Hi. If this is not okay to post, please disregard it, moderator.
I have created a beta version of an app that allows you to chat with your own documents. It is very basic for now – you upload a PDF, and then you can ask it anything from the PDF. The idea is to eliminate the need to go through a large document or use Ctrl + F to find something you vaguely remember reading but can’t recall where you read it. Additionally, the app provides proper explanations for any tricky content found in the document.
You can try asking questions about your training diary or technical PDF, seek advice on how to apply concepts from a training book to your workouts, or anything else you’d like. Please provide feedback as this is my first time developing such an app, and I appreciate your understanding if I’ve overlooked something crucial. The app is free to use, and registration is not required.
I would totally use a tool like this. Tried uploading an 11-page PDF and got:
Internal Server Error
The server encountered an internal error and was unable to complete your request. Either the server is overloaded or there is an error in the application.
This is totally on me as people are using the app and I dont know enough about servers to scale it properly. Try again and pm me the document if it does not work, then I will upload it manually so you can try chatting with it. If it is not too much trouble of course.
Adit: Actually, it seems to be an Openai-problem. Will look at an alternative embedding
I would totally use a tool like this. Tried uploading an 11-page PDF and got:
Internal Server Error
The server encountered an internal error and was unable to complete your request. Either the server is overloaded or there is an error in the application.
The problem is fixed now. Try uploading the same document again
Is this anything more than just uploading the document to openAI and passing prompts through?
Yes, the problem with GPT is that it hallucinates. So, to avoid this, I force it to use your document as its knowledge base. But there is no secret sauce here; the approach is straightforward.
I have set up a vector database (Pinecone). Then, I take your document and make it into smaller chunks by applying a sliding window. After that, I convert the text into vector embeddings and send the result to Pinecone, using vector similarity search when you ask your question.
The only prompt I have added is: ‘Answer only if the answer is in the document, but give all the information you can:’
I had a server issue but now it runs without any problems. It uploads 5-6 pages/second so please try it out and let me know what you think!
OK, I did manage to upload the file, entered the prompt and this is the outcome:
Your Question:
What are the benefits of cable pooling?
Answer:
I don’t have access to the specific content of the publication you mentioned, but I can provide some general information about cable pooling. Cable pooling refers to the practice of combining the electrical cables of different renewable energy sources, such as offshore wind and floating photovoltaic (PV) solar, into a single shared cable infrastructure. This approach offers several potential benefits
Thanks for the reply. What did your document state explicitly? I have mainly been getting feedback from medschool students where gpt interactions are like:
Q: How many hours after taking ASS should I take Metamizol?
A: I am not a medical professional, but I can provide some general information. Aspirin (acetylsalicylic acid or ASS) and metamizole (also known as dipyrone) are both medications commonly used as pain relievers and fever reducers. It’s essential to follow the specific dosing instructions provided by your healthcare provider or the instructions on the medication packaging.
Wheras the app answers 45min, which is what the text book says. And also not answering the question like:
Q: Which Na chanel does Ranolazin inhibit?
A: Ranolazine is a medication used to treat chronic stable angina (chest pain). It works by inhibiting the late sodium current (INa) in cardiac muscle cells. By inhibiting the sodium channel, it reduces the influx of sodium ions into the cardiac cells during the late phase of the action potential, leading to a decrease in intracellular calcium levels. This results in a decrease in myocardial oxygen demand, which helps relieve angina symptoms.
The inhibition of the late sodium current by ranolazine is considered its primary mechanism of action for its antianginal effects.
Wheras the app answers the Nav 1.5-chanel. These are obviously very specific questions with very specific answers. But the idea must be to create a framework in which any question the document answers can be asked with a good response. Would be happy if you could give it another go!