Class GenerativeAgentMemory

Class that manages the memory of a generative agent in LangChain. It extends the BaseMemory class and has methods for adding a memory, formatting memories, getting memories until a token limit is reached, loading memory variables, saving the context of a model run to memory, and clearing memory contents.

Example

const createNewMemoryRetriever = async () => {
  const vectorStore = new MemoryVectorStore(new OpenAIEmbeddings());
  const retriever = new TimeWeightedVectorStoreRetriever({
    vectorStore,
    otherScoreKeys: ["importance"],
    k: 15,
  });
  return retriever;
};
const tommiesMemory = new GenerativeAgentMemory(
  llm,
  await createNewMemoryRetriever(),
  { reflectionThreshold: 8 },
);
const summary = await tommiesMemory.getSummary();

Hierarchy

Constructors

constructor

Properties

addMemoryKey llm memoryChain memoryRetriever mostRecentMemoriesKey mostRecentMemoriesTokenKey nowKey queriesKey relevantMemoriesKey relevantMemoriesSimpleKey verbose reflectionThreshold?

Accessors

memoryKeys memoryVariables

Methods

addMemory clear formatMemoriesDetail formatMemoriesSimple getAddMemoryKey getCurrentTimeKey getMemoriesUntilLimit getMostRecentMemoriesTokenKey getRelevantMemoriesKey loadMemoryVariables saveContext

Constructors

constructor

Properties

addMemoryKey

addMemoryKey: string = "addMemory"

llm

llm: BaseLanguageModel<any, BaseLanguageModelCallOptions>

memoryChain

memoryChain: GenerativeAgentMemoryChain

memoryRetriever

memoryRetriever: TimeWeightedVectorStoreRetriever

mostRecentMemoriesKey

mostRecentMemoriesKey: string = "most_recent_memories"

mostRecentMemoriesTokenKey

mostRecentMemoriesTokenKey: string = "recent_memories_token"

nowKey

nowKey: string = "now"

queriesKey

queriesKey: string = "queries"

relevantMemoriesKey

relevantMemoriesKey: string = "relevant_memories"

relevantMemoriesSimpleKey

relevantMemoriesSimpleKey: string = "relevant_memories_simple"

verbose

verbose: boolean

Optional reflectionThreshold

reflectionThreshold?: number

Accessors

memoryKeys

  • get memoryKeys(): string[]

  • Returns string[]

memoryVariables

  • get memoryVariables(): string[]

  • Returns string[]

Methods

addMemory

  • Method that adds a memory to the agent's memory.

    Parameters

    • memoryContent: string

      The content of the memory to add.

    • Optional now: Date

      The current date.

    • Optional metadata: Record<string, unknown>

      The metadata for the memory.

    • Optional callbacks: Callbacks

      The Callbacks to use for adding the memory.

    Returns Promise<ChainValues>

    The result of the memory addition.

clear

  • Method that clears the memory contents.

    Returns void

    Nothing.

formatMemoriesDetail

  • Method that formats the given relevant memories in detail.

    Parameters

    • relevantMemories: Document<Record<string, any>>[]

      The relevant memories to format.

    Returns string

    The formatted memories as a string.

formatMemoriesSimple

  • Method that formats the given relevant memories in a simple manner.

    Parameters

    • relevantMemories: Document<Record<string, any>>[]

      The relevant memories to format.

    Returns string

    The formatted memories as a string.

getAddMemoryKey

  • Method that returns the key for adding a memory.

    Returns string

    The key for adding a memory as a string.

getCurrentTimeKey

  • Method that returns the key for the current time.

    Returns string

    The key for the current time as a string.

getMemoriesUntilLimit

  • Method that retrieves memories until a token limit is reached.

    Parameters

    • consumedTokens: number

      The number of tokens consumed so far.

    Returns Promise<string>

    The memories as a string.

getMostRecentMemoriesTokenKey

  • Method that returns the key for the most recent memories token.

    Returns string

    The key for the most recent memories token as a string.

getRelevantMemoriesKey

  • Method that returns the key for relevant memories.

    Returns string

    The key for relevant memories as a string.

loadMemoryVariables

  • Method that loads memory variables based on the given inputs.

    Parameters

    • inputs: InputValues

      The inputs to use for loading memory variables.

    Returns Promise<Record<string, string>>

    An object containing the loaded memory variables.

saveContext

  • Method that saves the context of a model run to memory.

    Parameters

    Returns Promise<void>

    Nothing.

Settings

Member Visibility

Theme

On This Page

Generated using TypeDoc