instance of ZeroShotAgent.
Static
getReturns the default output parser for the ZeroShotAgent.
Optional
fields: OutputParserArgsOptional arguments for the output parser.
An instance of ZeroShotAgentOutputParser.
Static
validateValidates the tools for the ZeroShotAgent. Throws an error if any tool does not have a description.
List of tools to validate.
Generated using TypeDoc
Agent for the MRKL chain.