Introduction
Knowledge databases help you integrate specific information into Alan, so that this information is taken into account when generating answers. For example, you can securely make internal company knowledge accessible to Alan and thereby tailor Alan to your needs.
Knowledge databases can be shared with your colleagues, so that a single configured knowledge database can benefit many people in your organization.
In this section, we will introduce you to using and managing knowledge databases in Alan.
Advantages and Limitations of Knowledge Databases
Knowledge databases are especially useful if you:
- Need regular access to the same set of resources.
- Want to manage and search large sets of documents effectively.
- Would like to build a reusable shared basis for teams or projects.
However, there are also limitations you should be aware of:
Prior Processing Before Alan can use the information from a knowledge database, the documents must be indexed and divided into smaller text sections called chunks. This is an additional step that can require more time if you are dealing with a large number of very large documents.
Not all content at once When answering a question, Alan does not load all data from a knowledge database into his context at once, but only those chunks that are relevant to answering the question according to a semantic search. This is generally beneficial because it allows for much larger document libraries than the language model could otherwise process at once. However, it also implies that at no point does the model have access to the entire content of the knowledge database, so you cannot simply request a full summary of the entire knowledge database or of a very large document.
Setup and Maintenance Knowledge databases have to be set up, structured, and updated if needed. If you only need a document for a brief analysis or a one-time project, the required effort may not pay off.
Limited by Search Results If some or all of the relevant chunks are not identified by the semantic search, they are not available to Alan in the context. This can occasionally lead to incomplete answers if certain information is not classified as “relevant.”
If you run into these limitations, or if you only want to upload and quickly analyze or compare one or more documents on a short-term basis, Chatting with Files might be the better option. It immediately attaches the documents to your chat, without requiring you to create a separate knowledge database. This does not mean one approach is “better” than the other - they are simply different and tailored to different use cases.
You can find a direct comparison between knowledge databases and Chatting with files in the Chatting with Files documentation.
Video Tutorials (in German)
Using specific information with knowledge databases
In this video, we will show you how to use Alan with your own data sources.
Creating knowledge databases
In this video, we will demonstrate how to create and share knowledge databases in Alan.
