Unleash OpenAI Assistant 2.0 New File Search with Vector Store - Step-By-Step
Vložit
- čas přidán 15. 06. 2024
- File Search augments the Assistant with knowledge from outside its model, such as proprietary product information or documents provided by your users. OpenAI automatically parses and chunks your documents, creates and stores the embeddings, and use both vector and keyword search to retrieve relevant content to answer user queries.
#openai #openaiassistant #chatgpt4 #filesearch #rag #pinecone #vectordatabase
📚 Learn Python & AI in my skool community:
www.skool.com/customgptaaa
🤝 Get A FREE Quote For An End-To-End AI Software/App: hcalendly.com/connect_imtiazh/...
🤝Linkedin: / hasanimtiaz
My Links 🔗
👉 Subscribe: / @customgpt.aiacademy
👉 Instagram: customgptaaa
👉 Tiktok: customgpt_aiaa
Resources/Links Mentioned:
👉Download the notebook from the resource
www.skool.com/customgptaaa
About Me 🙋♂️
Hello, I'm Imtiaz, a seasoned expert in data science, AI development, and prompt engineering. My mission is to empower your projects with Custom-GPT solutions. Whether you're looking to build, deploy, or sell GPT chatbots with no code/low code expertise, I'm here to partner with you. Let's leverage my extensive experience to elevate your projects, transforming innovative ideas into practical, profitable solutions. Join me in navigating the exciting world of AI and let's create something remarkable together.
Chapters:
0:00 Introduction to OpenAI's New File Search Tool
0:44 Framework Overview
1:06 Understanding Retrieval-Augmented Generation (RAG)
3:24 Keyword Search vs. Semantic Search
5:11 Semantic Search with Vector Databases
8:00 Navigating OpenAI Documentation
9:21 Implementing File Search Assistant with Python
13:14 Outro
Don't forget to like, share, and subscribe to support the channel - Věda a technologie
Have you integrated File Search into your Open AI Assistant? 🤔
Join my free skool community to enhance your python and AI skills
www.skool.com/customgptaaa
Get A Free Quote for your AI software/app
calendly.com/connect_imtiazh/30min
Help, I upload a TXT file to a vector store and the status appears as failed, why get this?
Convert it to docx and try again
Great content mate! Very informative
How can I make the assistant only answer question with the given knowledge? Mine keeps answering questions even when questioned with outside of the files scope.
You need solid prompt engineering bro or maybe fine tune the model. Lets hop on a google meet to sort it out
www.skool.com/customgptaaa/about
Hello,
I uploaded two files to my assistant, which are guidelines for writing case reports. The assistant automatically generates a vector store for these files. If I later use this assistant to input a long text interview record and ask the assistant to create a case report based on the uploaded guidelines, do I need to write an additional prompt specifying which file ID or vector ID to reference?
I'm not very experienced with programming and am currently only using the assistant playground.
depends on your use case. I would recommend uploading vector store to the assistant if you want to use general knowledge for the assistant. and then create vector store seperately and attach it to thread to have assistant reference that specific document. The only way i can think of doing this is programmatically. Send me a msg on skool and lets talk
www.skool.com/customgptaaa/about
@@CustomGPT.AIAcademy
Thank you very much. Currently, I am using the assistant playground to upload two Markdown files for knowledge. OpenAI assistant V2 automatically generates a vector store (VS). It seems that no matter how many files I upload, there will only be one VS, but I can still see the information of the two uploaded files and their file IDs in the storage.
Since the content of these two guidelines for writing case reports is relatively small, totaling about 15,000 tokens, it is similar in size to my interview records.
Can I remind the assistant in the prompt to use the unique VS ID of the attached files for summarizing reports?
However, I feel that if I specify the guidelines clearly in the instructions and prompt, the AI will understand better. Something like:
Please refer to the following materials to write the case report:
Summary Writing Guide: file-1 id
Mental Status Examination Guide: file-2 id
Vector Data: vs_3 id
@@CustomGPT.AIAcademy How can I add this? Does the discussion require a fee? Thank you.
@@user-mv9ul9tz1c nope. Its free bro
Is there any way to retrieve vectors for file id or to know which file id was used to answer a particular question?
this is not currenly available under openai framework. The way to get around this to use file naming, manual logging or considering adding custom metadata to your files
Include an identifier in the in the context of the file that will be returned after search is complete, it's easier to do if you do not use OpenAI's vector storage and carfefully construct your embeddings
If it is necessary for my company where I receive calls from customers. I want to make a voice assistant, which will answer questions about price and others. where can I find out how best to structure data for such a task
Hey man. You would keep your knowledge base the same as in docs, pdfs etc, you would need to create an application or leverage an app that also converts audio input to text and then text to speech for output like the whisper model. We can talk more on google meet if you want. Schedule it on my calendly link
how do you call a pre-built assistant on the openai platform using the assistant_id?
In the platform, you can select them by the assistant name
@@CustomGPT.AIAcademy So you just reference the assistant_id and initiate a run/thread?
@@innovationroom3701 are u running it inside the openai platform or in a development environment?
@@innovationroom3701 we can get on a quick google meet call in 15 mins if u want meet.google.com/wnh-hpoi-tyu
@@CustomGPT.AIAcademy dev environment
RAG which one is better? llamaindex or file search in openAI?
Great question! Depends on your use case. Llamaindex is only a query and index framework similar to RAG but can be connected to vector database and can offer dynamic updates. It is complex to set up. File search is easy to set up but has less customization features.
@@CustomGPT.AIAcademy Can you please tell more? I wonder if llamaindex or file search, which one can connect to database? As I want my assistant can understand more about my database. But I'm not sure how we can make an assistant have knowledge about our database.
@@ngocvikhanghuynh5424 filesesearch already use vector database bro.
Hi, I'm trying to make an app using gpt and I want to consult with you, is it possible in 3 or 4 hours?
Sure. Send me a msg here www.skool.com/customgptaaa/about