Replies: 6 comments 1 reply
-
|
I am thinking about same problem Might be implemented through embeddings (basically to search among documents for relevant parts ("indexing")) and feed chunk of this data as input for model, so we can have RAG I have planned to test how just embeddings work with local-ai and langchain, so I can search through doc or database, to find what I want using context similarity search For example, it might be useful through parsing user's data or to working with historical data Let's say you have a db with users-locations, which can be any location in any format like "Rim, Italy -- Rome, Italy -- Berlin -- Bay Area -- НижнийНовгород -- Украина Киев -- Украiна Кiев -- worlidwide -- Moon(planet earh)" and so on Working with historical data means, that you can just add dump of all local-news for a few years and make your model aware of context nowdays. Or it could be chatHistory with chat-sessions, or it could be CVE list dump, or even your own codebase but that just embeddings |
Beta Was this translation helpful? Give feedback.
-
|
But for a full RAG we also need some kind of |
Beta Was this translation helpful? Give feedback.
-
|
I found some kind of solution here: |
Beta Was this translation helpful? Give feedback.
-
|
Seconding this since a lot of agentic libraries have RAGs built in with reranking and embedding functions. |
Beta Was this translation helpful? Give feedback.
-
|
why you thinking of it ? |
Beta Was this translation helpful? Give feedback.
-
|
Could you do a whole list on what is available? Not sure if LMStudio does it but it ain't FOSS |
Beta Was this translation helpful? Give feedback.
-
|
lmstudio work at the moment only via an embedding model and send it via server to eg privateGPT i know only jan privategpt maybe it give 2more, but if, it was to complicate to install |
Beta Was this translation helpful? Give feedback.
-
any local docs/pdf i can talk with ?
some library all indexed read for use ;)
Beta Was this translation helpful? Give feedback.
All reactions