Scanned registers, photographed pages, and Devanagari documents become text you can search, extract from, and route.
HOW IT WORKS
Most institutional history is on paper that no search tool can read.
Registers, circulars, signed agreements and photocopies hold decades of decisions. Each page goes through the same five stages, whether it arrived as a clean scan or a photograph taken on a phone at a branch counter.
Straighten skewed pages, lift faded contrast, and remove speckle from photocopies so the text underneath can be read.
Work out where the columns, tables, stamps and signature blocks sit, so a two-column circular does not come back as interleaved nonsense.
Convert printed Devanagari and Latin script into text, including pages that mix both — which most Nepali official documents do.
Dates, reference numbers and the parties named in the document are extracted automatically as the file arrives.
The text is indexed next to everything else you hold, inherits the document’s permissions, and can start a workflow.
| Script | Printed | Handwritten | Notes |
|---|---|---|---|
| Devanagari (Nepali) | Supported | Planned | Printed registers, circulars, notices and agreements set in Devanagari. |
| Latin (English) | Supported | Planned | Correspondence, contracts and reports produced in English. |
| Mixed on one page | Supported | Planned | The common case in Nepali official documents — Devanagari body with English figures, dates and proper nouns. |
Each of these starts with a document that exists, is authoritative, and cannot currently be found by anyone who did not file it.
Decades of bound registers and ledgers sit in storage because finding one entry means sending someone to look for it. Once the pages are read, an entry can be found by name, date or reference number.
The executed copy carrying the signatures is often the only authoritative version, and it exists as one physical document. Reading it makes the terms searchable and the key dates extractable, while the original stays intact as evidence.
Circulars and notices frequently arrive as scanned PDFs with no selectable text in them. Reading them on arrival makes their content searchable the same day, and catches repeat copies rather than filing them twice.
Customer and applicant files are assembled from photographed identity documents and supporting paperwork. Extracting the reference numbers and dates at intake means the file is complete before it reaches whoever approves it.
Not yet. Printed Devanagari and Latin text is supported today. Handwriting recognition for older Nepali records is on the roadmap and is not something we claim to do now — if your archive is predominantly handwritten, say so early and we will tell you honestly what proportion is currently readable.
Pages are straightened, contrast-corrected and de-speckled before any text is read, which recovers a lot of faded photocopies. Where a page is genuinely illegible, it is stored and indexed by its metadata rather than being silently dropped, so the record still exists and can be found.
Yes. Mixed Devanagari and Latin on a single page is the normal case in Nepali official documents — a Devanagari body with English figures, dates and proper nouns — and is handled directly rather than requiring the document to be split.
Yes. The original scan or photograph is retained exactly as supplied. The extracted text is stored alongside it rather than replacing it, so the source document remains available as evidence.
Only whoever could already see the document. Extracted text inherits the document’s permissions, so reading a file never widens access to its contents — including through search results.