CiteWise: a local research desk that will not invent a citation

Описание и применение

CiteWise – локальный инструмент для поиска и проверки цитат в PDF, который не отправляет документы в облако и не выдаёт вымышленные цитаты.

Задача

Users need a way to search and cite passages from PDFs locally without sending documents to a hosted API and without the model inventing citations.

Результат

CiteWise enables local PDF search and citation verification without sending data to external services, and it transparently flags insufficient model answers, allowing users to inspect the source text directly.

Рабочий процесс

• Upload PDFs to the local app
• CiteWise extracts each page and keeps the original page number
• The text is indexed with rank‑BM25 and a local FAISS index
• User searches or asks a question
• CiteWise retrieves relevant passages and sends them to the local Ollama model with a JSON schema
• The model returns an answer; if it cannot provide a source, the answer is marked "insufficient"
• The user can view the retrieved page and the model’s answer side by side to verify the citation

Доказательства

CiteWise extracts each page, keeps the page number from the PDF, and indexes the text.

Источник

The answer is supposed to point at a passage. If the passage is not there, or the model will not commit to one, the answer is marked insufficient and no quotation is invented.

Источник

To run it locally you need Python 3.12, Node, and Ollama with a chat model pulled, for example ollama pull llama3.2:3b.

Источник