Skip to content

README.md

Latest commit

 

History

History
41 lines (26 loc) · 1.68 KB

README.md

File metadata and controls

41 lines (26 loc) · 1.68 KB

Prompts Directory

This directory contains all prompt templates used by the RAG Customer Support System.

Files

agent_system_prompt.txt

The main system prompt for the AI agent that processes customer emails. This defines the agent's behavior, tone, and constraints.

Usage: Loaded by AgentService to configure the LLM's behavior.

retriever_tool_description.txt

Description for the retriever tool that the agent uses to search the knowledge base.

Usage: Loaded by AgentService when creating the retriever tool.

email_classification_prompt.txt

Prompt used to classify incoming emails and determine if they should be processed by the agent or ignored.

Usage: Loaded by AgentService to filter emails before processing. The agent will only respond to emails classified as customer support inquiries.

comprehensive_prompt.md

A single, all-encompassing prompt to rebuild this entire application from scratch.

Usage: Not directly loaded by AgentService. This prompt is for development and understanding the full system context.

Experimenting with Prompts

You can freely edit these files to experiment with different prompts without modifying the code. Changes will take effect the next time the agent is started.

Tips for Prompt Engineering:

  • Be specific about the agent's role and constraints
  • Define the expected tone (professional, friendly, etc.)
  • Specify what to do when information is not available
  • Include examples if needed for complex behaviors

Configuration

The paths to these prompt files are configured in app/config/__init__.py and can be overridden via environment variables:

  • AGENT_SYSTEM_PROMPT_FILE
  • RETRIEVER_TOOL_DESC_FILE