Skip to content
Nappai Documentation

Chat Memory

This component lets you access and use past chat messages within Nappai. It’s like having a memory for your automated tasks, allowing you to refer to previous conversations to improve the context and accuracy of your automated processes. You can filter these messages to find exactly what you need.

Relationship with Langchain

This component works with Langchain, a powerful framework for building applications with large language models. It can retrieve messages from Langflow’s internal database or from an external memory system, depending on your configuration. This allows for flexible integration with various data sources.

Inputs

  • External Memory: Choose whether to retrieve messages from an external memory source (you’ll need to configure this separately) or use Nappai’s internal storage. Leaving this blank uses Nappai’s internal storage.
  • Sender Type: Filter messages by who sent them: “Machine and User” (default, shows messages from both), “Machine” (only AI messages), or “User” (only your messages).
  • Sender Name: (Advanced) Filter messages by the specific name of the sender.
  • Number of Messages: (Advanced) Specify how many past messages to retrieve. The default is 100.
  • Session ID: (Advanced) Filter messages by a specific session ID (useful for managing multiple conversations).
  • Conversation ID: (Advanced) Filter messages by a specific conversation ID (useful for managing multiple conversations).
  • Order: (Advanced) Choose whether to retrieve messages in ascending (oldest first) or descending (newest first) order. The default is ascending.
  • Template: (Advanced) Customize how the messages are formatted. You can use placeholders like {sender_name} and {text} to control the output. The default is “{sender_name}: {text}”.

Outputs

  • Messages (Data): This output provides the retrieved messages in a structured data format. This is useful for further processing within Nappai.
  • Messages (Text): This output provides the retrieved messages formatted as plain text, according to the template you specified. This is ideal for displaying the chat history in your dashboard or using it as input for other components.
  • Memory: This output creates a Langchain chat memory object. This is used internally by Nappai to manage the chat history and is typically used by other components in your workflow.

Usage Example

Imagine you have a chatbot that needs to remember previous interactions with a user. You can use the Chat Memory component to retrieve the user’s past messages, filter them by sender type, and then use the formatted text output to provide a more context-aware response.

Templates

This component is used in the “Memory Chatbot” template. In this template, it is configured to retrieve a certain number of recent messages and format them using a default template. You can adjust the settings within the template to customize its behavior.

  • Store Message: This component saves chat messages to be used by the Chat Memory component.
  • ConversationChain: This component uses the Chat Memory to maintain context during a conversation.
  • Retrieval QA: Chain for question-answering querying sources from a retriever: This component uses the chat history to answer questions more accurately.

Tips and Best Practices

  • Start with the default settings and only adjust the advanced options if you need more specific filtering or formatting.
  • Use the “Template” input to create clear and readable output for your chat history.
  • If you are using an external memory, ensure it is properly configured before using this component.

Security Considerations

Ensure that your external memory (if used) is secured appropriately to protect sensitive data. Review Nappai’s overall security settings to ensure the confidentiality and integrity of your chat data.