File Search

New chat interface

Screenshots and labels on this page show the new chat interface. It is currently available in the demo access and in pilot environments; in other environments, labels and steps may still differ.

File Search allows you to search the contents of files similar to the query and enrich the responses with internal, specific knowledge. File search is only available per user.

In the chat, you activate file search via the plus icon in the chat input (Attach and tools) with File Search under Tools. Attached files are then available to it when needed; you do not have to choose a separate upload destination. In an agent, you add files permanently under File Search; this allows the agent to search larger document collections as well.

What is file search suitable for?

It is best suited for longer text content that does not need to be processed in its entirety, but rather where individual sections relevant to the question should be searched for and included in the answers.

Further information here: File processing

The files are not added to the context in their entirety; instead, only relevant content is returned through a similarity search. This functionality is called RAG - Retrieval Augmented Generation and expands the knowledge of AI and AI agents in CompanyGPT with domain-specific, internal, or even confidential documents. The content is converted into embeddings (vectors) using an AI model. When searching, the search query is also converted into vectors and the proximity is calculated mathematically. This allows content proximity to be mapped much better than, for example, using a full-text search.