How often do plugins require server side node TiddlyWiki?

#1

In my continued series on the topic of MWS, we now come to the question of whether or not plugins generally require the server side instance or whether they are content to run on a client.

I mean obviously a lot of plugins are probably intended to work with single file wikis, so they won’t be affected, but MWS doesn’t load plugins on the server side the way Node TiddlyWiki does.

What plugins can I find in the wild which rely on running server-side JavaScript, or perhaps just set server-side filters for the client?

MWS is still going to support some form of server side plugins, we just won’t be loading the same plugins on both the server and the client, which in my opinion is a big improvement. So I’m trying to figure out what kind of service I’d plugins are currently out there so I can make sure there’s a solid path to making them compatible.

#2

Usually not. Even I use nodejs wiki (tidgi desktop), but I keep most of my plugins works on client side, so they also works on mobile (tidgi mobile app).

And due to no standard API to call server side task, it also requires set up new HTTP route to run server side feature. But TidGi as a secure tiddlywiki app, use IPC for client-server communication, so I don’t like to setup new HTTP route, so I’ve never done that.

I believe I’m a very typical power user of tw, I know how to put task to server side, but for many reasons I’ve never done that.

But I’m writing my personal AI assistant on TidGi tiddlywiki app, I may directly run filter on nodejs-side to RAG for user note, and I hope I could integrate MWS in TidGi app. So I care about what specifically do you mean here:

#3

I think that’s sensible. I personally think it is not needed because of security concerns.

#4

None of my plugins need to run on the server side. However some AI services I would like to use are blocked by cors, so I may end up doing something on the server side.

#5

CORS can easily be a setting, so it doesn’t need to be done with plugins. Good point.

#6

Basically what I mean by that is if you have plugins that set filters on the server which control what tiddlers are sent to the client.

#7

Which AI services are those that are getting blocked and you can’t override? :slight_smile:

#8

anthropic is blocking me when using firefox and chrome, I have to disable cors completely when booting up the browser to get claude to work (I tired writing an extension for chrome, to modify cors, but that failed) . I use linuxmint 22.1, I have not tried using any other OS.

#9

I had the same problem initially. If you look in the connection-pool.js file of WikiSage, you’ll see how I’ve overcome that by adding a special header:

            async fetch(url, options) {
                const maxRetries = this.config?.maxRetries || 3;
                const retryDelay = this.config?.retryDelay || 1000;
            
                // Determine if this is an Anthropic API call
                const isAnthropicAPI = url.includes('api.anthropic.com');
            
                for (let attempt = 1; attempt <= maxRetries; attempt++) {
                    try {
                        if (isAnthropicAPI) {
                            // Ensure headers are properly set for Anthropic API
                            options.headers = {
                                ...options.headers,
                                "anthropic-version": "2023-06-01",
                                "content-type": "application/json",
                                "x-api-key": options.headers["x-api-key"],
																"anthropic-beta": "pdfs-2024-09-25",
                                "anthropic-dangerous-direct-browser-access": "true"
#10

thanks that is working. :slightly_smiling_face: