AI is transforming our world. It’s being used in lots of headline catching ways but how would it change the little things, like doing some local history or finding out when your next Parish Council meeting is taking place. It’s this kind of demo that could also transform a business.

Want to just find out how a project has progressed from just reading the minutes of the meeting but you don’t want to trawl through all of them?
Get an AI to summarise the projects history for you.

Want to know how to disassemble and diagnose some faulty machinery but you are not sure where to look in the technical manuals?
An AI can handle that easily and give you step by step instructions augmented by its own knowledge.

For this demonstration, we partnered with Friends of Droxford Church who hold a number historical documents on their website that we could feed in to our database, Droxford also has a good online presence for local government and other resources that will add additional utility to villagers.

Adding these documents in to a database for retrieval by an AI or LLM (Large Language Model), is called either RAG (Retrieval Augmented Generation) or CAG (Cache Augmented Generation).

In RAG, you break your documents down in to small typically 1000 character chunks and use an AI to store them in a database. Using an AI to store them allows for much greater recall of information over normal database recalls, it’s cost-effective as you only get your information in 1000 character chunks, making AI use cheaper (they charge per token which is roughly equal to a character). With a users query an AI LLM can then attempt to build the correct query to return the relevant documents.

In CAG, we pass all of the documents to an AI LLM, along with the the users query. This was not possible until recently when the context window (how many characters an AI LLM can handle before getting confused and shutting down) was large enough to handle lots of text. LLMs like OpenAI’s gpt-4.1-mini and Google’s gemini-2.0-flash can now get up to ~1,000,000 characters or ~2 standard novels (~100,000 words each).