Hexagon Digital

Case Studies

An Internal Knowledge Assistant Over Google Drive and Dropbox

A private chat assistant that answers staff questions using only the company’s own documents, synced from Google Drive and Dropbox.

Case studyKnowledge Base

RAG Brain

Use
Internal company knowledge
Industry
Knowledge Base
Services
Custom AI Agents
Built with
React Router, PostgreSQL with pgvector, OpenAI, Claude

  • 2

    Cloud drives synced: Google Drive, Dropbox

  • 6

    Document formats read and indexed

  • 2

    AI providers supported: OpenAI and Claude

problem
Company knowledge sat in documents across Google Drive and Dropbox, and finding an answer meant knowing which file to open.
approach
We built a RAG chat system that syncs both drives, indexes the documents and answers questions from them alone.
outcome
Staff ask in plain language and get an answer that names its source document - or a clear “not in the documents”.

The situation

Policies, proposals, process documents and presentations were spread across Google Drive and Dropbox. People asked colleagues, or searched folder by folder.

A general chatbot was not an option: it does not know the company’s documents, and it will answer confidently anyway.

What we built

A retrieval-augmented (RAG) chat system for internal use.

  • Sources: an admin connects Google Drive and Dropbox, picks the folders to include and runs a sync.
  • Documents: PDF, Word, PowerPoint, CSV, text and Markdown files are parsed, split into sections and indexed.
  • Retrieval: each question is matched against the indexed sections, and the most relevant ones are passed to the model.
  • Answers: the model is instructed to answer only from that context and to name the documents it used.
  • Admin: a dashboard for sources, documents, users, chat history and settings, including which AI model to use.

No guessing

When the documents do not contain an answer, the assistant says the information is not available instead of filling the gap from general knowledge.

Access is limited to signed-in company users, with separate admin and sub-admin roles.

The result

Staff ask a question in plain language and get an answer with the source document named, so it can be checked. The model provider can be switched from the settings screen without changing the system.

More case studies

Connect with an expert

Let’s talk about your project

Tell us what you want to build or automate, and we’ll show where AI, web and marketing can make the biggest difference.

Book a discovery call