AI you can rely on
Your Source for Accurate Answers
PINOCCHAT Agents History
Celebrating European Brain Day
We are proud to announce that after more than one year of continuous improvements we released to public first AI Counsellor for SVÉ HLAVY advising patients and caregivers about treatment of brain.
CRAG (Polymorphic Self Checking Agent)
We gave a lot of effort in attempt to make Generative AI answers reliable. Focusing Agent on given task and adding self checks for accessed knowledge increased CRAG’s accuracy for more than 40% in general field, and over 60% to 85% in narrow knowledge field.
Custom Observability Dashboards
Extensive logging of interactions with AI Agent allows to understand Knowledge Base usage and refine Data Sources, increasing answers reliability.
Structured Data Improvements
To increase answers accuracy in specific fields of knowledge We have implemented Tags Extraction to SQL, leading in overall improvements in ability of Generative AI to navigate through data in given field.
SQL and API agents
As time passed PINOCCHAT team saw need in integrating API and SQL and using them as knowledgebases. Then there were created and thoroughly tested SQL and API agents including corresponding integrations.
Distribution Channels
Reviewed and improved UI/UX and Backend for MS Teams and Web Site (Iframe + Pop-up).
Added gRPC for high performance deployments.
MRKL (Modular Reasoning, Knowledge Language)
This agent was inspired by Jurassic-X whitepaper which moved interaction with AI models on the next level. It gave agents ability to use multiple Knowledge Bases and Tools. Still it wasn’t able to perform complex tasks on big data, causing non reliable answers and hallucinations.
RAG/QA (deprecated since March 2024)
Basic Q&A Agent which was first step in PINOCCHAT development. To understand if modern Generative AI is able to perform semantic search tasks.
First ChatBot Pinocchio
First ChatBot, inspired by idea of helping people with stroke or brain injuries in cooperation with SVÉ HLAVY


