Historical conversation: Sovi Research Librarian

# Conversation: Sovi Research Librarian

User: I keep losing papers, URLs, and quotes across Grok threads. I want a research librarian that treats every conversation as a potential library accession.

Assistant: We should not build a generic notes app. The librarian should:
1. Ingest conversations and documents
2. Extract sources with URL, author, date, reliability notes
3. Store an archived excerpt, never a dangling bookmark
4. Maintain a citation graph
5. Answer questions with citations back to the original corpus

User: Decided: evidence status is first-class. SOURCE-SUPPORTED vs WORKING HYPOTHESIS vs SPECULATIVE IDEA.

Assistant: Also extract methodologies and datasets. A paper is not the same object as a workflow described in a chat.

User: Must integrate later with the Capability Foundry source registry rather than duplicating it. Shared infrastructure opportunity.

Sources mentioned:
https://arxiv.org
https://en.wikipedia.org/wiki/Zettelkasten
https://www.zotero.org

Unresolved: how to handle paywalled PDFs without violating licenses.

We will build a local-first librarian with Markdown export and a citation graph. Integrate Sovi private corpus later. Do not fake publisher access.