govchat

what is govchat

GovChat is a retrieval augmented generation (RAG) app that uses GenAI to chat with and summarize civil service documents. It's designed to handle a variety of administrative sources, such as letters, briefings, minutes, and speech transcripts.

• Better retrieval. GovChat increases organizational memory by indexing documents

• Faster, accurate summarisation. GovChat can summarize reports read months ago, supplement them with current work, and produce a first draft that lets government employees focus on what they do best.

govchat features

• Robust Retrieval Augment Generation (RAG) based chat for data driven conversations with granular citations.

• Guided summarisation where the human in the loop can review and guide summaries of multiple documents.

• Advanced prompting techniques to ensure safety and balance in AI responses and outputs.

• Proactive data privacy and retention features to maintain compliance and manage data.

• Securely designed for on premises deployment and across clouds.

govchat Architecture

GovChat is intended to be as modular and multi-use as possible. Our initial design is build for AWS simply due to the cloud available to us. However, we are designing the system to be able to be used on any cloud provider or on-premises.

Some of our guiding architectural principles are:

• Modularity: We want to be able to swap out components easily. This means that we want to be able to swap out the database, the document processing pipeline, the large language model, etc.

• Scalability: We want to be able to scale the system horizontally. This means that we want to be able to add more instances of the system to handle more load.

• Resilience: We want the system to be able to recover from failures. This means that we want to be able to recover from a failure of a single instance of the system.

• Security: We want the system to be secure. This means that we want to be able to secure the system from unauthorized access.

• Simplicity: In taking a simple approach to data ownership we want to enable other deployers to introduce their own business logic separately from the core system.

govchat Document Processing Pipeline

The document processing pipeline is the core of the GovChat system. It is responsible for processing documents and extracting information from them. The pipeline is broken down into two main components: Worker and Core API.

govchat RAG Architecture

The Retrieval Augmented Generation (RAG) architecture grounds our Large Language Model in the context of uploaded documents.

chat interface