RECORD
- ID
- 004
- WORLD
- Knowledge
- TYPE
- Knowledge base
- STATUS
- Live
NOTES
The idea
A Gujarati archive you can query without paying anyone. Books, scans and audio are stored and searched on your own machine, and two kinds of question are answered: where exactly does this appear, and what does the archive say about it.
Why it exists
To have a free, grounded knowledge base for Gujarati content that does not depend on commercial services.
What it does
- References: exact phrase, several words, AND / OR / NOT, wildcards, nearby words and loose spelling — each result cited to a book and page, or an audio file and time
- Ask: the question is understood, the right passages are found, and a Gujarati answer is written with numbered sources
- Reads plain text, Word files and scanned PDFs, with OCR running on a local GPU
- Reference books can be kept out of Ask so they cannot mislead it, while staying searchable in References
- Narrow questions fetch one focused passage; broad ones fetch more
DETAILS
- PUBLIC SITE
- References only — Ask runs locally on a home machine
- BUILT WITH
- Python, SQLite, local language models, Streamlit