qa_agent.md

Alias for Question Answering

Alias-QA is a question-answering agent that integrates RAG (Retrieval-Augmented Generation) and GitHub MCP tools, capable of answering user questions based on both a private knowledge base and GitHub code repositories.

Environment Setup

Before using Alias-QA, you need to set the following environment variables:

export DASHSCOPE_API_KEY=your_dashscope_api_key
export GITHUB_TOKEN=your_github_token

GITHUB_TOKEN: Used to access GitHub MCP tools (search repositories, code, etc.). For how to obtain it, please refer to the GitHub documentation.

Using Custom Knowledge Base

If you want to use your own documents to build a knowledge base, refer to the following steps.

Usage Steps

1. Prepare Your Document File

   Place your document file in the text format supported in the alias/agent/agents/qa_agent_utils directory, or specify a file path.

2. Modify Script Parameters

   Edit the main() function in create_rag_file.py and change faq_file_path to your file path:

   async def main() -> None:
       """Main function for standalone execution."""
       # Read the FAQ samples file
   
       faq_file_path = SCRIPT_DIR / "as_faq_samples.txt"
       collection_name = "as_faq"
       await initialize_rag(
           faq_file_path=faq_file_path,
           collection_name=collection_name,
       )

3. Run the Script

   python alias/agent/agents/qa_agent_utils/create_rag_file.py

   The script will automatically:

   • Start the Qdrant vector database (if not running)
   • Read and process your document file
   • Chunk the documents and generate embedding vectors
   • Store them in the Qdrant database

Important Notes

Default Behavior

If no file path is specified, the script will default to processing the as_faq_samples.txt file and store the processed data at the following path: /alias/agent/agents/qa_agent_utils/qdrant_storage/collections/as_faq

• In this path, as_faq is the value of collection_name
• Running this script repeatedly will continually append processed file content to collections/as_faq

Creating and Using Different Collections

1. Create a New Collection

   You can create a new collection by editing the main() function in create_rag_file.py and modifying the collection_name parameter:

   collection_name = "your_own_collection_name"

2. Switch to a Different Collection

   In /alias/agent/tools/add_qa_tools.py, when referring to the knowledge base through knowledge = SimpleKnowledge(...), you can switch to a different collection by modifying the collection_name parameter:

   collection_name = "collection_name_you_want_to_use"

More Information

create_rag_file.py is a demo implementation of RAG functionality in AgentScope. For more advanced operations and customization options regarding RAG, please refer to the AgentScope RAG Official Documentation, including:

• Multimodal RAG
• Custom Reader, Knowledge, and Store components
• Agentic vs. Generic RAG