Several responses suggest using Google's NotebookLM, which the user found satisfactory. However, it was noted that NotebookLM is currently limited to 20 documents per notebook and only works with Google Drive. Another suggestion was to use OCR and pattern matching on text, which are computationally inexpensive and straightforward to implement. For instance, tax documents often contain the name of the tax authority and the year, which can be used as search parameters.
Key takeaways:
- The user is seeking a local-first AI solution to search through many PDFs, including scans and non-scans, for specific documents.
- The ideal solution would be a Mac software that can access iCloud, where the user stores all the documents.
- The user prefers a solution that does not require any tagging and is okay with false positives in the search results.
- One suggestion provided is Google's NotebookLM, which the user found satisfactory, but it is limited to 20 documents per notebook and only works with Google Drive.