Видео

Stick to Chinese

Thanks for watching!!

Thanks for watching!!

Hi I was using the highlights from Adobe PDF reviewer.

Thank you for watch!! Hope you have a nice day!

Thank you for the helpful tips!! I will do a better job. 😊

Thank you for watching!!!

Yes, you are right!! It cannot go beyond the content you have. Cuz the purpose of this is to find content in you PDF, not from ChatGPT. If you really want to go beyond your PDF, one solution: you can wrap your search function to Function Calling from ChatGPT endpoint, you can get some ideas from ruclips.net/video/gtbj4AgOkTo/видео.html. Hope this is helpful for you.

Yes, you can search another words, it will give you other related content. You can check this video, it shows more example ruclips.net/video/MERQHJBCq4c/видео.html.

Thanks for watching!!

Thank you for watching!

Thank you so much for watching. I will record another video how to use LLMs from Huggingface to replace OpenAI. Stay tuned. Thanks.

@@techwithray8943 Id prefer to be able to use oobabooga/textgen as my local-api, but id settle for anything that lets me run a local flowise pipeline utilizing a wizrdlm gptq model (preferably wizardLM-uncensored-falcon-7b-gptq, but again, ill take anyhting that works) .... PLEASE GOD HELP! xD

Thank you so much for your support!!! Will keep working on good content!

Thank you so much for watch! Do you mean you want to search info from your database and serp, then send all the relevant content to LLMs (OpenAI) ?

@@techwithray8943 This is almost the meaning. There is a lot of knowledge in openai that is outdated or incorrect. It needs to be re-aligned with the local database. If you only use serp to search, there will also be an error message. In specific problems, use a dedicated database first, and then use LLM would be a better choice

@@liuchuchenliu2802 ***I SPENT $9 DOLLARS IN OPENAI API CALLS BUILDING A CUSTOM DB YESTERDAY- DOING IT THIS WAY THROUGH FLOWISE IS NOT OPTIMUM*** add "conversation summery memory" to have persistant memory between queries, in-memory venctor store, vector store retreiver, hugging face embeddings, local file system fais vector retreiver, and fais vector store to write your db to local disk... i think thats, use text file and text splitter modules if you want to give it guidance on the stored vector emmbeddings, and have a realy long convo or prompt chain that stays on topic of what you would like it to "learn" (add vector store retrieve prompt template, and tell it its obsessed with a certain topic to laser-focus on specific data)... when you want to implement it , use fais retrever, , a context memory store of some kind, and conversational QA chain, vector db QA chain...

Hi thanks for watching. You can use LLM Chain to directly use prompt, or use Conversational Retrieval QA chain to add "inputs" along with embedings and send to OpenAI API.

Thank you so much.

Hi, I did not use azure openai api. It should be simple to use it. Let me try it first and get back to you. Thanks.

@@techwithray8943 Thank you, but I just succeeded after changing a set of APIs. Obviously, there is a certain difference between using azure and native openai to generate content😄

@@liuchuchenliu2802 I am so glad that you figured out. I checked the azure oai services, the authentication is a bit different and the gpt version might be different from what we choose from openai.

Hi Thank you for your question. Q1: Hugging Face Embedding is free, while OpenAI Embedding will charge fee based on the number of tokens. If we embedding PDF files with few hundreds of pages, the fee might be out of expectation. Q2: We do not need to put the model for embedding, as there is DEFAULT_MODEL_NAME. You can check Flowise source code here github.com/FlowiseAI/Flowise/blob/ff93d11913e00ef2daa390757a2a0ed14485c234/packages/components/nodes/embeddings/HuggingFaceInferenceEmbedding/HuggingFaceInferenceEmbedding.ts#L3, it use the HuggingFaceInferenceEmbeddings, which has the DEFAULT_MODEL_NAME, please refer here python.langchain.com/en/latest/_modules/langchain/embeddings/huggingface.html (python code and LangchainJS should be the same logic).

@@techwithray8943 Thank you

Thank you so much for watching. I am new to RUclips and still improving my video quality.

Thanks for watching. I will work on this request and record another video.

Thank you!!