I have released a new version of simplifiai at https://simplifai.tiddlyhost.com. I have updated the gemini models and it now supports claude. Support for claude is ‘experimental’ as I do not use it much. I have also made some minor additions to the interface.
1 Like
I have no idea but chatgpt says this:
- Make sure you’re signed into the correct Google account (personal vs work/school). Sign-in page: https://aistudio.google.com/app/apikey. ([AI Studio][1])
- If this is a work/school (Google Workspace) account, your Workspace admin may have disabled AI Studio or not granted access — contact them to enable Gemini/AI Studio for your org. (Workspace accounts often need admin enablement.) ([Google AI for Developers][2])
- Confirm region & eligibility: AI Studio / Gemini API is only available in supported regions and some account types; if your country/region isn’t supported you’ll be blocked. Also some features/API usage differ by region. ([Google AI for Developers][3])
- Check age / account type rules: work accounts typically require the user to be 18+; personal accounts have different minimums. If Google thinks the account is underage it can block access. If that’s the case you may need to verify your age via Google account settings or Family Link. ([Google Help][4])
- Quick local fixes that often help: sign out and sign back in, open an Incognito/private window, try a different browser or device, clear cookies, or do a hard refresh (Ctrl+F5). These have resolved “permission denied” flashes for other users. ([Reddit][5])
Ah. That may be it. I have a Workspace Account & I do keep it locked-down to prevent workers from sallying-forth. But the Workspace admin doesn’t mention the specific setting I’d need change.
Laters.
Since I can’t demo it without a key, could you explain what the advantage of wrapping gemini in TW is? Does it create tags? Or draw info from your local tiddler set?
Thanks!
1 Like
For the Ai side:
simplifai interface allow fine control over the AI, you can define mulitple ‘personality’ (system prompt) in tiddlers and switch between them with a click, similarly AI models are named in tiddlers which means they can also be switched between with a click
The latest thing in AI is called Context engineering (which at it base is loading content at the begin of a chat, ie into its working memory). You can:
1)tell the AI to read tiddlers so that it has their contents are in working memory at the beginning of a chat
2)each chat is a tiddler, and the simplifai interface allows you to create new chats (tiddlers) from parts of an existing chat - this is really import as the AI looses focus with long chats, - at some point you can ask it to summarize a chat and then with a click create a new chat with the summary at the beginning. You go back to a previous point in a chat acd restart a chat from there (like in chatgpt)
3)As chats are tiddlers you can tag them etc.
From the tw side:
The AI can read and write tiddlers and load tiddlers in the river. The AI can create/edit wikitext, Javascript for widgets.
Security with AI interaction: you can control how much of your content is expose to the AI, the AI can only read the tiddlers you give it access to (this can be controlled by a filter in the simpifai config tiddlers). I think this is much better idea than using MCP to read the files on you computer. As tw is in the browser it is in a ‘walled garden’ - browser producers have spent decades ensuring that webpages can not access your file system.
1 Like
Thank you for the explainer! I suggest you paste this into your opening tiddler.
Admittedly, my use of AI has been mainly using CoPilot to answer questions about Dutch grammar, and a fallback search engine when the regular search engines only give me the SEO top hits.
1 Like