Specialised knowledge
When the AI model needs to know information it was not trained on.
What is RAG? Retrieval-augmented generation, simply explained.
Retrieval-Augmented Generation (RAG) personalises AI by connecting models like ChatGPT to your own data, so they give accurate, context-aware answers based on your information.
What makes RAG powerful?
RAG extends an LLM’s knowledge beyond what was available during its pretraining by bringing in data that doesn’t exist on the public internet. While an LLM’s knowledge is effectively frozen at its training cutoff, RAG lets it securely access your personal or proprietary data at the moment you ask.
RAG turns your data into tokens and vector embeddings, the mathematical representation AI models understand. That lets the system find the information most relevant to your question, even when the context is spread across large volumes of documents.
The more specific and well-structured your question, the more effectively the system can map it to the nearest vectors that hold the most accurate answer.
How RAG works.
Document processing
Your documents are broken into chunks and processed in a way that preserves their original structure.
Vector embeddings
Each chunk is converted into a numerical address (a vector) derived from the relationships between the words and phrases in it.
Retrieval
When you ask a question, AI uses the proximity between vectors in high-dimensional space to find the right answer. This is called semantic search.
Generation
The retrieved information is injected into the LLM’s context window, so it generates an answer grounded in that information.
When do you need RAG?
Fresh knowledge
When your information changes often and the AI must work from current data.
Large volume of data
When the answer has to be found across more data than fits in any LLM’s context window.
Diverse file formats
When you use formats LLMs don’t natively support: Google files, MS Office, Kindle, large video and audio files.
Knowledge sharing
When you want to share specific domain knowledge as an AI chat that others can ask questions of.
Source attribution
When you need the AI to answer from your information and cite the document it used.
Why AskBnB’s RAG platform.
Advanced RAG techniques
Pre- and post-processing that significantly improves RAG performance: context continuity across chunks, original structure preserved, and metadata added for smart filtering.
Share collections
Share a collection to a URL fully hosted by AskBnB, where anyone can ask your knowledge questions in a chat.
Broad file format support
Google Docs, Sheets and Slides, MS Office, Kindle, video and audio, complex PDFs, multi-tab XLSX, and even web URLs and social posts.
Google Drive and OneDrive
Connect files to sync changes to your collection automatically, so the AI always has the latest information.
Data security
SOC 2 Type II, HIPAA and GDPR compliant, with end-to-end encryption. Your data is never sold or used to train other models.
Zero hallucinations
Guardrails make the AI answer only from the data in your AI library. Links to the source file it used are provided.
Ready to build your AI library?
Turn the information you care about into a knowledge base you can ask. 14 days free.