LangChain Beginner's Tutorial for Typescript/Javascript

[@chatwithdata]

Timestamps: 01:00 What problems are LangChain solving? 07:40 LangChain usecases + components 09:30 LangChain demo 11:07 Simple LangChain chain 14:34 LangChain prompt templates 21:11 LangChain agents 25:57 LangChain memory 29:19 LangChain indexes + embeddings (vectorstores) 42:27 Question answering docs

[@techknowledge5069]

can you prepare tutorial about pinecone how to use it and embedded it into the project .

[@MusicBox.Melodies]

Really really interesting! My big realization was at the end when you explained that we do the vector comparing offline and then we use the top search as context, that is how we save on openAi tokens! Thank you!

[@Grahfx]

Thanks to Andrej Karpathy I was able to find you.

[@resistance_tn]

Great video ! Looking forward for the next one :) It would be very interesting to discuss in a video how to manage data updates for a usecase such as QA on a big pile of Docs. As these docs could be for example Programming documentation that are meant to evolve and be enriched quite often, it would be so cool to find an efficient way (price, speed, reliability, etc..) to add these changes on the fly to the vectordb/index and not re-do the whole thing.

[@bakanndy]

Your github code just works and that is excellent !

[@nathanie1637]

Can LangChain JS work with a locally running Llama.cpp? if so, do you have any materials on getting that up and running?

[@yusakmanullang7656]

Two months late to this video. But still, pretty valuable. I want to say thank you for providing such complete informations.

[@junaidiqbal4104]

i want some guidance if you want to guide me 😀

[@altered.thought]

are you able to connect langchain with n8n? as I noticed langchain only have the zapier tool api at the moment.

[@chatwithdata]

Not officially. Maybe it's possible using a custom agent.

[@nivekmugo9660]

Which application do use to make the illustrations? Great video.

[@kaitakami8641]

Excalidraw

[@cerumai]

This was awesome, thank you for such simple explanations! Joined your newly launched course and looking forward to it!

[@mushroomthump]

Great tutorial! Do you notice any downsides to using the JS/TS version over the python version?

[@ConorBrown-tl9dv]

Can you explain how you set up your pinecone index? For example what values you chose for dimensions and metrics? I keep encountering an error with mine despite the fact that I have a working key and a discoverable index

[@robertodelgado1234]

Same here. I'm getting an upsert request error when I run vectorstores.ts despite correct keys. I'm assuming I have set the wrong dimensions or something.

[@b1chler]

@@robertodelgado1234 you can read up about the dimensions in the pinecone documentation. I assume the problem is you are on the free plan and your data is too big for the small free plan.

[@DeepakSwami-p3r]

@@b1chler i am also getting the same error when i try to insert one text file data which is almost 5 lines of data and getting the same upsert related error.

[@b1chler]

@@DeepakSwami-p3r your text document is probably too long. use supabase or other opensource vector storage solutions or pay for a bigger database on pincone.

[@DeepakSwami-p3r]

@@b1chler i am only using one text file it has only one line of 10 words.

[@robertodelgado1234]

Thank you for a great introduction!

[@MuhammadJunaid-g4e]

All good you explained very well. But I have some question I'm Sure I missed something. During question_answer_docs run every time i'm getting the result "I don't know." ---> even i don't know why. So i tried a diff question same answer. after checking code i didnt find anything so i tried a question "give me the summary" and response was way different. this is the content :"This context is about the importance of understanding different perspectives. It emphasizes the need to be open to different points of view in order to form a more complete understanding of a situation. It suggests that having a well-rounded perspective is necessary in order to make informed decisions." let me know if anyone have idea about this

[@pleabargain]

8:45 surprisingly there is no mention of coding tools. What of tools designed specifically to support legacy apps or to build new apps on existing modules| libraries?

[@chatwithdata]

Can you provide more details of what you mean by "coding tools"? Do you mean transitioning existing AI apps to use LangChain?

[@avg_ape]

Great explanation & walk through. Thank you.

[@tomb.8375]

Thank you. I have a question please. So what I understood at the end, is to answer a question from our data, we convert the query to an embedding, then do vector similarity search, get the block of text, make a prompt + user question + context text --> get the answer. That's okay, but that only gives us first level reference answers. For example: "How to do XYZ?" Our document is large and contains: "To do XYZ see ABCED topic". I want the answer not to be "See ABCED topic", instead I want it to keep internally asking to get the final answer. Any thoughts?

[@hyper_channel]

As things are right now, your use case might require to ingest a sanitized version of your document after doing a search and replace of those misleading statements

[@carlosvillu]

Great video! I am looking forward for the next one.

[@chatwithdata]

Thanks carlos. Let me know if you have any ideas for upcoming videos.

[@carlosvillu]

@@chatwithdata I would like to watch a Tutorial about how consume the notion API and create a Q&A bot. I am going to try to do it and it will be super helpful for me.

[@chatwithdata]

@@carlosvillu Here you go, just released this video: kzbin.info/www/bejne/pqPFnaKLfNGpgqc Hope you find it useful

[@Assoehosselaar]

Thank you very much for this detailled explaination!

[@arunraajj]

Thanks for the tutorial! really helpful

[@ravisojitra1037]

Super useful! Your discord community link is invalid, Please can you update it?

[@georgigeorgiev1576]

Great one, thank you!

[@aibeginnertutorials]

Langchain began as a Blockchain Project.

[@Gohans1107]

thanks very much!

[@SonAyoD]

Fantastic video

[@rajchandra2]

In openai's defence, chatGPT only has information till 2021 and since langchain was officially launched in 2022, it's expected to not know what it is.

[@snapdeus]

thanks for doing what you do :]

[@rodgetech]

Super helpful

[@chatwithdata]

Thanks. What related video tutorial would you like to watch next?

[@rodgetech]

@@chatwithdata Would be nice to see a video about the new Langchain Chat Models in action as they're relatively new.

[@chatwithdata]

@@rodgetech using gpt turbo?

[@chatwithdata]

@@rodgetech Good idea!

[@aadarshsingh910]

29:30 plz make a video on embeddings

[@csersin]

appreciate the work but way too many ads, very distracting for people who want to learn...

[@chatwithdata]

I don't monetize my videos. I think youtube puts them in by default.

[@csersin]

It is absurd. They were showing me an ad every five minutes.

[@gordansimpson1045]

'PromoSM'

[@peifeili67]