very nice -- that translation from LCEL to LangGraph makes it really clear what LCEL is doing. I'm never quite sure what data is passed over that LCEL pipe character. With the langgraph example, it's obvious what is passed, it's right there in the node.
Do you use FAISS for production as your vector store or what do you use and based on how do you use it? One of the things I'm working on right now which I am sort of consfused on how to approach it is role based. 1) split each user based on email 2) Each user will have a role 3) Each role will have list of documents to do RAG on 4) Specific roles will be able to delete or add documents from specific roles such as CEO can remove a doc for the Intern Role 5) Specific roles can add/remove roles but the documents do not get deleted. Just that the role doesn't exist anymore
Great video as always, thanks a lot! Could you please upload the notebook? I took a look at the repository and I think this latest notebook is not there yet...
@@codingcrashcourses8533 Thanks for the tips. Can you please elaborate what kind of knowledge are you referring to? Is it the LangChain Ecosystem knowledge or the general LLM App knowledge? I got a decent understanding of general LLM App knowledge like RAG, vector Store, Flow engineering etc, but dont have much LangChain Ecosystem knowledge. Do I still need to learn LangChain first?
@@alsonyang230 No, you can use any other library or framework too. There is no "have to", but these frameworks bring their own ideas which you might like or dislike. I can´t say "do this, 100%"
This found me at just the right time. Day 3 of sitting down with langchain and absolutely hate the inconsistency, verbosity and lack of standardized documentation. Above all the breaking changes, every other release. LangGraph is shit as well. Haystack seems way better though and will probably catch up with langchain in features soon.
Same thoughts on langchain now. When I was a beginner, I used Langchain because of strick timeliness of the deliverables and lc made it quite easy to do things. Now I just want to get rid of langchain and implement using the LLM api available.
@@codingcrashcourses8533just working through your RAG Udemy course, my main ambition is more complicated production environments, what do you recommend in place of Langchain/Langraph?
very nice -- that translation from LCEL to LangGraph makes it really clear what LCEL is doing. I'm never quite sure what data is passed over that LCEL pipe character. With the langgraph example, it's obvious what is passed, it's right there in the node.
The testing aspect is huge! Thanks for pointing this out, really useful :-)
I don't see this notebook committed to the repo yet. just going to Rewind and FastForward a lot for now.
sorry it seems i lost it somewhere in the process. I was not able to find it somewhere :(
Do you use FAISS for production as your vector store or what do you use and based on how do you use it? One of the things I'm working on right now which I am sort of consfused on how to approach it is role based.
1) split each user based on email
2) Each user will have a role
3) Each role will have list of documents to do RAG on
4) Specific roles will be able to delete or add documents from specific roles such as CEO can remove a doc for the Intern Role
5) Specific roles can add/remove roles but the documents do not get deleted. Just that the role doesn't exist anymore
@@yazanrisheh5127 I use pgvector
Great video as always, thanks a lot! Could you please upload the notebook? I took a look at the repository and I think this latest notebook is not there yet...
shit it got lost somewhere in the process. Gonna check that, sorry
The gold question is, do I need to learn LangChain first before learning LangGraph tho? Will I miss anything by directly jumping onto LangGraph?
You probably should learn LangChain first. Without that knowledge, you can not build LLM based apps.
@@codingcrashcourses8533 Thanks for the tips. Can you please elaborate what kind of knowledge are you referring to? Is it the LangChain Ecosystem knowledge or the general LLM App knowledge? I got a decent understanding of general LLM App knowledge like RAG, vector Store, Flow engineering etc, but dont have much LangChain Ecosystem knowledge. Do I still need to learn LangChain first?
@@alsonyang230 No, you can use any other library or framework too. There is no "have to", but these frameworks bring their own ideas which you might like or dislike. I can´t say "do this, 100%"
How about Swarm AI in comparison?
I already made a video about that last week
This found me at just the right time. Day 3 of sitting down with langchain and absolutely hate the inconsistency, verbosity and lack of standardized documentation. Above all the breaking changes, every other release. LangGraph is shit as well. Haystack seems way better though and will probably catch up with langchain in features soon.
Why u think like this about langgraph?
Same thoughts on langchain now. When I was a beginner, I used Langchain because of strick timeliness of the deliverables and lc made it quite easy to do things. Now I just want to get rid of langchain and implement using the LLM api available.
Yes, like I said some of the abstraction is awesome. Retrievers, templates, Chat classes - all great. But LCEL for complex workflows. Hello no!
@@codingcrashcourses8533just working through your RAG Udemy course, my main ambition is more complicated production environments, what do you recommend in place of Langchain/Langraph?
I try to keep things simple; I do not use LangChain.
@@micbab-vg2mu:(