Document Analysis
Requires v1.17.0 and later
Last updated
Requires v1.17.0 and later
Last updated
BoltAI allows you to attach a document to the chat and let you ask a LLM about it. Think of it as "Chat with PDF", "Chat with CSV"... kind of application.
It's straightforward: add the documents to your chat, then start asking questions. There are multiple ways you can add a file to the chat:
Drag and drop files to the chat, or
Copy the file and paste into the chat input or
Open these files with BoltAI in Finder
You can manage the attached documents in the Project section. Open the right sidebar (1) and select the Project tab (2)
By default, all attached documents are included in the System Instruction and are not in sync with your file system. You can exclude, or delete a document from the chat easily.
Note: Tokens are estimated using the GPT-4 tokenizer can might be inaccurate.
When working with ChatGPT or Claude app, you need to upload your files to their server. Whenever you make a change, you will need to upload the new version, again and again.
BoltAI allows you to "link" directly to your files in the filesystem with the File Sync feature. When enabled, BoltAI will always use the file's latest content. This is handy when working with source code or other text documents.
Does BoltAI support Retrieval Augmented Generation (RAG)? Not yet. The current implementation adds the document's full content to the System Instruction so make sure you're using a LLM with a large enough context window. For more advanced use cases, consider using PDF Pals instead.
What document types are supported? Text documents and source code (js, txt, markdown, csv...), PDF, docx, xlsx.
Can you add support for <insert_file_type>? Please create a feature request on our Canny Board.
Click the plus button and select "Add Document (Full Content)", or
To enable File Sync, click on the ellipsis button and select "Enable File Sync"
