Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add explanation of embeddings #500

Merged
merged 4 commits into from
Aug 16, 2024
Merged

Add explanation of embeddings #500

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8d9b768
Add explanation of embeddings
InAnYan Aug 14, 2024
f0ce409
Fix from code review
InAnYan Aug 14, 2024
e8b0c71
Rework text
koppor Aug 15, 2024
3052a57
Fix typo
InAnYan Aug 16, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 9 additions & 1 deletion en/ai/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,15 @@ In this window you can see the following elements:

## How does the AI functionality work?

In the background, JabRef analyses the linked PDF files of library entries. The information used after the indexing is then supplied to the AI, which, to be precise, in our case is a Large Language Model (LLM). The LLM is currently not stored on your computer. Instead, we have many integrations with AI providers (OpenAI, Mistral AI, Hugging Face), so you can choose the one you like the most. These AI providers are available only remotely via the internet. In short: we send chunks of text to AI service and then receive processed responses. In order to use it you need to configure JabRef to use your API key.
JabRef uses external AI providers to do the actual work.
You can choose between OpenAI, Mistral AI, and Hugging Face.
They all run "Large Language Models" (LLMs) to process the requests.
The AI providers need chunks of text to work.
For this, JabRef parses and indexes linked PDF files of entries:
The file is split into parts of fixed-length (so-called *chunks*) and for each of them, an *embedding* is generated.
An embedding itself is a representation of a part of text and in turn a vector that represents the meaning of the text.
Each vector has a crucial property: texts with similar meaning have vectors that are close to (so-called *vector similarity*).
As a result, whenever you ask AI a question, JabRef tries to find relevant pieces of text from the indexed files using vector similarity.

## More information

Expand Down
Loading