Ministerial Correspondence
Overview
The MCU AI project is a Proof of Concept (PoC) LLM system that showcases how large language models could be used to enhance the handling of ministerial communications.
It aims to:
- Triage documents to extract key facts and generate summary
- Automate initial drafts of responses
- Reduce administrative workload for correspondence drafters
Features
- Triages and summarises received correpondence using an LLM
- Retrieval Augmented Generation (RAG) to generate responses based on policy documents
- Built to accomodate a range of LLMs from various providers:
- Azure OpenAI
- Amazon Bedrock
- Ollama (for local deployment)
- Uses custom personas to tailor response tone and language
- Supports full audit capabilities via response history page and IG Log
Gallery
-
LLM Selection
-
Persona Selection
-
Knowledge Selection
-
Sources
Architecture
-
MCU Architecture
