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.
OptionalcallbackManager: 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
OptionalcallbackManager: CallbackManagerStaticcreateCreate prompt in the style of the zero shot agent.
List of tools the agent will have access to, used to format the prompt.
Optionalargs: ZeroShotCreatePromptArgsArguments to create the prompt with.
StaticdeserializeLoad an agent from a json-like object describing it.
StaticfromLLMAndCreates 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.
Optionalargs: ZeroShotCreatePromptArgs & AgentArgsOptional arguments for creating the agent.
A new instance of ZeroShotAgent.
StaticgetReturns the default output parser for the ZeroShotAgent.
Optionalfields: OutputParserArgsOptional arguments for the output parser.
An instance of ZeroShotAgentOutputParser.
Staticvalidate
Agent for the MRKL chain.
Example
Deprecated
Use the method instead.