hi Michael, I'm still new to chatbot applications. May I know if a prompt template is required for this setup or to make it more reliable? Thanks and great tutorial!
Have you tried ingesting a mix of pdf documents and slide decks? I'd like to create my own model for my own writings but also presentations I've created and I'm wondering if something like this will work. I assume I continue adding documents to the model after the initial upload. Loved the explanation, I'm just starting to think about how to ingest my own data into a model I can reference.
why is my bot way slower and super 'retarded' when it comes to questions - I am using same annual report from Macquarie, copied your setup and bot still cannot figure out what was the ebitda, should I play around with settings or sth.?
would this be able to do further functionality? im trying to make a chatbot trained on a textbook (which I have in pdf format) and I would like to be able to ask the bot to produce me flashcards or multiple choice questions, would this work at all?
yes, you would have to use custom langchain code, the output from the vector database is just 4 chunks of relevant text, see my first video on the channel for more understanding
Thank you for the production tip, subscribed.
hi Michael, I'm still new to chatbot applications. May I know if a prompt template is required for this setup or to make it more reliable? Thanks and great tutorial!
This is so cool, I watch 3 other guys on this flowise ... and this one was the most explained, and direct to the point!! well done 🙂
Glad you liked it!
Super newbie here. Could you set this up for the LLM to input pdf txt and then learn the specific "voice"? Anyone point me to a video?
Cool demo, thanks Michael!
Have you tried ingesting a mix of pdf documents and slide decks? I'd like to create my own model for my own writings but also presentations I've created and I'm wondering if something like this will work. I assume I continue adding documents to the model after the initial upload. Loved the explanation, I'm just starting to think about how to ingest my own data into a model I can reference.
So every time i send a request will it send 1000 tokens * 450 = 450,000 token
silly question but need to know. As it will increase cost
Is there a simple way to bring in PDFs? I can see the pinecone is more suited for efficient long term memory. What about short/ medium term?
This is the best way otherwise your run into context length issues
so, it requires the PAID version of the OPENAI?
why is my bot way slower and super 'retarded' when it comes to questions - I am using same annual report from Macquarie, copied your setup and bot still cannot figure out what was the ebitda, should I play around with settings or sth.?
would this be able to do further functionality? im trying to make a chatbot trained on a textbook (which I have in pdf format) and I would like to be able to ask the bot to produce me flashcards or multiple choice questions, would this work at all?
yes, you would have to use custom langchain code, the output from the vector database is just 4 chunks of relevant text, see my first video on the channel for more understanding
Any free alternatives to pinecone? Quite expensive. Anything self hosted?
Chroma, its open source
does using openAI API like you did cost or it is free ?
Yes its costs money
Is flowise free to use?
yes
@@michaelborman no it very expensive. I tried it with my documents, and in 10 minutes chat, generated more than 500K tokens on openai
it is going to be depreciated