How Kapa takes the pain out of finding accurate technical answers

Summary

Using Weaviate's vector database, Kapa created an innovative AI platform that converts complex technical documentation into responsive support chatbots, enabling companies of all sizes to streamline their technical support operations. Kapa powers the “Ask AI” widget on Weaviate’s documentation.

The Challenge

Technical documentation is vast and complex, making it challenging for users to quickly find the specific information they need. Traditional search methods often fall short when dealing with technical queries, requiring users to spend considerable time piecing together answers from multiple sources. Kapa focused on building an intelligent platform that could effectively process and respond to technical questions while maintaining accuracy and reliability.

This required a robust solution that could:

• Handle large volumes of documentation efficiently
• Process queries quickly and accurately
• Scale across distributed systems
• Provide reliable performance under heavy memory consumption

Why Weaviate?

After evaluating various vector database solutions, Kapa chose Weaviate for several key reasons:

• Docker Compatibility: Weaviate's ability to run in Docker containers was crucial for their deployment and local development needs, unlike alternatives such as Pinecone.
• Built-in Hybrid Search: Weaviate was one of the first vector databases to offer hybrid search out-of-the-box.
• Scalability: The distributed nature of Weaviate’s multi-tenancy feature made it ideal for scaling across nodes, which was essential for Kapa as they were experiencing rapid growth with users and data quantities.

What's Next?

Kapa is actively exploring new opportunities with Weaviate's latest features, including a focus on multi-vector support. This will allow them to have backups of embeddings enhancing reliability with a cost-effective layer of redundancy. The team is continuously working on improving accuracy benchmarking and measurement systems to ensure optimal performance for their customers. Running on Google Cloud and written in Python, Kapa continues to evolve their tech stack while maintaining Weaviate as their core vector database and embeddings layer solution.