Skip to content

Ideas collective #1

Description

@sigman78

Its a good start but there is so much ideas how to grow

Search providers

Each search provider could have own alignment by government (google, yandex, baidu etc) - makes sense give a choice.

Specialized search engines: reddit, Wikipedia, web-archive, youtube transcript

Richer search query lang (file type, ignore kw, require kw)

Richer metadata per search result if provider supports

Scrapers / extractors

Specific extractor types: pdf, images, web-page-source?

Richer metadata: relevance score?

Split: render/grab (playwright) then filter/extract (trafilatura)

Research DB

Large deep research would eventually run out of context, ideas:

  • Keep each scraped result with index like {@source-query, [kwd/tags], "title", "summary", "raw-text"}
  • RAG it
  • Perform facts 'aggregation'

Results

Allow variable research results output: full, thesis, relevant links, info-graphics


As it turns out - its a command line LMNotebook O_O

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions