Welcome to our tutorial on creating a retrieval chat using Volt Vector. This guide will walk you through the process of setting up a retrieval chat, which is a core component of the Volt Vector system. The retrieval chat is designed to work seamlessly with prompts, embeddings, namespaces, and profiles, enabling efficient data retrieval and interaction with a Language Model (LLM). Let’s dive in!

What is a Retrieval Chat? #

A retrieval chat is a feature in Volt Vector that allows for efficient data retrieval and interaction with a Language Model (LLM). It utilizes prompts, embeddings, and namespaces to organize and retrieve data.

Step 1: Creating a New Retrieval Chat #

1. Start by Adding a New Chat: Navigate to the ‘all chats’ section and select ‘add new’. Name this new chat ‘retrieval’.
2. Namespace: It’s essential to assign a namespace, as the chat won’t function without it. Embeddings always require a namespace, so ensure you specify one here.

Step 2: Setting Up Context and Language #

1. Context: Select ‘retrieval default system prompt’ for your context.
2. Language Options: The default language setting is English, but you can choose other languages like Spanish. For instance, if you ask a question in Spanish, the chat will respond in Spanish.

Step 3: Customizing Chat Features #

1. Chat Tone and Welcome Message: You can modify these according to your requirements. The welcome message is a greeting for site visitors but isn’t sent to the LLM.
2. Send Button and Text Area: Customize the send button text and the placeholder text in the message input area.

Step 4: Profiles and LLM Connection #

1. Selecting Profiles: Choose ‘vault simple, default chunk’ for profiles. This means the chat will find embeddings in the default namespace under the simple MySQL.
2. LLM Profile: Use the base profile to connect to OpenAI.

Step 5: Managing Cited Sources and Restrictions #

1. Cited Sources Limit: Set a limit for the number of cited sources – adjust this to manage API costs.
2. Cosine Similarity Threshold: Set a threshold (e.g., 0.8) for including cited sources based on their relevance.
3. No Results Message: Customize the message for when no relevant cited sources are found.

Step 6: Chat Customization and Deployment #

1. Message and Style Settings: Set a message limit and choose between light or dark themes.
2. Icons and Naming: Customize icons and name your chat (e.g., User Chat or Assistant Chat).
3. Saving and Implementing: Save your settings and use a shortcode block in the block editor to implement the chat on your site.

Step 7: Testing Your Chat #

1. Sample Queries: Test the chat with queries like “How do you set up WordPress debug?” in different languages.
2. Checking Responses and Sources: Ensure the chat provides correct responses and cited sources.

Conclusion #

Congratulations! You’ve successfully set up a retrieval chat in Volt Vector. This chat will be an invaluable tool for providing information and support on your site, enhancing user engagement and efficiency.

Additional Tips:

• Regularly test your chat with different queries to ensure it’s functioning correctly.
• Monitor API usage to manage costs effectively.
• Continuously update and refine your chat settings based on user feedback and usage patterns.

Thanks for following this tutorial. We hope your new retrieval chat enhances your Volt Vector experience!