01 · extract
Extract
Text from PDFs, Word, RTF, spreadsheets and web pages, plus AI-powered OCR for scans, faxes and photographs that keyword search can't read.
Product · Archive Intelligence
Ask your archive. Get answers with sources.
Scanned letters, reports, photos and transcripts, in several languages. Ask a question in plain English and get a written answer that cites the exact documents it came from, with the original one click away.
Run it in the cloud, or entirely on your own hardware, so your documents are never shared with anyone.
Answers
Answers, not keyword hits
Keyword search tells you which 400 files contain your word. Archive Intelligence gives you the answer and the documents that prove it.
$ ask "Who approved the 1958 shipping contract, and was it ever amended?"
The contract was approved by the board chair, E. Lindqvist, on 12 May 1958 [1].
A first amendment extending the delivery window was signed in March 1959 [2];
a Norwegian memo confirms no further changes were made [3].
[1] Board minutes 1958-05-12.pdf Swedish → English
[2] Amendment no. 1.docx
[3] Internal memo (scan).tiff Norwegian → English · OCR
How it works
From a document dump to an archive you can question
02 · normalise
Normalise
Every document is language-detected and translated to English, and images get a written description. Your team no longer needs to read Swedish, French or Norwegian to find things.
03 · index
Index
Full-text and semantic (vector) search combined inside PostgreSQL with pgvector, so a question finds the right document even when it uses different words.
04 · answer
Answer
A language model, running locally if you choose, reads the most relevant passages and writes the answer, citing every source. Click a citation to see the original next to it.
Confidential by design
Your documents, shared with no one
Archive Intelligence runs in the cloud, but it can also run entirely on your own hardware. For legal evidence, government records, investigative material or company history, that means the AI comes to the documents, and no third party ever sees them.
- run locally
- OCR, translation, search and answers can all run on one GPU server you control. No third-party AI APIs, no training on your data.
- or in the cloud
- When the documents allow it, the same system runs in the cloud.
- traceable
- Numbered citations open the original scan or file inside the app, so researchers verify instead of trust.
- no copies
- Originals are viewed in the browser, not downloaded. Access is limited to named users.
- multilingual
- Mixed-language collections become one consistent English corpus, with the original-language text one click away.
- reports
- Turn an answer and its sources into a private report that becomes searchable alongside the archive.
Who it's for
Archives nobody can read end to end
For records management, research and investigations, and any collection too large or too multilingual to read by hand.
- Law firms & e-discovery
- National & municipal archives
- Commissions of inquiry
- Investigative journalism
- Corporate records & compliance
- Museums & research institutes
- formats
- PDF, Word, RTF, spreadsheets, web pages, scans, faxes and photographs
- search
- PostgreSQL full-text + pgvector semantic search
- models
- OCR, translation and language models, on your own GPU server or in the cloud
- operations
- Installed, monitored and maintained by a.s.k. solutions
Next step
See it on your own documents
We will show it working on a sample of your own documents, on your premises if you prefer. The same team builds the system, runs it, and supports it.