Project Overview

Kakapo is an AI powered chat-bot trained on your custom data, built to serve as a 24/7 customer care assistant for website visitors.

Challenges

Deliver a full-stack chatbot builder that could:

• Support multiple bots per user with secure auth and data isolation.
• Ingest URLs and files, then index them for reliable RAG retrieval.
• Provide embeddable experiences (share pages, iframe, floating chat bubble).
• Add observability, caching, and conversation logs without slowing responses.

Our Solution

We built a scalable AI chatbot platform with:

• Core Product: Bot management, URL/file ingestion, and Supabase vector embeddings.
• RAG Pipeline: LangChain + OpenAI with Upstash Redis caching and Langfuse tracing.
• Embeddable Delivery: Public share pages, iframe embeds, and a floating chat bubble.
• Customization: Branding controls, welcome messages, and suggested prompts.

Results

• Production-Ready Platform: Teams can launch multiple bots, train them on custom data, and embed them to their website.
• Faster, Smarter Responses: RAG plus caching improves accuracy and response time.
• Operational Visibility: Conversation logs and tracing make performance measurable.