parlant-qna 0.2.3

A questions and answers tool service for Parlant

Parlant Q&A: A Questions/Answers Tool Service for Parlant

Website | Introduction | Installation | Tutorial | About

A dynamic question-answering tool service that helps your Parlant agents provide accurate, traceable responses based on your FAQ data.

What is this?

This tool service enables your Parlant agents to answer questions based on a managed set of questions and answers that you provide.

✨ Key Features ✨

The service carefully evaluates each question against your FAQ, providing:

• 📝 Detailed explanation of how the question was interpreted in light of the FAQ
• 📚 References to specific source material used (question IDs and relevant quotes)
• 🔍 Complete traceability for each response through session events in Parlant sessions (via tool events)
• 🛡️ Protection against hallucination - if the information isn't in your knowledge base, the agent won't make up an answer

🚀 Getting Started

Install parlant-qna:

$ pip install parlant-qna

You can run the service in two ways:

# As a standalone server
parlant-qna serve

# Or as a hosted module within your Parlant server
parlant-server --module parlant_qna.module

📚 Adding FAQs

Add question/answer pairs dynamically through the CLI:

$ parlant-qna add \
    -q "What are your business hours?" \
    -a "We're open Monday through Friday, 9 AM to 5 PM Eastern Time."

⚡⚙️ How It Works

When your agent needs to answer a question, it calls the find_answer tool through the qna service. The service then:

1. Evaluates the question for proper context and intent
2. Searches your FAQs for relevant information
3. Constructs an answer using only verified information
4. Records the entire process as a tool event in the session
5. Returns both the answer and detailed metadata about sources used

Example response:

{
  "answer": "We're open Monday through Friday, 9 AM to 5 PM Eastern Time.",
  "evaluation": "Question seeks information about operational hours, which is provided in the background information",
  "references": [
    {
      "question_id": "dja8-108fj",
      "quotes": ["Monday through Friday, 9 AM to 5 PM Eastern Time"]
    }
  ]
}

💡 Use Cases

Perfect for:

• Building customer service agents that need accurate, verifiable answers
• Creating internal support bots that can explain their reasoning
• Developing agents that need to provide source references for their responses
• Ensuring compliance by preventing agents from making up information

🤝🛠️ Contributing

We're actively developing this tool service. If you'd like to contribute, please:

1. Check our issue tracker for current needs
2. Join our Discord community for discussion
3. Submit pull requests with improvements