Updated January 28, 2025
Fileread uses a combination of large language models and a technique called retrieval augmented generation (RAG) to create an all-purpose conversational AI. By using RAG to verify information accuracy and to direct the GenAI Model, Fileread is able to focus the model’s attention on a specific document universe and interpret meaningful insights.
<aside> ✅
Duplicate documents negatively impact result quality. We strongly recommend removing duplicates and cleaning up your data. For optimal results, include only the most comprehensive documents.
</aside>
Before you begin, you will need to index the documents you’d like Fileread to have access to. Please note that these documents count towards your active document allocation.
To index, go to the Documents view → Select your documents (folder / saved search / click to select) → Mass Actions → Fileread Indexer
You should then see a modal where you enter the password given to you and click continue. Click Start Indexing to initiate the indexing of your documents.
<aside> ⚠️
Currently only PDFs are compatible for indexing. Filetypes that are not PDFs are not support for this version. We are rapidly adding support for more datatypes.
</aside>
Fileread’s AI can analyze documents with diagrams, images, or presentations that don’t generate extracted text for analysis. This leads to more comprehensive understanding of the data and richer insights. Multi-modal is best for data that is presentations, images, and diagrams that do not have extracted text.
To use multi-modal, select Run multimodal analysis. Sets of up to 5000 documents can be uploaded at a time. We recommend creating a saved search with the documents you’d like to run this analysis on.
You can monitor your progress in the Settings Page of the Fileread App. Simply navigate to the Fileread App and click the ⚙️ icon on the top left.