Search local files with AI
Answer a focused question from a directory of documents, spreadsheets and PDFs, with source locations you can check for yourself.
You can search local files with AI when a reliable answer depends on locating the right page or spreadsheet row. The kb-retriever skill starts from the directory’s own index files and narrows the search by topic. It reads short extracts around likely matches instead of putting an entire manual or workbook into context.
Suppose an events organiser wants to know whether a venue deposit changed after a particular date. The updated figure may be in a spreadsheet, while the exception for existing bookings is in a plain-text policy. Ask for each finding with its source. That makes it easier to spot a superseded figure or check whether two documents really refer to the same booking period.
What it does
- Locate relevant files through nested directory indexes.
- Search for a term and inspect the surrounding passage.
- Read selected sheets and rows from a large workbook.
- Find a page or figure in an authorised PDF.
- Report conflicting details with their original locations.
Install and use
- Give the agent read access to an authorised local directory with its data_structure.md indexes. Use the appropriate PDF or spreadsheet reader; the bundled page-image helper needs Python, pdf2image and Poppler.
- Ask, “Check when the venue deposit changed and tell me whether existing bookings were exempt. Show where each fact came from.”
- Open the cited locations and narrow the question if the source material is inconclusive.
FAQ
Source
Made by ConardLi. Original skill: kb-retriever, published on GitHub under the MIT licence.
View on GitHub