Skip to content

Latest commit

 

History

History
59 lines (39 loc) · 2.24 KB

assemble_module.md

File metadata and controls

59 lines (39 loc) · 2.24 KB
Raw

🧩 Assemble Module

The Assemble Module manages chat history and context assembly for the AI assistant.

🌟 Key Features

  • 📚 Maintains chat history
  • 🔗 Assembles context for AI prompts
  • 🧠 Manages memory chunks
  • 🔀 Flexible prompt assembly options

🔧 Main Functions

  • 🔄 assemble_prompt_with_history(current_prompt: str, chat_history_only: bool = False) -> str: Combines chat history (and optionally chunk history) with the current prompt.

    • current_prompt: The current user input or query.
    • chat_history_only: If True, only includes chat history in the assembled prompt. If False (default), includes both chat history and chunk history.

  • ➕ add_to_chat_history(prompt: str, response: str): Adds new interactions to the chat history.

  • 📋 get_chat_history_tuples() -> List[Tuple[str, str]]: Retrieves chat history as a list of (prompt, response) tuples.

  • ✂️ truncate_chat_history(n: int): Truncates the chat history to the last n entries.

🚀 Usage

This module is used to provide context to the AI model, ensuring coherent and contextually relevant responses. The chat_history_only option allows for flexible prompt assembly based on specific use cases.

Example usage:

from src.modules.assemble import assemble_prompt_with_history

# Assemble prompt with both chat and chunk history
full_prompt = assemble_prompt_with_history("What's the weather like today?")

# Assemble prompt with only chat history
chat_only_prompt = assemble_prompt_with_history("What's the weather like today?", chat_history_only=True)

🧪 Testing

While there's no dedicated test file for this module, its functionality is indirectly tested through the test_save_history.py file, which covers:

  • Adding entries to chat history
  • Saving and loading chat history
  • Managing memory chunks

To run related tests:

python -m unittest src/tests/test_save_history.py

🔍 Key Considerations

  • Optimize the history assembly process for performance with large histories
  • Implement a strategy for managing very long conversations (e.g., summarization)
  • Ensure proper synchronization if implementing concurrent access to chat history
  • Consider the impact of including or excluding chunk history on the AI's responses