Feat: Multi-User Chat Support with Named Messages

[raffareis]

Feature: Multi-User Chat Support with Named Messages

Summary

This PR introduces support for multi-user chat environments by leveraging the name field in BaseMessage from Langchain. This enhancement allows the chatbot to differentiate between multiple users, enabling personalized responses and better context management.

Key Changes

1. Message Naming: Added the name field to chat messages, allowing the bot to identify and personalize responses for different users.
2. Database Schema Update: Modified the chat_message table to include the name column.
3. Memory Management: Enhanced zep memory node to store and retrieve messages with user names, improving context handling in multi-user chats.

Benefits

• Personalized Responses: The bot can now address users by name and provide tailored responses.
• Enhanced Context: Ability to ask the bot to summarize or enumerate opinions based on user names.
• Improved Multi-User Interaction: Supports environments where multiple users interact, such as group chats or channels.

Example Use Cases

• Personalized Greetings: "Hi Henry! What can I help you with today?"
• Contextual Summaries: "Summarize everything that Henry said, please."
• Opinion Enumeration: "Enumerate what everyone's opinion is in this discussion."

Testing

• Use flow:
  Test Name Chatflow.json

• Send prediction with humanPrefix
  

Relevant Links

• Zep Memory Base Class:
  https://github.com/langchain-ai/langchainjs/blob/main/libs/langchain-community/src/memory/zep.ts

• Langchain BaseMessage
  https://github.com/langchain-ai/langchainjs/blob/3ae84b64ed32a09a1d214516ac10af443b724d4d/langchain-core/src/messages/base.ts#L165

• OpenAI Chat Completions Message Schema
  https://platform.openai.com/docs/api-reference/chat/create

Feedback

• This is a prototype with Zep Memory and Tool Agent, for validation. If approved, we may extend this to other memory nodes / agents.

---

Thank you for considering this enhancement. Looking forward to your feedback and suggestions.

Rafael Reis