[Feature Request] Local docs sync and numerous dx improvements

[pzoltowski]

This is great repo but I wish this had some improvements:

1. Local llms.txt sync

Right now it fetches directly to context. I think it would be better that context7 was a combination of MCP / CLI / VSCode plugin. The goal would be to be a similar docs repository as dash app (https://kapeli.com/dash) but for AI. The flow would be like this: a) in VSCode Plugin or CLI or MCP function we would query e.g. for 'expo' or provide url 'https://expo.dev/', plugin would check if there is official llms.txt in official web page. If there is it would download either to global local docs repository (and symlink it in current project folder) or download to project folder (user can choose). It official is not available then we would use 3rd party or generate one from official docs page. b) search results should show some badge if 'llms.txt' is official one / llms.txt generated from official web page / 3rd party - I have more trust in official one since those will be more likely maintained and up-to-date and more secure c) plugin would show list of installed docs and badges (official, 3rd party, genefated), when last time synced. There would be a button to allow resync manually. Maybe even some checks and icon to showing if there is potentially new uploaded llms.txt in given official website (by checking document upload date and comparing to one from last sync or by checking hash)

I think there are few benefits of such solution:

• official docs up-to-date, maintained and more secure
• if we have docs in project folder windsurf / cursor can index it
• we don't have to each time download docs since those will be already stored locally in global folder and need to just symlink or copy
• in cursor / windsurf rules we able to directly link to those llms.txt e.g. "in our techstack we are using use `vercel ai sdk' [vercel-ai-sdk.llms.txt]
• some llms.txt are very heavy - if we have it indexed locally by any agent we can reduce cost of inference (e.g. when using cline/roocode) or speedup agent because don't have pollute context with 0.5M tokens (e.g. tailwindcss)

2. Some tweak for current website: a) right now search is not very accurate: when looking for 'rerun' it list so many but not this popular library that already has llms.txt: https://rerun.io/llms.txt b) the table on website is too small in width so when searching for tailwindcss it find few options but hard to see full REPO url. Would be also good to have some badges here: official or 3rd party or generated. And project thumbnail or github thumbnail