OpenAI + LangChain Wrote Me 100 Custom Sales Emails

[@nattapongthanngam7216]

Thank you, Sensei Greg, for this amazing demonstration of LangChain's capabilities!

[@menpathy]

Hot dang. My brain is buzzin and I love it. This looks like it was a lot of work to put together and I appreciate your generosity in sharing it with us.

[@DataIndependent]

Nice! What sorts of ideas do you want to do? I'd love to help with demo videos

[@menpathy]

@@DataIndependent Great question. I would love to use transcripts from my own recordings as a database to use as a chatbot for my own website. Is that a ridiculous idea?

[@DataIndependent]

@@menpathy Not at all, I've done a bit of transcript work as well. Check out this tweet for my demo project using transcripts: twitter.com/GregKamradt/status/1630717735398428672 It'll take a bit of pre processing but you'll be able to do it.

[@alessandroferrari2166]

thank you for all your amazing videos you've been releasing lately!

[@DataIndependent]

Awesome thanks Alessandro

[@ac_cobra8540]

These videos are superb Greg. I've hit Notification alerts for the all rest. Thank you.

[@DataIndependent]

Awesome, thank you!

[@lasanthasanjaya9918]

Superb Greg. very informative as a student, Thank you.

[@DataIndependent]

Nice! Glad to hear it - gpt4 makes this process way better

[@enlaichu]

Loving your videos on Langchain. Thanks for making them! Btw i’m trying to find the video you mentioned about summarising sales calls on Twitter but can’t seem find the tweet. Would you mind linking it here please? Thanks!

[@enlaichu]

Ah found it. I had followed @DataIndependent but just saw it on @GregKamradt

[@DataIndependent]

twitter.com/GregKamradt/status/1630717735398428672

[@Sean-pc8iu]

This is really helpful stuff, I've been going through all these docs. Question: is there a way to download the code itself? I'm new to python (and programming in general), and having a hard time trying to patch it all together. Seems like some line items are hidden. I run it all in visual studio and the command prompt on windows. Is there a better way? Thanks & keep up the great work

[@DataIndependent]

Yep totally - you can head to the github page in the description. If you need help with downloading it you can ask chatgpt on how to "clone a repo". Or you could just copy and paste it into your own jupyter notebook

[@cstu107]

Thank you for your video. May I ask if LangChain can set the output language? For example, Spanish.

[@DataIndependent]

Try it out! You can add that command to your output and see what it does. I wouldn't bet on OpenAIs Spanish though...

[@ChenXibo]

super cool!

[@DataIndependent]

Nice! Thank you

[@ZZ-yw4xf]

This is brilliant! Thanks for sharing it. In my use case, I have a lot of pre-composed personalised emails by human writers, and we want GPT to 'learn' the tone of these emails and maybe also understand what to sell to the customers (e.g., cheapest price, warranty, or product novelty). Again there are examples from our database of past emails. Can I ask what is the best way to use that kinds of data?

[@oliva8282]

Why is it needed to put input documents in cell 16, minute 11:26?

[@德艺双馨比利周]

Thanks Greg I love you videos! I have been trying to come up with my own chat bot based on a bunch of txt files. Do I add all of the text into one txt files and give that one file to Pinecone? Or do I convert them into CSV file before I give them to CSV? Thank you so much man. Your videos have been super helpful to me. It would be great if you can do a video on something like this

[@DataIndependent]

You can do either. You can add all the embeddings in one go, or consistently add them as you go. Up to you. Thank you for the kind words!

[@saadkhattak7258]

Hi Greg, Thanks a lot for making valuable content :) Quick question, would it be possible if you can share the dataset that you have used in the video? :)

[@AlmazLab]

treasure! tnx!

[@DataIndependent]

Nice!

[@lorenzoleongutierrez7927]

Great !

[@DataIndependent]

Nice - thank you

[@AyushSharma-ux4fk]

hey - loving the series. can you please share the slides as well along with the code? Looking back at the slides makes it easy to create a mind map of what we are achieving.

[@DataIndependent]

I don't have them in a ready state to share right now. But I will do that going forward

[@RyanSmith-rb1ch]

Would love a version of this but pre-training it with web content.

[@DataIndependent]

What pre-training are you referring to? What types of content?

[@sersoage]

are you using gpt.3.5? or davinci?

[@DataIndependent]

The default for OpenAI LLM (from LangChain) is davinci right now. I haven't tried to convert this process to a message based system to see if it would work better. How's your experience been with gpt3.5?

[@sersoage]

@@DataIndependent it’s not bad but prompts need some rework

[@DataIndependent]

@@sersoage Do you get better results even after the rework? If the value is there I might try switching over to it. I'll investigate either way

[@sersoage]

@@DataIndependent considering the 10x decrease in price yes. The no fine tuning (so far) is a big drawback

[@alirezakhatami7813]

@@DataIndependent davinci is pretty expensive when you scale. use the model_name param within the OpenAI class initialization and set it to gpt-3.50-turbo

[@ksanjaykumar4412]

hey, do you have any ideas related to GPT-3 having access to internet? I haven't found any resources (to built it). Let me know if you have any resources to share!!

[@DataIndependent]

No word on that yet. I imagine that it won't have access to the internet for a while. The work around is you loading what whatever you need from the internet, then giving it to openai to process for your need

[@ksanjaykumar4412]

@@DataIndependent got it!!

[@SahilSanganis]

Can you create a video on how to create unlimited facebook video ad scripts for a particular product of e-commerce using major marketing books of all time available on the internet? It can become a saas app too. We find it difficult to find ideas fr our products sometimes and this is what I’m trying to build and now i found your videos really helpful. Thanks and please respond

[@DataIndependent]

This sounds like something jasper would be used for. They are experts at marketing material