Chat Memory

In xFlow, Chat Memory is a crucial feature that allows AI agents to retain context across multiple interactions. This capability is essential for creating coherent and contextually relevant conversations. Chat Memory in xFlow supports different types of memory configurations that can be customized to fit specific requirements. This documentation explains the details of the types of memory supported and how to configure them.

Chat Memory Properties

Memory ID

• Name: memoryId

• Type: string

• Description: The unique identifier used to store and retrieve the agent's memory. This ID ensures that the conversation history is correctly associated with the specific AI agent.

• Example:

  memoryId: "session123"

Memory Type

• Name: memoryType

• Type: string

• Description: Defines the type of memory to use. The supported types are message_window and token_window. The default type is message_window.

• Enum Values:

  • message_window: Retains a fixed number of messages.

  • token_window: Retains a fixed number of tokens.

• Default: message_window

• Example:

  memoryType: "token_window"

Max Messages

• Name: maxMessages

• Type: integer

• Description: The maximum number of messages to retain in memory. If the memory exceeds this limit, the oldest messages are evicted. This property is used when the memory type is message_window.

• Example:

  maxMessages: 10

Max Tokens

• Name: maxTokens

• Type: integer

• Description: The maximum number of tokens to retain in memory. The memory will retain as many of the most recent messages as can fit within the maxTokens limit. If an old message does not fit within the limit, it is completely evicted. This property is used when the memory type is token_window.

• Example:

  maxTokens: 1000

Types of Chat Memory

Message Window Memory

message_window memory retains a fixed number of recent messages. This type of memory is useful when the exact number of past interactions needs to be remembered, regardless of the length of the individual messages.

• Configuration Example:

  memory:
    memoryId: "session123"
    memoryType: "message_window"
    maxMessages: 10

Token Window Memory

token_window memory retains a fixed number of tokens. This type of memory is beneficial when controlling the total size of the memory is more important than the number of messages. It ensures that the memory does not exceed a certain number of tokens, making it easier to manage the cost and performance of the AI agent.

• Configuration Example:

  memory:
    memoryId: "session123"
    memoryType: "token_window"
    maxTokens: 1000

Usage Example

Here is an example of how to configure Chat Memory in an AI Agent State within a xFlow workflow:

- name: ExampleAIState
  type: aiagent
  agentName: ExampleAgent
  aiModel: gpt-4o
  systemMessage: "You are an assistant designed to provide accurate answers."
  userMessage: '${ "User: " + .request.question }'
  output: 
    {
      "type": "object",
      "properties": {
          "response": {
              "type": "string",
              "description": "The AI's response to the user question"
          }
      },
      "required": ["response"]
    }
  maxToolExecutions: 5
  memory: 
    memoryId: "session123"
    memoryType: "message_window"
    maxMessages: 10
  tools:
    - name: SEARCH_DOCUMENTS
      description: "Search for relevant documents based on the user's query."
      parameters: 
        {
          "type": "object",
          "properties": {
              "query": {
                  "type": "string",
                  "description": "The search query"
              }
          },
          "required": ["query"]
        }
      output: 
        {
            "type": "object",
            "properties": {
                "documents": {
                    "type": "array",
                    "items": {
                        "type": "string",
                        "format": "uri"
                    }
                }
            },
            "required": ["documents"]
        }
  agentOutcomes:
    - condition: '${ $agentOutcome.returnValues.response != null }'
      transition: SuccessState
    - condition: '${ $agentOutcome.returnValues.response == null }'
      transition: ErrorState

For more detailed information and advanced configurations, refer to the AI Agent State spec.