Learn LangChain.js - Build LLM apps with JavaScript and OpenAI

At the end of this course, you will be able use LangChain's Expression Language to build GPT-powered chatbots that have specific knowledge about an underlying dataset. You will use text embeddings and a vector database to perform retrieval-augmented generation (RAG).

The course requires an understanding of JavaScript. Here is a free JS course if you need a refresher:

✏️ If you have any feedback to the teacher Tom Chant, please reach out to him on X here:

⭐️ Code ⭐️

⭐️ Chapters ⭐️
⌨️ (0:00:00) Introduction
⌨️ (0:02:01) What you'll learn and build
⌨️ (0:05:36) App Flow Diagrams
⌨️ (0:09:04) What are embeddings
⌨️ (0:15:20) Supabase Setup
⌨️ (0:19:27) Split the text
⌨️ (0:21:28) Split the text II
⌨️ (0:27:47) Upload to supabase
⌨️ (0:31:53) Starter code
⌨️ (0:33:19) Explainer The Standalone Question
⌨️ (0:35:31) Aside Prompt Templates
⌨️ (0:38:59) Aside Prompt Templates II
⌨️ (0:42:08) Adding the first chain
⌨️ (0:46:11) Retrieval
⌨️ (0:52:45) Add StringOutputParser
⌨️ (0:55:15) Fetching the answer the template
⌨️ (1:00:39) Serialize the docs
⌨️ (1:02:53) Aside RunnableSequence
⌨️ (1:07:54) Aside RunnableSequence 2
⌨️ (1:10:57) Aside RunnableSequence 3 RunnablePassthrough
⌨️ (1:14:47) Super Challenge - add the RunnableSequence
⌨️ (1:17:10) Super Challenge - solution
⌨️ (1:21:47) Wire up the UI
⌨️ (1:24:34) Setting up the memory
⌨️ (1:27:27) Super Challenge Wire up the memory
⌨️ (1:33:51) Performance Issues Check-list
⌨️ (1:37:44) Outro

🎉 Thanks to our Champion and Sponsor supporters:
👾 davthecoder
👾 jedi-or-sith
👾 南宮千影
👾 Agustín Kussrow
👾 Nattira Maneerat
👾 Heather Wcislo
👾 Serhiy Kalinets
👾 Justin Hual
👾 Otis Morgan
👾 Oscar Rahnama

--

[kasper]

Scrimba teachers have mastered the art of teaching. Thank you for this video.

[jessecasey]

I was just teaching myself this, you guys always seem to know what we need

[paul-olof]

I'm a few lessons into the free Js course above and it's really good. Without a doubt I'm following both of these to the end. Thanks really!

[abdelhaibouaicha]

📝 Summary of Key Points:

📌 The video discusses the use of Lang chain, a tool for building AI applications, specifically focusing on the concept of output passers and the use of runnable sequences.
🧐 Lang chain offers various output passes for different situations, such as outputting in JSON or binary data. The video demonstrates the use of the string output passer, which passes the output as a string.
🚀 The video introduces the concept of a runnable sequence, which allows for more complex chains of prompts and models. The video provides examples of how to use the pipe method and the runnable sequence to chain together prompts, models, and output passers.
🚀 The importance of formatting the output and using input variables effectively is discussed.
🚀 Performance can be improved by adjusting parameters such as chunk size, overlap size, and OpenAI settings.

💡 Additional Insights and Observations:

💬 "Lang chain offers a powerful and flexible way to build AI applications by chaining together prompts, models, and output passers."
📊 No specific data or statistics were mentioned in the video.
🌐 Lang chain is a tool provided by OpenAI, a reputable source in the field of artificial intelligence.

📣 Concluding Remarks:

The video provides a comprehensive overview of using Lang chain to build AI applications. It explains the concept of output passers and demonstrates the use of the string output passer. The video also introduces the concept of a runnable sequence and provides examples of how to chain together prompts, models, and output passers. It emphasizes the importance of formatting the output and using input variables effectively. Overall, Lang chain offers a powerful and flexible way to build AI applications, and performance can be optimized by adjusting various parameters.

Made with Talkbud

[huangvivi]

How do you setup all the langchain and supabase dependencies locally?

[ItsMe-rymk]

if some one like me (noob) encountering error while document load add the just 3 columns in supabase metadata jsonb, content text, embedding and disable database->edit (RLS) if im wrong please correct me

[AB-kqxm]

The scrimba ui with the video on the code editor is dreadful. Improve it if you want it to thrive

[vedicmusings]

I think they have updated. When we do the punctuation Prompt it is corrected the grammar also, so no need to run the grammarPrompt

[MSs_.]

Hello, I really like your videos and I'm able to learn a lot from the courses you have here. As a Brazilian, I only wonder if there is a way to translate some videos, in addition to the Brazilian language into other languages. YouTube's own translation is not that good and some words often make no sense at all.

[samshosho]

great tutorial indeed. Thanks a bunch.
Just one question, can this technique be used with another LLM downloaded on one's computer? like llama for example. So doing the same thing without having to pay for tokens and such.
Thanks,

[girishanker]

That's a great tutorial ⭐ Thank you

[FatemehDamarchilou]

You exactly know what we need.
Awesome

[inaamullah]

Perfect, thanks. How does that works with data bases with more than one parameters retrieval?

[stephenjames]

I got a quota failure when I tried to upload to supabase, is the free tier not sufficient to complete this tutorial? BTW, great course.

[dhiransapkota-ug]

when the conversation increases and in every invikation we need to pass the whole conversation history, doesn't that increases the token size we are consuming which could be very costly. Is there any solution to that. Suppose chatbot responses with more that 1000 tokens for 10 token question and we need to give this size of token along with next question, that doesnot seems efficient way or am i missing something here? What is the solution.

[neerosity]

can we push new data embeddings in same document / table which we already have embeddings? or should we have new table for each new docuemnt?

[BuyCarsTVPakistan]

Thanks but when I tried your example on my webpage, I ran into all sorts of problems, don't know how it works on your end. For me, it required to have "<script type="importmap">" above the script, that did it but now i am stuck at dotenv and fs as it seems they cannot be used inside webpage. How did you run in your demo please ?

[soorajthekkepatt]

will this work with the latest langchain updation

[quanxom]

The database has a large number of records of about 16666 records and has error 57014 canceling statement due to statement timeout null. Is there any way to fix it?

[lucasfrutig]

why in my application node can't load the file with fetch