Chatbot Icon

August 31, 2023

AI chatbots made easy

Written By

Imogen Eales


AI/ Machine Learning

Read time

2 minute

AI chatbots have been revolutionising customer interactions, but despite their advancements, they often struggle to incorporate specific domain knowledge. Recognising this gap, we at Contic have been focusing on a cutting-edge project to showcase a Node.js, Prisma, and Postgres example demonstrating how to augment OpenAI's GPT models with your unique knowledge base.
Models like ChatGPT are adept at generating human-like responses, yet they falter when it comes to integrating domain-specific or company-specific knowledge. Existing solutions, such as retraining the model or directly embedding the knowledge into the ChatGPT prompt, are either expensive, time-consuming, or hampered by token limitations.
Our approach is a bit off the beaten track. Instead of overburdening the model, we use embedding similarity search to associate ChatGPT with a custom knowledge base. We convert text into embeddings, vector representations that facilitate the comparison and identification of similar pieces of text. This enables us to dynamically extract the most relevant knowledge to answer a user's query.
The beauty of our solution? It's not confined to FAQs or customer support scenarios. It's adaptable and can be leveraged for various applications - from Notion workspaces, web scraping, and replacing traditional intranets to streamlining business operations.
By integrating custom knowledge bases, we're expanding the potential of AI chatbots. These chatbots can now provide responses that are not just intelligent but deeply contextual and tailored to your business.
Our approach lets businesses deliver outstanding customer experiences without the recurring costs or time typically associated with continuous AI model retraining. Keeping chatbots updated with the latest company-specific knowledge is as simple as updating the custom knowledge base.
Let's shape the future together. We're at the forefront of AI chatbot innovation and invite you to join us. If you would like to introduce this into your product, we’d love to speak to you about it. Get in touch!


How did you find your chatbot experience today?

Thank you for your feedback

There has been an unexpected error, please try again later