AI search & RAG on your content
Let visitors and staff ask questions in plain English and get answers based on your own content, documents and policies, with links back to the source pages. It works with a Drupal site, a Symfony application or a custom PHP platform, and answers come only from the content you choose.
engagement: Proof of concept, then production stack: PHP, AWS Bedrock, vector search works with: Drupal, Symfony, custom PHP data: Kept in your own cloud account
Sounds familiar?
- Visitors can't find answers on a large site or in long policy documents
- Staff spend time searching an intranet and shared PDFs
- Your support team answers the same questions every day
- You'd like to use AI, but worry about wrong answers and where your data goes
The work, in detail.
Retrieval-augmented generation (RAG) pipelines
Questions are matched to the most relevant passages of your content, and the model answers from those passages only.
Website content and documents synced to a vector index
Pages, PDFs and documents indexed and kept up to date as editors publish, with unpublished or restricted content left out.
AWS Bedrock and other LLM providers
A model chosen for your needs and budget, and switchable later without rebuilding.
Integration with Drupal, Symfony or custom PHP
A search box, a chat panel or an API, built into the site you already have.
Guardrails, citations and answer logging
Every answer links to its sources, off-topic questions are declined, and questions and answers are logged for review.
Data kept in your own cloud account
The index and the model calls run in your account, so content stays under your control.
From first call to finished work.
Every engagement starts with a free 30-minute discovery call and a written proposal. Nothing starts until the scope, price and terms are agreed in writing.
Use case and content review
We agree who will ask questions, what about, and which content the answers should come from.
Proof of concept
A working version on a sample of your content, tested against real questions, with results written up.
Production build
If the proof of concept works for you, it's built properly: secure, monitored and integrated with your site.
Monitor and improve
Logged questions show where content is missing or answers are weak, so both can improve.
Frequently asked.
Will it make things up?
Any language model can give a wrong answer. RAG reduces the risk by answering from your content and showing its sources, and guardrails tell it to decline questions your content doesn't cover. The proof of concept measures how well it does on your real questions before you commit to a production build.
Is our data used to train the model?
That depends on the provider's terms, so we check them together before anything goes live. The setup runs in your own cloud account and region, and you choose which content is indexed.
What about personal data and UK GDPR?
If the content or the questions include personal data, you may need a data protection impact assessment before launch, and we sign a data processing agreement. I can provide the technical description your data protection lead needs.
Do we need a Drupal site?
No. It works with Drupal, Symfony or a custom PHP platform, and documents can be indexed from shared storage as well as from a website.
Tell me what your site needs.
The first conversation is free and comes with no obligation. Tell me about your site and I'll come back with a written proposal.