Class: Agent

agents.Agent

Class responsible for calling a language model and deciding an action.

Remarks

This is driven by an LLMChain. The prompt in the LLMChain must include a variable called "agent_scratchpad" where the agent can put its intermediary work.

Hierarchy

• BaseSingleActionAgent

  ↳ Agent

  ↳↳ ZeroShotAgent

  ↳↳ ChatAgent

  ↳↳ ChatConversationalAgent

Constructors

constructor

• new Agent(input)

Parameters

Name	Type
input	AgentInput

Overrides

BaseSingleActionAgent.constructor

Defined in

langchain/src/agents/agent.ts:170

Properties

llmChain

• llmChain: LLMChain

Defined in

langchain/src/agents/agent.ts:158

Accessors

allowedTools

• get allowedTools(): undefined | string[]

Returns

undefined | string[]

Overrides

BaseSingleActionAgent.allowedTools

Defined in

langchain/src/agents/agent.ts:162

---

inputKeys

• get inputKeys(): string[]

Returns

string[]

Overrides

BaseSingleActionAgent.inputKeys

Defined in

langchain/src/agents/agent.ts:166

---

returnValues

• get returnValues(): string[]

Returns

string[]

Inherited from

BaseSingleActionAgent.returnValues

Defined in

langchain/src/agents/agent.ts:31

Methods

_agentType

▸ Abstract _agentType(): string

Return the string type key uniquely identifying this class of agent.

Returns

string

Overrides

BaseSingleActionAgent._agentType

Defined in

langchain/src/agents/agent.ts:198

---

_stop

▸ _stop(): string[]

Returns

string[]

Defined in

langchain/src/agents/agent.ts:231

---

constructScratchPad

▸ constructScratchPad(steps): string | BaseChatMessage[]

Construct a scratchpad to let the agent continue its thought process

Parameters

Name	Type
steps	AgentStep[]

Returns

string | BaseChatMessage[]

Defined in

langchain/src/agents/agent.ts:245

---

extractToolAndInput

▸ extractToolAndInput(_input): Promise<null | { input: string ; tool: string }>

Extract tool and tool input from LLM output.

Parameters

Name	Type
_input	string

Returns

Promise<null | { input: string ; tool: string }>

Defined in

langchain/src/agents/agent.ts:179

---

finishToolName

▸ finishToolName(): string

Name of tool to use to terminate the chain.

Returns

string

Defined in

langchain/src/agents/agent.ts:238

---

llmPrefix

▸ Abstract llmPrefix(): string

Prefix to append the LLM call with.

Returns

string

Defined in

langchain/src/agents/agent.ts:193

---

observationPrefix

▸ Abstract observationPrefix(): string

Prefix to append the observation with.

Returns

string

Defined in

langchain/src/agents/agent.ts:188

---

plan

▸ plan(steps, inputs): Promise<AgentAction | AgentFinish>

Decide what to do given some input.

Parameters

Name	Type	Description
steps	AgentStep[]	Steps the LLM has taken so far, along with observations from each.
inputs	ChainValues	User inputs.

Returns

Promise<AgentAction | AgentFinish>

Action specifying what tool to use.

Overrides

BaseSingleActionAgent.plan

Defined in

langchain/src/agents/agent.ts:297

---

prepareForOutput

▸ prepareForOutput(_returnValues, _steps): Promise<Record<string, any>>

Prepare the agent for output, if needed

Parameters

Name	Type
_returnValues	Record<string, any>
_steps	AgentStep[]

Returns

Promise<Record<string, any>>

Inherited from

BaseSingleActionAgent.prepareForOutput

Defined in

langchain/src/agents/agent.ts:67

---

returnStoppedResponse

▸ returnStoppedResponse(earlyStoppingMethod, steps, inputs): Promise<AgentFinish>

Return response when agent has been stopped due to max iterations

Parameters

Name	Type
earlyStoppingMethod	StoppingMethod
steps	AgentStep[]
inputs	ChainValues

Returns

Promise<AgentFinish>

Overrides

BaseSingleActionAgent.returnStoppedResponse

Defined in

langchain/src/agents/agent.ts:307

---

createPrompt

▸ Static createPrompt(_tools, _fields?): BasePromptTemplate

Create a prompt for this class

Parameters

Name	Type	Description
_tools	Tool[]	List of tools the agent will have access to, used to format the prompt.
_fields?	Record<string, any>	Additional fields used to format the prompt.

Returns

BasePromptTemplate

A PromptTemplate assembled from the given tools and fields.

Defined in

langchain/src/agents/agent.ts:208

---

deserialize

▸ Static deserialize(data): Promise<Agent>

Load an agent from a json-like object describing it.

Parameters

Name	Type
data	Object

Returns

Promise<Agent>

Defined in

langchain/src/agents/agent.ts:347

---

fromLLMAndTools

▸ Static fromLLMAndTools(_llm, _tools, _args?): Agent

Construct an agent from an LLM and a list of tools

Parameters

Name	Type
_llm	BaseLanguageModel
_tools	Tool[]
_args?	Record<string, any>

Returns

Agent

Defined in

langchain/src/agents/agent.ts:217

---

validateTools

▸ Static validateTools(_tools): void

Validate that appropriate tools are passed in

Parameters

Name	Type
_tools	Tool[]

Returns

void

Defined in

langchain/src/agents/agent.ts:229