File Upload
in progress
l
leo
It will be really helpful if I can upload files directly into the chat
By files, I mean a variety of formats: text documents, code, PDFs, and even compressed ZIP files. I believe many of us frequently refer to certain documents during our chat sessions. With a file upload feature, we could share these documents right within the chat, reducing any unnecessary back-and-forths and enhancing our productivity.
Moreover, this feature could come in handy when we're dealing with tasks like data analysis or summarizing data. Quite often, these data are compiled into ZIP files.
Daniel Nguyen
Update: I've added the ability to attach documents for analysis in v1.17. But it uses full content as context.
I will add support for RAG and Code Interpreter soon.
Daniel Nguyen
Update: I've added the ability to attach documents for analysis in v1.17. But it uses full content as context.
I will add support for RAG and Code Interpreter soon.
Vec
Daniel Nguyen Could we somehow keep full context or RAG would be better? I think previously claude I just uploaded the full context as well.
Daniel Nguyen
Vec if cost is not an issue, I find that full context is almost always better. But it has a limit in max context length so RAG is still useful.
Vec
Daniel Nguyen I see. Currently I'm using the free mode of Google Gemini with 50 times Pro usage per day, so personally I think full context may be better helped.
a
avijit ghosh
Another ask for this. A PDF reader in BoltAI would be incredibly powerful.
If you combine it with a RAG such that it reads say a folder of PDFs (or say a Zotero library via a plugin ) I think you'd have an unstoppable scientific research platform.
L
Lee Simpson
Any progress with this feature. Its the number 1 missing feature compared to the default OpenAI app.
Daniel Nguyen
Sorry for the delay. I'm back to building this. One quick question, do you think we should change the name of this "Assistant" in BoltAI?
Currently, an AI Assistant in BoltAI means a custom GPT configuration and can be used accross LLM models. The configuration is stored locally and is platform-agnostic.
The OpenAI's Assistant can only be used with OpenAI.
I'm afraid that using the same term can be problematic.
One of my solution was to use "Smart Assistant" or "Cloud Assistant" for OpenAI's Assistant API, but it's probably even more confusing.
What do you think? cc Victor R leo Vec
V
Victor R
Daniel Nguyen I understand the dilemma with the naming. However, since "OpenAI Assistant API" or even "Assistant API" should be synonymous with OpenAI's implementation, I believe it would reduce confusion. Even if the name is shortened to "Assistant API," there could be some text or wording indicating that "Assistant API" refers to the OpenAI Assistant API component.
P
Paul Webber
Daniel Nguyen Why don't call the new feature "File Assistent" or just "file upload"? I hope I got your question right.... however it will be called, I'm really looking forward to finally can upload CSVs and PDFs 💪🙏
l
leo
Daniel Nguyen I think you can use another name for it, like the Paul Webber said.
I suggest something like "File Analyst". It is not user-friendly and also distinguishes it from "Assistant" label of OpenAI.
Daniel Nguyen
in progress
Daniel Nguyen
planned
I've been working on supporting the official Code Interpreter API from OpenAI.
However please note that you might be charge extra for using these tools
Is that what you want?
l
leo
Daniel Nguyen: My main focus was in the file upload and to ask question about it. Like a pdf or a csv or a zip with some files. In some cases will be extract a json from a file, or summarize it or even generate a chart to use on a presentation.
About the charge, i understand that it will cost more than normal text conversations, i don't have any problem with that.
A side question, is a custom function needed to upload the file ?
V
Victor R
Daniel Nguyen Yes this would be a great addition to Bolt AI to be able to process common file types like XLSX, PDF, CSV, JSON, YAML, PROGRAMMING LANGUGES file extensions etc. using the OpenAI Assistant API.
Vec
Daniel Nguyen I think this is acceptable. OpenAPI's interpreter is strong in function actually. Meanwhile, I think what they want here is something like chat with pdf.
A
Alessandro
Vec For PDF there is PDF Pals already
Vec
Alessandro they're disucssing on data analysis, not fetching answers from pdf
Rodrigo Paiva
Daniel Nguyen, pls, make sure that it supports Markdown files (.md). In the future, we could sync the chat with a folder to use it as a second brain with some vector db inputting the right context to a query, it would be great.